AGRISAR. AGRISAR 2006 Agricultural Bio-/Geophysical Retrievals from Frequent Repeat SAR and Optical Imaging. Final Report.

Transcription

1 AGRISAR 2006 Agricultural Bio-/Geophysical Retrievals from Frequent Repeat SAR and Optical Imaging Final Report Prepared for European Space Agency Prepared by German Aerospace Center Microwaves and Radar Institute (HR) Oberpfaffenhofen and 14. Januar 2008 Version 1

2 1 CONTENTS 1 Contents Introduction Campaign Objectives Campaign Institutions Campaign Participants Description of the Study Area AGRISAR Campaign Schedule Airborne Data Acquisition Synthetic Aperture Radar Data Acquisition Mission Logistics Calibration of the E-SAR system Main Measurement Campaign Radar Data Acquisition Optical Flights Mission Logistics Calibration Flight Main Measurement Campaigns Optical Data Acquisition Satellite Data Acquisition ALOS-PALSAR ENVISAT-ASAR ENVISAT-MERIS CHRIS AATSR Atmospheric Measurements Radio sounding Sunphotometer CIMEL (direct sun irradiance) Sunphotometer CIMEL (sky radiance) Sunphotometer Microtops II Aureole Sunphotometer FUBISS /01/2008 Page 1 of 259

3 7 Ground Radiometric Measurements Solar range ground radiometric measurements Thermal infrared ground radiometric measurements Soil and Vegetation Measurements Continuous measurements Intensive Measurements Intensive Measurements Intensive Measurements Intensive Measurements Intensive Measurements Surface Energy Budget Bowen-ratio station Scintillometer (LAS) Goniometer AGRISAR Data Base Data Quality Airborne Data Quality Synthetic Aperture Radar Data Optical Data Atmospheric Data Sunphotometer CIMEL (direct sun irradiance) Sunphotometer CIMEL (sky radiance) Sunphotometer Microtops II Aureole Sunphotometer FUBISS Ground Radiometric Measurements Solar range ground radiometric data Soil and Vegetation Data Continuous data Intensive data Surface Energy Budget Bowen-ratio station Scintillometer (LAS) /01/2008 Page 2 of 259

4 Goniometer Preliminary Data Anaylsis Synthetic Aperture Radar Derivation of soil surface roughness dynamics using L-Band PolSAR data by the University of Kiel Investigation of Soil Moisture by ISSIA Decomposition of different scattering mechanisms for soil moisture estimation under the vegetation: Preliminary Analysis on the polarimetric AGRISAR data by DLR-HR Extinction Estimation by Corner Reflectors Located on Crop Fields by University of Alicante Land cover classification by TUB LAI and Soil Water content by University of Naples Optics Analysis of Chlorophyll by University of Valencia Using CASI for LAI estimation in Sentinel-2 configuration by University of Naples Water and Energy Budget Analysis by ITC Land Processes Integration of RS-Derived Information into Hydrological Models by LHWM Large Aperture Scintillometers (LAS) by ITC Assimilation of surface soil moisture information into land surface models to compensate for uncertainties in precipitation information by TUM Summary and Conclusion Appendix References /01/2008 Page 3 of 259

5 2 INTRODUCTION This document aims to describe the AGRISAR campaign carried out from the 18 th of April to the 2 nd of August In the following the main objectives of the AGRISAR campaign are summarised and a description of the test site, the acquired airborne data, atmospheric and ground data are given as well as the quality of the acquired data and the preliminary analysis are presented. In the frame of its Earth Observation Envelope Programme of the European Space Agency (ESA), AGRISAR aims to support geophysical algorithm development, calibration/validation and the simulation of future space borne Earth Observation missions. The next generation of ESA Earth Observation satellites include a series of Sentinel Missions to be developed and operated within the framework of GMES (Global Monitoring for Environment and Security). These will include SAR and Optical satellites with new imaging configurations and spectral bands, and much improved capabilities for frequent repeat coverage. The AGRISAR 2006 campaign aimed to collect in-situ, airborne SAR and optical measurements, as well as Satellite data in support of decisions being taken on satellite instrument configurations for the first Sentinel Missions. In addition it aims to provide an important database for the study of longer term mission concepts. 1.1 Campaign Objectives The AGRISAR 2006 campaign was established to address important specific programmatic needs of Sentinel-1 and -2: To assess the impact of Sentinel-1 and Sentinel-2 sensor and mission characteristics for land applications (land use mapping, parameter retrieval). To provide a basis for the quantitative assessment of sensor or mission trade-off studies, e.g. spatial and radiometric resolution, revisit time Simulate Sentinel-1 and Sentinel-2 image products over the land In the context of Sentinel-1, AGRISAR 2006 is aimed primarily at the investigation of radar signatures throughout the crop growing season at time intervals of 7/10 days which are consistent with the mission concept. An important dataset of coordinated in-situ and airborne SAR measurements were collected which will provide support both to studies of the Sentinel-1 technical concept, as well as contributing to studies of future mission concepts involving parameter retrieval at C- and L-band. As part of the refinement and verification of the Sentinel-1 technical concept, AGRISAR data will be used for the assessment of land use classification using the proposed nominal operating configuration (i.e. IW mode, VV + HH polarisation plus co-polarisation, weekly revisit). Simulation of Sentinel-1 image products is planned. By including an optical data acquisition component, the campaign also aims to provide feedback on key issues relating to definition of the ESA Sentinel-2 multi-spectral mission requirements. 18/01/2008 Page 4 of 259

6 Attention focuses on the investigation of the optimum position and width of spectral bands for land cover/change classification and retrieval of bio/geophysical parameters (e.g. improved surface classification, quantitative assessment of vegetation status at different crop growth stages). The imaging spectrometer data acquired as part of AGRISAR 2006 will be used to simulate Sentinel-2 L1b products using the different proposed configurations, and to investigate compatibility with the envisaged L2/L3 products. In addition space borne data are used to support the algorithm retrieval and test the performance of available satellite. The main space borne data included are: ENVISAT/ASAR, ALOS/PALSAR, CHRIS/PROBA, ENVISAT/MERIS, MODIS. Current ESA studies being carried out which are expected to benefit from the availability of AGRISAR data include those being carried out in two parallel contracts started in February 2006 to address the topic of Exploiting Longer wavelength SAR data for the Improvement of Surface process Modelling. 1.2 Campaign Institutions AGRISAR involved in total 60 people from 16 different institutes coming from eight different countries, with around 125 people participating during the intense measurements periods. Participants included German teams from DLR-HR, DLR-FB, DLR-DFD, ZALF, IG-Demmin and Universities in Munich, Berlin, Kiel and Jena along with Spanish teams from University of Valencia, University of Alicante and INTA, Italian teams from University of Naples and the National Research Council (ISSIA), teams from Canadian ITRES, from Technical University of Denmark, from University of Ghent in Belgium, from University of Canfield in UK and from the International Institute for Geo-Information Science and Earth Observation (ITC) in the Netherlands, along with participants from ESA. AIRBORNE TEAM: German Aerospace Center (DLR), Microwaves and Radar Institute (HR) P.O. Box 11 16, D Wessling, Germany in the following referred to as DLR-HR German Aerospace Center (DLR), Flight Operations (FB) P.O. Box 11 16, D Wessling, Germany in the following referred to as DLR-FB ITRES Research Limited Suite 110, ST Street NW Calgary, Alberta, Canada T2L 2K7 in the following referred to as ITRES Instituto Nacional de Técnica Aeroespacial (INTA) Área de Teledetección, Dpto.de Observ.de la Tierra, Teledetección y Atmósfera 18/01/2008 Page 5 of 259

7 Carretera de Ajalvir, p.k Torrejón de Ardoz (Madrid) in the following referred to as INTA Figure 2.1: Airborne team at the third intensive campaign Figure 2.2: AGRISAR ground team at the second intensive campaign (a part of the ground team) 18/01/2008 Page 6 of 259

8 GROUND TEAM: German Aerospace Center (DLR), German Remote Sensing Data Center (DFD) Kalkhorstweg 53, Neustrelitz, Germany in the following referred to as DLR-DFD Technical University of Denmark, Ørsted DTU, Section for EM Systems Ørsteds Plads 348, DK-2800 Kgs. Lyngby, Denmark in the following referred as DTU Friedrich-Schiller-University Jena Dep. of Earth Observation Loebdergraben 32, D Jena in the following referred as FSU Free University Berlin Institute for Space Sciences Carl-Henrich-Becker-Weg 6-10 D Berlin in the following referred as FUB GEO-Informatik Buchholz Alt Tellin in the following referred as GEO-INF Istituto di Studi sui Sistemi Intelligenti perl'automazione (ISSIA) Consiglio Nazionale delle Ricerche (CNR) Via Amendola 122/D, I-70126, Bari, Italy in the following referred to as ISSIA International Institute for Geo-Information Science and Earth Observation P.O. Box 6, 7500 AA Enschede, The Netherlands in the following referred as ITC Laboratory of Hydrology and Water Management Ghent University, Coupure link 653 B-9000 Ghent, Belgium in the following referred as LHWM Ludwig Maximilians Universität München (University of Munich) Department of Earth and Environmental Sciences Geography and Remote Sensing Luisenstr. 37, Munich (Germany) in the following referred to as LMU 18/01/2008 Page 7 of 259

9 University of Alicante DFISTS, EPS, Signals, Systems and Telecommunications Group (SST) P.O. Box 99 E Alicante, Spain in the following referred as Uni Alicante University of Kiel (Christian-Albrechts-Universität), Department of Geography Ludewig-Meyn-Strasse 14, D Kiel, Germany in the following referred to as Uni Kiel University of Naples Frederico II Dept. Agricultural Engineering and Agronomy University of Naples "Federico II" Via Università 100, I Portici (Naples) ITALY in the following referred as Uni Naples Universidad de Valencia Facultad de Fisica, Departamento de Termodinamica Remote Sensing Unit Laboratory of Earth Observation (LEO) Solar Radiation Unit c/dr. Moliner, 50, E Burjassot, Valencia, Spain in the following referred as Uni Valencia Leibnitz-Zentrum für Agrarlandschaftsforschung (ZALF) Institut für Bodenlandschaftsforschung Eberswalder Str. 84 D Müncheberg in the following referred as ZALF 18/01/2008 Page 8 of 259

10 2.1 Campaign Participants ORGANISATION Team: ESA RSAC DLR-HR DLR-DFD Remo Bianchi Malcolm Davidson Mike Wooding Irena Hajnsek Carolin Wloyczyk AIRBORNE Team: DLR-FB DLR-HR ITRES INTA Andrea Hausold Philip Weber Stefan Grillenbeck Klaus Dietl Roland Welser Michael Grossrubatscher Emil Sauer Helmut Kirner Siegfried Judt Roman Koch Frank Schirmer Ralf Horn Bernd Gabler Peter Hackenberg Martin Keller Steve Math Jason Howse Alix Fernandez-Renau Jose Antonio Gomez Eduardo de Miguel 18/01/2008 Page 9 of 259

11 GROUND/ATMOSPHERE Team: DLR-DFD DLR-HR DTU DWD FSU FUB GEO-Info Carolin Wloczyk Erik Borg Heike Gerighausen Adolf Günther Bernd Fichtelmann Hans-Hermann Vajen Matthias Rosenberg Manuel Schulz Martin Becvar Irena Hajnsek Martin Keller Rolf Scheiber Christian Andres Henning Skriver Irina Nischan Peter Schierbaum Christiane Schmullius Christian Thiel Tania Riedel Martin Herold Sören Hese Jacqueline Sambale Enrico Stein Kathrin Woellner Robert Lux Johannes Reiche Marcel Urban Martin Lindner Corina Manusch Susan Hanisch Michael Schulz Jürgen Fischer Thomas Ruhtz Paul Zieger Stefan Stapelberg Edgar Zabel 18/01/2008 Page 10 of 259

12 Görmin Farm ISSIA ITC LHWM LMU Uni Alicante Uni Kiel Uni Napoli Uni Valencia - Global Change Unit Karsten Trunk Peter Paschen Francesco Mattia Giuseppe Satalino Laura Dente Bob Su Wim Timmermans Ard Blenke Remco Dost Joris Timmermans Kitsiri Weligepolage Valentijn Pawels Gabrielle De Lannoy Davy Loete Alex Löw Ingo Keding Michael Robert Natalie Ohl Thomas Wanderer Claudia Hundseder Juan Manuel Lopez-Sanchez Josep David Balester-Berman Ralf Ludwig Karsten Krüger Philip Marzahn Freya Hensgens Janina Marx Holger Zebner Swen Meyer Torben Gerdes Guido D Urso Mario Palladino José Antonio Sobrino Juan Carlos Jimenez Guillem Soria 18/01/2008 Page 11 of 259

13 Juan Cuenca Maria Malena Zaragoza Mireia Romaguera Mónica Gómez Yves Julien Anäis Barella Mariam Atitar Uni Valencia - Remote Sensing Unit Uni Valencia - Solar Radiation Unit ZALF Jose F. Moreno Luis Alonso-Chordá Gloria Fernandez Soledad Gandia Jordi Garcia Jose A. Martinez-Lozano Maria P. Utrillas Victor Estellés Michael Sommer Marc Wehrhan Gernot Verch K. Hantel 18/01/2008 Page 12 of 259

14 3 DESCRIPTION OF THE STUDY AREA DEMMIN (Durable Environmental Multidisciplinary Monitoring Information Network) is a consolidated test site located in Mecklenburg-Western Pomerania in North-East Germany, approximately 60 km north of Neustrelitz and 150 km north of Berlin (Figure 3.1). DLR s German Remote Sensing Data Center (DFD) is cooperating with the local farmers to operate the longterm test site DEMMIN. This test site is well established, since it has been used regularly since 1999 as a test site for the simultaneous collection of airborne and in situ data. The altitudinal range within the test site is around 50 m. The main crops grown are winter wheat, winter barley, winter rape, maize and sugar beet. The DEMMIN site comprises of four large-area farms combined within a farming association (the so-called IG-Demmin ) managing approximately ha (Figure 3.2). Single land parcels are very large in this area, in average 80 ha. The Görmin farm forms the north-eastern part of the DEMMIN site. The main attention of the huge area is paid to the Görmin farm in the northeastern part of the test site. Figure 3.4 shows the 2006 cropping for the Görmin farm. Sowing and harvest dates for the main crops are typical of those for northern Europe: Crop Sowing Date Harvest Date Winter Wheat 05.September 31.October August Winter Barley September July Winter Rape 08. August- 5.September 28.July-10.August Maize (for silage) 20.April-05.May 25.September 10.October Sugar Beet 25.March 20.April 25. September-31.October The main geographical coordinates in the UTM Zone 33, Date WGS84 of the Görmin farm location is Görmin Farm X-Coordinate Y-Coordinate In front of the main office This farmer uses advanced precision farming technologies and is open for the investigation of new methods, as an example he uses satellite navigation, the N-Sensor instrument and automated yield mapping. There are a large number of data bases available from the previous research carried out in the test site. These include digital quasi-static-data (GIS data layers), data derived from precision farming, and data from previous campaigns and monitoring activities. The common reference system to be used for data storage and distribution is UTM WGS84. 18/01/2008 Page 13 of 259

15 AGRISAR Figure 3.1: Location of the DEMMIN test site in the German federal state Mecklenburg-Western Pomerania Figure 3.2: Map of farms within the Demmin test site 18/01/2008 Page 14 of 259

16 Digital quasi-static data Geological map Soil map Hydrological map Agricultural field map Digital Elevation Model Digital dynamic data Annual yield maps (Combine measurements) Annual nitrogen-sensor measurements (Track measurements) Annual application maps Macro and micro nutrients Measurements of vegetation stages Campaign data I (In-situ data) Destructive measurements for determination of leaf area and leaf area index (since 2004) Biomass measures (dry and wet), Nitrogen measures (SPAD), Photographical documentation for measuring crop height and crop density Spectrometric measures on ground Campaign data II (airborne data) Annual hyperspectral flying campaigns (e.g. HyMap), Simultaneous ground measurement program for data validation. Precision farming data are available for the AGRISAR team, including yield and nitrogen application maps, collected by farmers in the 2006 growing season. There are two agro-meteorological stations, one at the Görmin test site (station Görmin) and one near by (station Kletzin). Both have the following atmospheric and soil science instruments: Pyranometer (up- and downwelling short-wave radiation) Pyrgeometer (up- and downwelling long-wave radiation) Relative air moisture Air temperature Leaf wetness Electronic rain sensor Wind direction and speed (in 2 m height) Soil moisture (C-Probes) at 10, 30 and 90 cm depth Soil temperature at 0, 5, 15, 20 and 50 cm depth The sampling rate of the system is programmable (currently 15 minutes for all parameters). The measured agrarian meteorological and soil data are transmitted automatically to a receiving station and a data server. The Görmin area is a very flat area with very small topographic variations. The highest topographic variation is between 0 m at the Penne River to 40 m 37 m MSL in the middle of the corn field 222. In principal there is a slight decrease of topography from North to South in direction of the river Peene. A digital elevation model for this area is available from the SRTM mission derived from an X-band single pass interferometer. 18/01/2008 Page 15 of 259

17 Also a soil map of this area is available (Figure 3.3). The main soil texture that exists at the site is IS and SL which is a variation between loamy sand to strong loamy sand. With respect to the soil texture the continuous sample points have been selected as described in the Experimental Plan. 18/01/2008 Page 16 of 259

18 Figure 3.3: Soil map of the Görmin test site. 18/01/2008 Page 17 of 259

19 Figure 3.4: 2006 Cropping for Görmin farm 18/01/2008 Page 18 of 259

20 3.1 AGRISAR Campaign Schedule The contribution from each AGRISAR team partner is displayed in the time schedule plan of the AGRISAR campaign in the following table: Topic Month Mai June July August Days Field Condition Bare fields X X Main plant growth X X X X X X X X X X Harvest X X X X IN-SITU DLR-DFD/ ZALF/GEO-INF X X X X X X X X X X X X Uni of Kiel X X X X X X X X X X X Uni of Valencia X X X Uni Naples X X Uni. Berlin X X Uni. Alicante X X ITC X ISSIA X X Uni Ghent X X X TUD X X LMU X X X FSU X X AIRBORNE DLR-HR/FB (E-SAR) 3x X X X X 2x X X 2x X X X INTA/ITRES (AHS/CASI) X X 18/01/2008 Page 19 of 259

21 The collection of airborne and in-situ data took place over a whole vegetation period, starting in April, where still some fields have been bare and ending in August with the start of the harvesting. Radar data collection has been done closely weekly in coordination with a weekly in-situ measurements campaign. In total three intensive ground measurements period have been identified where intensive ground measurements have been collected from teams of different institution providing several instruments. During the intensive ground measurements period twice two optical systems have been operated in order to collect hyperspectral and thermal data. 18/01/2008 Page 20 of 259

22 4 AIRBORNE DATA ACQUISITION Three airborne sensors have been operated during the AGRISAR flight campaign to acquire valuable data for bio-/geo-physical parameter estimation. Two common flight directions have been chosen in order to acquire the most important landscape features. The most important one for the hydrological observation of the site is the N-S flight track that covers the watershed between two small river systems. This track has been flown at the three main intensive campaigns. The main radar operation flight track throughout the vegetation period was the E-W flight track as it represents the most appropriate direction with respect to the main wind direction. The two flight tracks are plotted (red rectangles) in the topographic map, where in blue the borders of the agricultural fields are presented (Figure 4.1). The flight tracks are around 10 km long and 3 km wide. Both flight tracks have been covered with data from the three airborne systems. Figure 4.1 Flight tracks coverage of the Demmin test site (Görmin farm) for the continuous radar flight. The two possible tracks are shown in red. The extent of the farm is shown in blue. 18/01/2008 Page 21 of 259

23 4.1 Synthetic Aperture Radar Data Acquisition Mission Logistics Before the AGRISAR campaign start the E-SAR system was tested. The direct flight to the test site has been performed as defined in the mission planning that is described in the Experimental Plan. The ferry flight of the Do228 took approximately 2.5 h starting in Oberpfaffenhofen and landing at the airport Neubrandenburg in Germany. They performed the data acquisition over the E-W track and flow back to Oberpfaffenhofen the same day. During the intensive ground measurements campaign the crew and the radar operator stayed at the airport Neubrandenburg in order to fly also the second track (N-S) Calibration of the E-SAR system In order to secure high quality E-SAR data, recording of good quality kinematic phase differential GPS measurements a GPS monitoring station has been installed at the airport. Static DGPS surveying using data of a permanent GPS station at a range of less than 50 km has been performed to determine its geographical position with a relative accuracy of less than 5 cm. The chosen test sites were located not too far from the airport, so that only few reference points needed to be set-up. Six corner reflectors were set up in the test site for digital terrain elevation generation over the whole Görmin site. From the six four has been kept fixed on the site during the whole data acquisition period. Mr. Zabel from Geo-Inf has checked the position of the corner reflector before each flight. 18/01/2008 Page 22 of 259

24 AGRISAR Figure 4.2 Flight map of the Görmin test site Figure 4.3: Corner reflector deployment in the overlapping area of the two flight tracks Main Measurement Campaign A total of 16 radar flights were executed in the period of and over the Görmin test site in northern Germany. All data as already described in the Experimental Plan were able 18/01/2008 Page 23 of 259

25 to be collected, without any delays or technical problems. The data were recorded on HHDT Tapes and are after the data acquisition transcribed digital to a hard disk at the DLR site. After transcription the radar data processing started, with an integrated radar data quality check. The following modes were collected for the 16 flight campaigns: Frequency Polarisation Passes Remarks X-band DEM VV 1 1.7m (single pass baseline) X-band HH / VV 2 Coregistered C-band Dual/Quad 2 Coregistered L-band quad 1 Coregistered Table 4.1: Acquisition modes of E-SAR system Whereas the X-band DEM has been only once acquired for both flight tracks at the beginning of the AGRISAR campaign (Figure 4.4). 18/01/2008 Page 24 of 259

26 Figure 4.4: X-band DEM processing and data acquisition of the Görmin farm, where the red box represents the border of the DEM. 18/01/2008 Page 25 of 259

27 4.1.4 Radar Data Acquisition In the following the acquired radar data are listed. The X-band single pass interferometry mode has been used to perform a digital elevation map covering not only the two flight stripes but the whole area as displayed in Figure 4.4 Scene-ID Location Radar Mode Freq.-Band Polarisation Track-ID Baseline Heading 1-st Date 18. Apr 06 06agrsar0101x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0102x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0103x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0104x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0105x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0106x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0107x1 Goermin 2-Kanal,HR-NS X-XTI VV agrsar0108x1 Goermin 2-Kanal,HR-NS X-XTI VV nd Date 19. Apr 06 06agrsar0201x1 Goermin1 1-Kanal,HR-NS X VV agrsar0203x1 Goermin1 1-Kanal,HR-NS X HH agrsar0205x1 Goermin1 2-Kanal,HR-NS C DP agrsar0206x1 Goermin1 2-Kanal,HR-NS C DP agrsar0208x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM rd Date 20. Apr 06 06agrsar0301x1 Goermin2 1-Kanal,HR-NS X VV agrsar0303x1 Goermin2 1-Kanal,HR-NS X HH agrsar0304x1 Goermin2 2-Kanal,HR-NS C DP agrsar0305x1 Goermin2 2-Kanal,HR-NS C DP /01/2008 Page 26 of 259

28 06agrsar0308x1 Goermin2 4-Kanal,HR- L PM 2 0 NS,PM th Date 03. Mai 06 06agrsar0401x1 Goermin1 1-Kanal,HR-NS X VV agrsar0403x1 Goermin1 1-Kanal,HR-NS X HH agrsar0405x1 Goermin1 2-Kanal,HR-NS C DP agrsar0407x1 Goermin1 2-Kanal,HR-NS C DP agrsar0409x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 11. Mai 06 06agrsar0501x1 Goermin1 1-Kanal,HR-NS X VV agrsar0503x1 Goermin1 1-Kanal,HR-NS X HH agrsar0505x1 Goermin1 2-Kanal,HR-NS C DP agrsar0507x1 Goermin1 2-Kanal,HR-NS C DP agrsar0509x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 16. Mai 06 06agrsar0601x1 Goermin1 1-Kanal,HR-NS X VV agrsar0603x1 Goermin1 1-Kanal,HR-NS X HH agrsar0605x1 Goermin1 2-Kanal,HR-NS C DP agrsar0607x1 Goermin1 2-Kanal,HR-NS C DP agrsar0609x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 24. Mai 06 06agrsar0701x1 Goermin1 1-Kanal,HR-NS X VV agrsar0703x1 Goermin1 1-Kanal,HR-NS X VV agrsar0704x1 Goermin1 1-Kanal,HR-NS X HH /01/2008 Page 27 of 259

29 06agrsar0706x1 Goermin1 2-Kanal,HR-NS C DP agrsar0708x1 Goermin1 2-Kanal,HR-NS C DP agrsar0710x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 06. Jun 06 06agrsar0801x1 Goermin1 1-Kanal,HR-NS X VV agrsar0803x1 Goermin1 1-Kanal,HR-NS X HH agrsar0805x1 Goermin1 2-Kanal,HR-NS C DP agrsar0807x1 Goermin1 2-Kanal,HR-NS C DP agrsar0809x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 07. Jun 06 06agrsar0901x1 Goermin1 1-Kanal,HR-NS X VV agrsar0902x1 Goermin1 1-Kanal,HR-NS X VV agrsar0904x1 Goermin1 1-Kanal,HR-NS X HH agrsar0906x1 Goermin1 2-Kanal,HR-NS C DP agrsar0910x1 Goermin1 2-Kanal,HR-NS C DP agrsar0912x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 13. Jun 06 06agrsar1001x1 Goermin1 1-Kanal,HR-NS X VV agrsar1003x1 Goermin1 1-Kanal,HR-NS X HH agrsar1005x1 Goermin1 2-Kanal,HR-NS C DP agrsar1007x1 Goermin1 2-Kanal,HR-NS C DP agrsar1010x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 21. Jun 06 18/01/2008 Page 28 of 259

30 06agrsar1101x1 Goermin1 1-Kanal,HR-NS X VV agrsar1103x1 Goermin1 1-Kanal,HR-NS X HH agrsar1105x1 Goermin1 2-Kanal,HR-NS C DP agrsar1107x1 Goermin1 2-Kanal,HR-NS C DP agrsar1109x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 05. Jul 06 06agrsar1201x1 Goermin1 1-Kanal,HR-NS X VV agrsar1203x1 Goermin1 1-Kanal,HR-NS X HH agrsar1205x1 Goermin1 2-Kanal,HR-NS C DP agrsar1207x1 Goermin1 2-Kanal,HR-NS C DP agrsar1208x1 Goermin1 2-Kanal,HR-NS C DP agrsar1210x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 06. Jul 06 06agrsar1301x1 Goermin2 1-Kanal,HR-NS X VV agrsar1303x1 Goermin2 1-Kanal,HR-NS X HH agrsar1305x1 Goermin2 2-Kanal,HR-NS C DP agrsar1307x1 Goermin2 2-Kanal,HR-NS C DP agrsar1309x1 Goermin2 4-Kanal,HR- L PM 2 0 NS,PM th Date 12. Jul 06 06agrsar1401x1 Goermin1 1-Kanal,HR-NS X VV agrsar1403x1 Goermin1 1-Kanal,HR-NS X HH agrsar1405x1 Goermin1 2-Kanal,HR-NS C DP agrsar1407x1 Goermin1 2-Kanal,HR-NS C DP agrsar1409x1 Goermin1 4-Kanal,HR- L PM /01/2008 Page 29 of 259

31 NS,PM 15-th Date 26. Jul 06 06agrsar1501x1 Goermin1 1-Kanal,HR-NS X VV agrsar1503x1 Goermin1 1-Kanal,HR-NS X HH agrsar1505x1 Goermin1 2-Kanal,HR-NS C DP agrsar1507x1 Goermin1 2-Kanal,HR-NS C DP agrsar1509x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM th Date 02. Aug 06 06agrsar1601x1 Goermin1 1-Kanal,HR-NS X VV agrsar1603x1 Goermin1 1-Kanal,HR-NS X HH agrsar1605x1 Goermin1 2-Kanal,HR-NS C DP agrsar1607x1 Goermin1 2-Kanal,HR-NS C DP agrsar1609x1 Goermin1 4-Kanal,HR- L PM 2 0 NS,PM 270 Table 4.2: Radar data acquired by E-SAR system 18/01/2008 Page 30 of 259

32 4.2 Optical Flights Two optical systems, the INTA S AHS and the ITRES CASI system, have been operated by the two teams and have been flown on the CASA aircraft from the Spanish Air force. The systems have been already described in the Experimental Plan of the AGRISAR campaign Mission Logistics Both optical systems have been installed in the CASA aircraft already before the ferry flight to Neubrandenburg in Spain. The CASI system has been shipped by ITRES to Madrid, where the INTA team together with the ITRES team performed the installation in two days. A test flight has been performed already in Spain. The ferry flight took one day to Neubrandenburg, which was the main base for the measurement flights. The team stayed every time one week in order to wait for the best weather condition. The team were on standby everyday ready to undertake a measurement flight. Together with a meteorological team in Spain and Berlin a weather forecast for every day has been checked. Finally, a team at the test site gave some recommendation for a go or no go for an optical flight Calibration Flight The bundle adjustment flight for the AHS and CASI system has been performed over the City Neubrandenburg in Northern Germany. Precise allocated and marked signs on the street have been made and provided by the University of Neubrandenburg. They provided the DGPS coordinates of the sings on the ground, which are used for the adjustment of the optical image (Figure 4.5). Figure 4.5: Precise marked signs for the bundle adjustment flights (NB 1019) of the CASI system (Photo from the University of Neubrandenburg). 18/01/2008 Page 31 of 259

33 The following flight lines have been performed during one flight over the City Neubrandenburg on 6 th of July 2006 with 72 spectral bands of the CASI system and 80 spectral bands of AHS spectral configuration (VNIR, SWIR, MWIR & LWIR). Flight Line Starting Point Ending Point Easting Northings Zones Easting Northings Zones CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U CASI-BAF P U U Table 4.3: Optical data acquired by CASI system Figure 4.6: AgriSAR 2006 Mission 2. CASI-1500 bundle adjustment flight over Neubrandenburg; Flight line pattern for CASI-1500 GSD 0.4m x 1.32m survey (lines in blue color). Ground target points in yellow. (WGS84/ETRS89 E1: topographic map provided by DLR) 18/01/2008 Page 32 of 259

34 4.2.3 Main Measurement Campaigns The main measurements campaign has been during two intensive ground campaigns performed in the beginning of June and July On both date the condition for an optical data acquisition has been good and the data quality seen from the quick-looks are very high. At the first campaign the first flight day has been nearly perfect for an optical flight. The second flight day needed to be stopped during the flight, as too many clouds were disturbing the data acquisition. After the two days of waiting for better weather condition a second flight could be performed with the AHS sensor. The CASI sensor had some technical problems and could not perform a acquisition flight. At the second flight campaign the weather and the technical condition have been perfect and all planned data could be acquired. Figure 4.7: CASA at Neubrandenburg Airport 18/01/2008 Page 33 of 259

35 4.2.4 Optical Data Acquisition During the first and second optical flight campaign the main flight days have been the 6th and 10 th June 2006 and the 4 th and 5 th of July 2006, with the following flight lines: Flight Line Starting Point Ending Point Easting Northings Zones Easting Northings Zones AHS-P U U AHS-P U U AHS-P U U AHS-P U U AHS-P U U AHS-P U U Table 4.4: UTM WGS84/ETRS89 flight lines co-ordinates for the AgriSAR June 2006 GSD 2.4m and 6.9m AHS surveys. (Agreed at the pre-campaign coordination meeting held on 05/June/2006 at Greifswald) Flight Line Starting Point Ending Point Easting Northings Zones Easting Northings Zones CASI-P U U CASI-P U U CASI-P U U CASI-P U U CASI-P U U CASI-P U U Table 4.5: UTM WGS84/ETRS89 flight lines co-ordinates for the AgriSAR June 2006 CASI-1500 survey (Agreed at the pre-campaign coordination meeting held on 05/June/2006 at Greifswald) The following optical data have been acquired in June and July 2006: DATE FLIGHT LINE FILE TRUE HEADING TIME UTC REMARKS 1st Optical Campaign 2006-JUNE-06 AHS-P : JUNE-06 AHS-P : JUNE-06 AHS-P : JUNE-06 AHS-P : JUNE-06 CASI-P :21 18/01/2008 Page 34 of 259

36 2006-JUNE-06 CASI-P : JUNE-06 CASI-P : JUNE-06 CASI-P : JUNE-06 CASI-P : JUNE-06 CASI-P : JUNE-06 CASI-P2 N/A :27 AHS image not available for this line 2006-JUNE-06 AHS-P : JUNE-06 AHS-P : JUNE-07 AHS-P : JUNE-10 AHS-P : JUNE-10 AHS-P : JUNE-10 AHS-P : JUNE-10 AHS-P : JUNE-10 AHS-P : JUNE-10 AHS-P :25 2nd Optical Flight Campaign 2006-JULY-04 AHS-P : JULY-04 AHS-P : JULY-04 AHS-P : JULY-04 AHS-P3 N/A : JULY-04 AHS-P :35 reflight 2006-JULY-04 CASI-P : JULY-04 CASI-P : JULY-04 CASI-P : JULY-04 CASI-P : JULY-04 CASI-P : JULY-04 CASI-P : JULY-04 AHS-P : JULY-04 AHS-P : JULY-05 AHS-P :18 18/01/2008 Page 35 of 259

37 2006-JULY-05 AHS-P : JULY-05 AHS-P : JULY-05 AHS-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P : JULY-05 CASI-P :30 Table 4.6: Optical data acquired in June and July 2006 Figure 4.8: AgriSAR 2006 Mission 1 Flight Campaign (05-18 June 2006) Flight line pattern for AHS and CASI-1500 survey (lines in blue color) (WGS84/ETRS89 E1: topographic map provided by DLR) 18/01/2008 Page 36 of 259

38 5 SATELLITE DATA ACQUISITION In addition to the airborne data also satellite data from existing radar and optical satellites have been collected. Small proposals have been submitted to ESA in order to acquire the satellite data. The following data are available or will be made available: 5.1 ALOS-PALSAR A proposal has been submitted and accepted by ESA. Fully polarimetric L-band data have been ordered and acquired during the commissioning phase. It is still not clear which data are available. 5.2 ENVISAT-ASAR In the following all data acquisitions have been listed that are available and can be used for further investigations: Id Product Mission Sensor Start Pass Orbit Track Swath Polarisation 1 ASA_IM Envisat ASAR/IM :36 A I2 H/H 2 ASA_IM Envisat ASAR/IM :18 D I6 H/H 3 ASA_IM Envisat ASAR/IM :35 D I2 V/V 4 ASA_IM Envisat ASAR/IM :24 D I4 V/V 5 ASA_IM Envisat ASAR/IM :35 D I2 V/V 6 ASA_IM Envisat ASAR/IM :35 D I2 V/V 7 ASA_WS Envisat ASAR/WS :39 A WS V/V 8 ASA_WS Envisat ASAR/WS :50 A WS V/V 9 ASA_WS Envisat ASAR/WS :32 D WS V/V 10 ASA_WS Envisat ASAR/WS :56 A WS V/V 11 ASA_WS Envisat ASAR/WS :17 D WS V/V 12 ASA_WS Envisat ASAR/WS :42 A WS V/V 13 ASA_WS Envisat ASAR/WS :41 D WS V/V 14 ASA_WS Envisat ASAR/WS :40 D WS V/V 15 ASA_WS Envisat ASAR/WS :15 D WS V/V 16 ASA_WS Envisat ASAR/WS :15 D WS V/V 17 ASA_WS Envisat ASAR/WS :26 D WS V/V 18 ASA_WS Envisat ASAR/WS :50 A WS V/V 19 ASA_WS Envisat ASAR/WS :56 A WS V/V 20 ASA_WS Envisat ASAR/WS :17 D WS V/V 21 ASA_WS Envisat ASAR/WS :29 D WS V/V 22 ASA_APH Envisat ASAR/AP :53 A I5 H/HV 23 ASA_APH Envisat ASAR/AP :41 D I1 H/HV 24 ASA_APH Envisat ASAR/AP :15 D I7 H/HV 25 ASA_APH Envisat ASAR/AP :38 D I1 H/HV 26 ASA_APH Envisat ASAR/AP :24 D I4 H/HV 27 ASA_APH Envisat ASAR/AP :29 D I3 H/HV 28 ASA_APH Envisat ASAR/AP :32 D I3 H/HV 29 ASA_APH Envisat ASAR/AP :38 D I1 H/HV 30 ASA_APH Envisat ASAR/AP :18 D I6 H/HV 18/01/2008 Page 37 of 259

39 31 ASA_APH Envisat ASAR/AP :32 D I3 H/HV 32 ASA_APH Envisat ASAR/AP :38 D I1 H/HV 33 ASA_APH Envisat ASAR/AP :38 D I1 H/HV 34 ASA_APV Envisat ASAR/AP :59 A I7 V/HV 35 ASA_APV Envisat ASAR/AP :45 A I3 V/HV 36 ASA_APV Envisat ASAR/AP :31 A I1 V/HV 37 ASA_APV Envisat ASAR/AP :36 A I2 V/HV 38 ASA_APV Envisat ASAR/AP :48 A I4 V/HV 39 ASA_APV Envisat ASAR/AP :53 A I6 V/HV 40 ASA_APV Envisat ASAR/AP :59 A I7 V/HV 41 ASA_APV Envisat ASAR/AP :33 A I1 V/HV 42 ASA_APV Envisat ASAR/AP :39 A I2 V/HV 43 ASA_APV Envisat ASAR/AP :45 A I4 V/HV 44 ASA_APV Envisat ASAR/AP :56 A I6 V/HV 45 ASA_APV Envisat ASAR/AP :31 A I1 V/HV 46 ASA_APV Envisat ASAR/AP :42 A I3 V/HV 47 ASA_APV Envisat ASAR/AP :48 A I4 V/HV 48 ASA_APV Envisat ASAR/AP :53 A I6 V/HV 49 ASA_APV Envisat ASAR/AP :59 A I7 V/HV 50 ASA_APV Envisat ASAR/AP :34 A I1 V/HV 51 ASA_APV Envisat ASAR/AP :39 A I2 V/HV 52 ASA_APV Envisat ASAR/AP :56 A I6 V/HV 53 ASA_APV Envisat ASAR/AP :54 A I6 V/HV 54 ASA_APV Envisat ASAR/AP :59 A I7 V/HV 55 ASA_APV Envisat ASAR/AP :53 A I5 V/HV 56 ASA_APV Envisat ASAR/AP :59 A I7 V/HV 57 ASA_APC Envisat ASAR/AP :35 D I2 HV/HV 58 ASA_APC Envisat ASAR/AP :21 D I5 HV/HV 59 ASA_APC Envisat ASAR/AP :27 D I4 HV/HV 60 ASA_APC Envisat ASAR/AP :24 D I4 HV/HV 61 ASA_APC Envisat ASAR/AP :30 D I3 HV/HV 62 ASA_APC Envisat ASAR/AP :24 D I4 HV/HV 63 ASA_APC Envisat ASAR/AP :30 D I3 HV/HV 64 ASA_APC Envisat ASAR/AP :35 D I2 HV/HV 65 ASA_APC Envisat ASAR/AP :41 D I1 HV/HV Table 5.1: ENVISAT-ASAR data acquisition 5.3 ENVISAT-MERIS Id Product Mission Sensor Start Orbit Track 1 MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : /01/2008 Page 38 of 259

40 10 MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : /01/2008 Page 39 of 259

41 60 MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : /01/2008 Page 40 of 259

42 110 MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : /01/2008 Page 41 of 259

43 160 MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : MER_FR Envisat MERIS : Table 5.2: ENVISAT-MERIS data acquisition 5.4 CHRIS From the optical system CHRIS the following data are available: ID Date Acquisition Angle Remarks 6BA0_Demin_ May :20 MZA = -21 High quality 6D65_Demin_ Jun :32 MZA = 1 Strong cover cloud F81_Demin_ Jul :15 MZA = -22 High quality 6FA8_Demin_ Jul :38 MZA = 11 Table 5.3: CHRIS data acquisition Strong cover cloud From the four acquisitions only two scenes are useful for further investigation and comparison with airborne data. (MZA= Minimum Zenith Angle) Figure 5.1: CHRIS/PROBA quicklook corresponding to central image of the 08-May-2006 acquisition 18/01/2008 Page 42 of 259

44 5.5 AATSR The Advanced Along-Track Scanning Radiometer (AATSR) provides information about the land surface temperature. ID Date Time Remarks AATSR_ _ql :09 L1b and L2 images acquired AATSR_ _ql :58 L1b and L2 images acquired AATSR_ _ql :03 L1b and L2 images acquired These images are available on ESA database to be acquired. ID Date Time ID Date Time ATS 25/06/ :40:26.59 ATS 25/04/ :57:35.20 ATS 26/06/ :34:09.54 ATS 27/04/ :19:41.33 ATS 27/06/ :02:32.69 ATS 30/04/ :25:26.64 ATS 28/06/ :46:11.94 ATS 01/05/ :09:05.88 ATS 30/06/ :08:17.98 ATS 02/05/ :37:29.00 ATS 01/07/ :51:57.22 ATS 03/05/ :31:11.89 ATS 03/07/ :14:03.21 ATS 05/05/ :43:14.21 ATS 04/07/ :57:42.43 ATS 07/05/ :05:20.12 ATS 06/07/ :19:48.39 ATS 08/05/ :48:59.33 ATS 07/07/ :03:27.60 ATS 10/05/ :11:05.18 ATS 09/07/ :25:33.54 ATS 11/05/ :54:44.37 ATS 10/07/ :09:12.74 ATS 13/05/ :16:50.16 ATS 11/07/ :37:35.82 ATS 14/05/ :00:29.33 ATS 12/07/ :31:18.64 ATS 16/05/ :22:35.07 ATS 14/07/ :43:20.89 ATS 17/05/ :06:14.23 ATS 16/07/ :05:26.70 ATS 19/05/ :28:19.93 ATS 17/07/ :49:05.88 ATS 21/05/ :40:22.06 ATS 19/07/ :11:11.66 ATS 22/05/ :34:04.74 ATS 20/07/ :54:50.84 ATS 23/05/ :02:27.70 ATS 22/07/ :16:56.61 ATS 24/05/ :46:06.84 ATS 23/07/ :00:35.77 ATS 26/05/ :08:12.43 ATS 25/07/ :22:41.51 ATS 27/05/ :51:51.56 ATS 26/07/ :06:20.66 ATS 29/05/ :13:57.09 ATS 28/07/ :28:26.36 ATS 30/05/ :57:36.19 ATS 30/07/ :40:28.48 ATS 01/06/ :19:41.98 ATS 31/07/ :34:11.15 ATS 02/06/ :03:21.30 ATS 01/08/ :02:34.11 ATS 04/06/ :25:27.70 ATS 02/08/ :46:13.25 ATS 05/06/ :09:07.02 ATS 10/08/ :19:48.20 ATS 06/06/ :37:30.27 ATS 11/04/ :22:30.37 ATS 07/06/ :31:13.39 ATS 12/04/ :06:09.70 ATS 09/06/ :43:15.95 ATS 14/04/ :28:16.10 ATS 11/06/ :05:22.24 ATS 16/04/ :40:18.66 ATS 12/06/ :49:01.54 ATS 17/04/ :34:01.72 ATS 14/06/ :11:07.76 ATS 18/04/ :02:24.94 ATS 15/06/ :54:47.04 ATS 19/04/ :46:04.24 ATS 17/06/ :16:53.21 ATS 21/04/ :08:10.46 ATS 18/06/ :00:32.48 ATS 22/04/ :51:49.75 ATS 20/06/ :22:38.71 ATS 24/04/ :13:55.93 ATS 21/06/ :06:17.99 ATS 23/06/ :28:24.15 Table 5.4: ENVISAT-AATSR data acquisition 18/01/2008 Page 43 of 259

45 6 ATMOSPHERIC MEASUREMENTS Aerosols and water vapour profiles have been estimated at the Görmin test site during the airborne optical campaign. These estimates are needed in order to correct for disturbances in the optical data. The following parameters have been measured and instruments operated: Radio Sounding: Relative Humidity and dew point (DWD) Sunphotometer CIMEL (direct sun irradiance): Water vapour and aerosol optical depth (Uni Valencia) Sunphotometer CIMEL (sky radiance): Aerosol size distribution, phase function, refractive index, single scattering albedo, asymmetry parameter, columnas mass and volume concentration, effective aerosol radius (Uni Valencia) Sunphotometer Microtops: Columnar ozone and water vapor (Uni Valencia) Aureole Sunphotometer (FUBISS): optical thickness (Uni Berlin) 6.1 Radio sounding Radio sounding from the German Meteorological Service (DWD) has been performed on three days before or during the optical flight campaign. Date Position Time 10/06/2006 at the field :45 MESZ 04/07/2006 on the Görmin Farm 13:55 MESZ 05/07/2006 on the Görmin Farm 13:08 MESZ Table 6.1: Dates of radio sounding The sounding instrument of the type RS 80 from Vaisala (Finland) (Figure 6.1) reached a height of 18 km and has measured the wind speed, wind direction, dew point, relative humidity and the temperature with from a height of 2 km (Figure ). Figure 6.1: Installation of the balloon for the radio sounding measurements (right) and position of the first measurement taken in the middle of the field /01/2008 Page 44 of 259

46 Figure 6.2: Vertical profile of different parameters estimated with the radio sounder at the 10/06/2006 Figure 6.3: Vertical profile of different parameters estimated with the radio sounder at the 04/07/ /01/2008 Page 45 of 259

47 Figure 6.4: Vertical profile of different parameters estimated with the radio sounder at the 05/07/ Sunphotometer CIMEL (direct sun irradiance) The CE318 is a commercial sunphotometer designed for the automatic measurement of direct solar irradiance and sky radiance. The unit employed in this campaign was an extended version, measuring in channels centred at 340, 380, 440, 500, 670, 870, 940, 1020 and 1640 nm, with the 940 nm channel being dedicated to obtain the atmospheric columnar water vapour. The AOD at 1640 nm was not actually operative for this field campaign due to calibration problems. The full width at half maximum of each channel varies from 2 to 40 nm, shorter for the UV and wider for the 1640 nm. In the VIS range, the bandwidth is about 10 nm. The sensor head is equipped with a double collimator with a 1.2º field of view. The CE318 was deployed on the roof of the Farm Administration Building at Demmin from 18 th to 5 th June 2006 (Figure 6.5). During this first phase, the measurements were performed following the default instrument protocol, based on the measuring guidelines of the Aerosol Robotic Network (AERONET). On 5 th June, the instrument was moved to the Görmin test site in order to attend the intensive field campaign. Therefore the measurements were more representative for the correction of airborne and satellite data. In this case, the default measuring protocol was 18/01/2008 Page 46 of 259

48 modified, by the addition of continuous series of direct sun and diffuse sky radiance measurements during the airborne sensor acquisitions. On 10 th June, the CE318 sunphotometer was finally removed from the test site in order to attend a calibration field campaign in Spain. In Figure 6.6 the data availability is graphed. Figure 6.5: Cimel sunphotometer at Demmin and Görmin. Figure 6.6: Dates and times of the day when direct sun measurements have been performed with the CIMEL. These measurements include cloud contaminated measurements, to be filtered in the processing of data. 18/01/2008 Page 47 of 259

49 6.3 Sunphotometer CIMEL (sky radiance) In addition to the direct measurements from the CIMEL sunphotometer, sky radiance measurements were also performed for retrieval of more complex aerosol optical and physical properties. In Figure 6.7. the measurements availability is presented. Figure 6.7.: Dates and times of the day when sky radiance measurements were performed for more complex aerosol characterisation. Quality filters are applied in the processing step, so these measurements include cloud contaminated data. 6.4 Sunphotometer Microtops II The MICROTOPS II is a commercial handheld sunphotometer. It measures direct sun irradiance at 305, 312, 320, 936 and 1020 nm, for deriving columnar contents of ozone, water vapour and aerosol (AOD at 1020 nm). Because of its manual operation, it was only used for auxiliary measurements during the June intensive campaign. In Figure 6.8. the performed measurements are presented. 18/01/2008 Page 48 of 259

50 :00 02:24 04:48 07:12 09:36 12:00 14:24 16:48 19:12 Figure 6.8.: Dates and times of the day when Microtops II data is available. 6.5 Aureole Sunphotometer FUBISS The ground based measurements were performed with the FUB-Trailer. The trailer was equipped with the aureole sunphotometer system Fubiss-ASA2 of the FUB. In addition to these measurements an airborne CASI (Compact airborne spectrographic imager) mission on the has been performed with simultaneous Fubiss-ASA2 sunphotometer measurements and an AOT profile in the vicinity of the AGRISAR test site. The analyses of these data are not part of the contract, but could be made available. Instruments: Fubiss-ASA2 (Free University Berlin Spectrometric system Aureole Sunphotometer Adapter 2) optional data acquisition (only occasionally): Fubiss-ASA1, Fubiss-Polar, Casi + Fubiss ASA2 flight from /01/2008 Page 49 of 259

51 Available data are listed in the following: Date Instrument Instrument setup FUBISS-ASA FUBISS-ASA (to much clouds -> no measurements) FUBISS-ASA FUBISS-ASA FUBISS-ASA FUBISS-ASA FUBISS-ASA2 Figure 6.9: Fubiss-ASA2 installed on the FUB-Trailer Data products: aerosol optical thickness (AOT) spectral channels 412nm, 450nm, 500nm, 609nm, 778nm, 862nm 18/01/2008 Page 50 of 259

52 7 GROUND RADIOMETRIC MEASUREMENTS Several instruments have been available in order to perform measurements, which could be used for optical data validation. The main class of instruments can be divided into Solar range ground radiometric measurements and Thermal infrared ground radiometric measurements 7.1 Solar range ground radiometric measurements Three ASD field spectrometers have been available during two intensive ground measurements campaigns in June and July. One has been provided by the University of Valencia in Spain, one from the Friedrich-Schiller-University in Jena and one from the University of Naples in Italy. The main purpose of these measurements is to obtain sufficient representative spectra from different surface types of the test area. These spectra are to be used for calibration and validation of the atmospheric correction of airborne and spaceborne hyperspectral images. The instrumentation used was: ASD FS/FR spectro-radiometer, radiometrically calibrated from 350 to 2500 nm. Spectralon white reference panel. Leaf Transmittance/Reflectance portable dark chamber. Garmin Gecko GPS, connected to the radiometer. Cal/Val measurements were taken following this pattern: first a static surface radiance measurement consisting of five consecutive spectra, with white reference measurements before and after to allow assessment of illumination stability. Then surface radiance was measured continuously while walking to the next stop, in this way each spectrum collected corresponds to the integration over a stripe of surface around 5m long, which would better correlate with airborne measurement of similar footprint. The pattern is repeated until the field is characterized. Figure 7.1: Calibration and test of ASD from the Uni Valencia (left) and FSU (right) 18/01/2008 Page 51 of 259

53 For validation purposes the time span before and after flight overpass should not be larger than one hour, therefore the different fields usually cannot be completely covered; therefore the strategy followed was to try to cover the larger variability possible within each field. Later GPS positioning allows accurate collocation with imaging data. Date Field ASD Uni ASD ASD Valencia FSU Unina Type Notes 06 June 2006 Farm concrete X X Cal/Val Cross calibration 230 Wheat X X Cal/Val Soccer Field X Cal/Val 102 Sugar Cal/Val X Beet 07 June 2006 Farm concrete X Cal/Val 08 June Rape X Cal/Val 102 Sugar Beet X Leaf Refl/Trans Together with Chlorophyll sampling 10 June 2006 Soil X X Cal/Val In front of Barley X Cal/Val 230 Wheat X Cal/Val Dry Grass X Cal/Val Next to 230 Thermal station GPS malfunction Farm concrete X Cal/Val 04 July 2006 Farm concrete X X Cal/Val Cross calibration NO GPS 440 Barley X Cal/Val NO GPS 222 Corn X X Cal/Val Scintilometer transect NO GPS Cal/Val Scintilometer 05 July Wheat X transect Bowen station 06 July Corn X 102 Sugar Beet X 222 Corn X X Cal/Val Leaf Refl/Trans Leaf Refl/Trans Leaf Refl/Trans Table 7.1: Dates of acquisition: ASD field spectrometer Scintilometer transect Soil moisture transect 18/01/2008 Page 52 of 259

54 7.2 Thermal infrared ground radiometric measurements The Global Change Unit from the University of Valencia performed thermal measurements during the two acquisition date of the optical flight campaign with the following instruments: CIMEL radiometer CIMEL ASTER radiometer RAYTEK ST8 Infrared radiometer RAYTEK Thermalert MID radiometers Thermocouple Type K. EVEREST 1000 calibration source LICOR LI-1000 Dataloggers Thermo Tracer TH9100 Pro and Irisys- Iri1011 thermal cameras Measurements table (June 2006): (*BT means Brightness Temperature) Figure 7.2: Thermal infrared ground radiometric measurements at the second intensive campaign in June Date Local time Field Parameter Measured Instrument BT transect RAYTEK ST6 222 BT temporal series Thermal camera NEC TH BT continuosly RAYTEK MID BT continuosly CIMEL CE /06/06 BT transect RAYTEK ST BT continuosly RAYTEK MID BT transect RAYTEK ST BT continuosly RAYTEK MID BT temporal series Thermal camera NEC TH BT continuosly CIMEL CE /06/ BT continuosly RAYTEK MID BT continuosly RAYTEK MID 2 08/06/ Emissivity CIMEL ASTER CE Emissivity CIMEL ASTER CE /06/ Emissivity CIMEL ASTER CE /06/ BT continuosly RAYTEK MID 2 BT continuosly CIMEL CE312-1 Emissivity CIMEL ASTER CE312-2 BT transect Thermal camera NEC TH /01/2008 Page 53 of 259

55 10-14 BT continuosly RAYTEK MID BT transect Thermal camera NEC TH Measurements table (July 2006): *BT means Brightness Temperature BT continuosly RAYTEK MID 3 BT continuosly BT transect Emissivity BT IFOV CIMEL ASTER CE312-2 Thermal camera NEC TH9100 CIMEL ASTER CE312-2 Thermal camera NEC TH9100 Date Local time Field Parameter Measured Instrument BT continuosly CIMEL CE BT transect RAYTEK ST BT IFOV Thermal camera NEC TH /07/ BT continuosly CIMEL CE ASTER BT transect RAYTEK ST BT IFOV Thermal camera NEC TH BT continuosly RAYTEK MID BT transect RAYTEK ST BT IFOV Thermal camera NEC TH Asphalt (Farm) BT transect RAYTEK ST BT continuosly CIMEL CE Emissivity CIMEL CE Emissivity CIMEL CE ASTER BT temporal series Thermal camera NEC TH /07/ BT continuosly CIMEL CE ASTER BT temporal series Thermal camera NEC TH BT continuosly RAYTEK MID Emissivity CIMEL CE Emissivity CIMEL CE ASTER BT IFOV Thermal camera NEC TH Pavement Emissivity CIMEL CE (Farm) Emissivity CIMEL CE ASTER /01/2008 Page 54 of 259

56 8 SOIL AND VEGETATION MEASUREMENTS As described in the Experimental Plan two procedures have been chosen for the estimation of soil and vegetation parameters (Section 3.1). Continuous measurements performed simultaneously to the radar flights Intensive measurements performed on three times during the AGRISAR campaign In the acquisition matrix (Figure 8.11) the fields with the acquisition dates and the kind of parameter estimated from every team is listed. 8.1 Continuous measurements During the continuous measurements a reduced amount of parameters have been sampled every time of the radar over flight. In total 16 ground measurements campaigns have been conducted. Involved were DLR-DFD, ZALF, GEO-Inf and University of Kiel with around 13 persons collecting the following vegetation and soil parameters. The vegetation parameter sampling from the individual teams is listed in the following: Date Parameter Samples (no.) Uni Kiel DLR-DFD and ZALF Phenology 9 9 LAI 9 9x2 Vegetation height 9x2 9 Crop density (plants/m) 9x2 9x2 Row distance 9 - Crop coverage (photo) 9x2 15 Biomass (wet/dry) 9x2 9x2 Plant water content 9x2 9x2 Chlorophyll (SPAD) Phenology 9 10 LAI 9 9x2 Vegetation height 9x2 9 Crop density (plants/m) 9x2 9x2 Row distance 9 - Crop coverage (photo) 9x2 15 Biomass (wet/dry) 9x2 9x2 Plant water content 9x2 9x2 Chlorophyll (SPAD) Phenology 9 - LAI 9 - Vegetation height 9x2 12 Crop density (plants/m) 9x2 - Row distance 9 - Crop coverage (photo) 9x2 15 Biomass (wet/dry) 9x2 - Plant water content 9x Phenology /01/2008 Page 55 of 259

57 Biomass (wet/dry) - 9x2 Plant water content - 9x2 Chlorophyll (SPAD) - 9 LAI - 9x2 Crop density (plants/m) - 9x Phenology - 13 Biomass (wet/dry) - 9x2 Plant water content - 9x2 Chlorophyll (SPAD) - 9 LAI - 9x2 Crop density (plants/m) - 9x Phenology 12 - LAI 9 - Vegetation height 9x2 15 Crop density (plants/m) 9x2 - Row distance 9x2 - Crop coverage (photo) 12x2 15 Biomass (wet/dry) 9x2 - Plant water content 9x Phenology LAI 9 9x2 Vegetation height 12x2 15 Crop density (plants/m) 12x2 15x2 Row distance 12 - Crop coverage (photo) 12x2 15 Biomass (wet/dry) 12x2 15x2 Plant water content 12x2 15x2 Chlorophyll (SPAD) Phenology - 15 Biomass (wet/dry) - 15x2 Plant water content - 15x2 Chlorophyll (SPAD) - 15 LAI - 9x2 Crop density (plants/m) - 15x Phenology 12 - LAI 9 - Vegetation height 12x2 15 Crop density (plants/m) 12x2 - Row distance 12 - Crop coverage (photo) 12x2 15 Biomass (wet/dry) 12x2 - Plant water content 12x Phenology - 15 Biomass (wet/dry) - 15x2 Plant water content - 15x2 Chlorophyll (SPAD) - 15 Crop coverage (photo) - 15 LAI - 15x2 18/01/2008 Page 56 of 259

58 Crop density (plants/m) - 15x2 Vegetation height Phenology 12 - LAI 9 - Vegetation height 12x2 - Crop density (plants/m) 12x2 - Row distance 12 - Crop coverage (photo) 12x2 - Biomass (wet/dry) 12x2 - Plant water content 12x Phenology LAI 9 15x2 Vegetation height 12x2 15 Crop density (plants/m) 12x2 15x2 Row distance 12 - Crop coverage (photo) 12x2 15 Biomass (wet/dry) 12x2 15x2 Plant water content 12x2 15x2 Chlorophyll (SPAD) Phenology - 15 Biomass (wet/dry) - 15x2 Plant water content - 15x2 Chlorophyll (SPAD) - 15 LAI - 15x2 Crop density (plants/m) - 15x Phenology - 15 Biomass (wet/dry) - 15x2 Plant water content - 15x2 Chlorophyll (SPAD) - 12 LAI - 15x2 Crop density (plants/m) - 15x Phenology 12 - LAI 12 - Vegetation height 12x2 15 Crop density (plants/m) 12x2 - Row distance 12 - Crop coverage (photo) 12x2 15 Biomass (wet/dry) 12x2 - Plant water content 12x Phenology 9 12 LAI 9 9x2 Vegetation height 9x2 12 Crop density (plants/m) 9x2 12x2 Row distance 9 - Crop coverage (photo) 9x2 12 Biomass (wet/dry) 9x2 12x2 Plant water content 9x2 12x2 Chlorophyll (SPAD) Phenology /01/2008 Page 57 of 259

59 LAI 9 9x2 Vegetation height 9x2 12 Crop density (plants/m) 9x2 12x2 Row distance 9 - Crop coverage (photo) 9x2 15 Biomass (wet/dry) 9x2 12x2 Plant water content 9x2 12x2 Chlorophyll (SPAD) Crop coverage (photo) - 15 Vegetation height - 9 Table 8.1: Vegetation parameter collected during the continuous campaign The soil parameter sampling from the individual teams is listed in the following: Date Parameter Samples (no.) Uni Kiel DLR-DFD and ZALF Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) - 15x3x Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 - Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 9x3x2 12x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) 12x3x2 15x3x2 Surface roughness (photogrammetric) Volumetric soil moisture (0-5cm, 5-10 cm) - 15x3x2 No. of samples = (number of elementary sampling units) x (number of recurrences) x (number of sampled depths) Table 8.2: Soil parameter collected during the continuous campaign 18/01/2008 Page 58 of 259

60 8.2 Intensive Measurements Three intensive measurements campaigns of one week have been performed during the four month of the AGRISAR project: April ( ): First intensive measurements campaign dedicated to soil samples and installation of permanent stations collecting soil moisture samples, weather parameters and atmosphere samples June ( ): Second intensive measurements campaign dedicated to vegetation/atmosphere samples July ( ): Third intensive measurement campaign dedicated to vegetation/atmosphere samples Intensive Measurements The measurement strategy was focused on soil parameters characterisation, as the amount of vegetation at this stage of the year is very short and sparse. Four teams contributed to this campaign: ISSIA, LHWM and LMU. In figure 8.3 the type of field crops of the Görmin test site in combination of the flight tracks are displayed. In figure 8.11 the parameter, that were sampled are listed in matrix corresponding to the field. The soil samples estimated are listed in the following: Date Parameter Institution Samples (no.) TDR SSIA/LHWM 237 samples 80 (Field 102) Surface roughness (Laser Profiles 3 (Field 102, length:20m, 15m, 10m) TDR LMU locations TDR ISSIA/LHWM 312 samples 30 (Field 250) 15 (Field 391) 23 (Field 230) 19 (Field 843) 17 (Field 160) Surface roughness (Laser Profiles 6 (Field 391, length: 3 of 20m & 3 of 5m) TDR LMU locations Surface roughness 5 samples on field 391 (Photogrammetry) Surface roughness LMU 15 samples on field 102 (Photogrammetry) Surface roughness (Laser ISSIA/LHWM 4 (Field 460, length: 2 of profiles 20m, 1 of 10m and 1 of 5m) Table 8.3: Soil parameter collected during the 1 st intensive campaign 18/01/2008 Page 59 of 259

61 For the estimation of the soil samples two methods were chosen, using predefined cylinders of a 5 cm length and TDR probe of 10 cm length. The sampling strategy is explained in the Experimental Plan and the TDR estimates for the first and second day of the 1. Intensive measurements campaign is displayed in figure Figure 8.1: Left: Surface roughness measurements with the ESA s Laser profile on field 102. Right: Soil moisture measurements with the TDR on the field 280. In addition to soil samples collected and soil parameters estimated two stable soil moisture stations from the LMU and the University of Kiel was installed at the first intensive ground measurements campaign on the field 250 at two different positions. The positions were selected with respect to the topography. The reinstallation of the stations was done after the last intensive measurements campaign. Hourly the data have been recorded during the four month. Weekly the data storage was exchanged. The description of the soil moisture stations are given in the Experimental Plan. Figure 8.2: Installation of the soil moisture station from the LMU at different depth. Further the Bowen-Ratio station was installed from the LHWM team on the field 250. Please see the description in section 9. 18/01/2008 Page 60 of 259

62 Figure 8.3: Sampling location for the continuous and intensive ground measurements. 18/01/2008 Page 61 of 259

63 Intensive Measurements The second intensive measurements campaign was focused on the vegetation parameter and atmosphere sampling. But also soil parameter sampling was estimated. The teams involved at this campaign are Uni Valencia, LMU, FSU, LHWM and DTU, see figure Date Parameter Institution Samples (no.) Leaf pigments Uni Valencia For chemical analysis of leaf pigments and SPAD- 502 calibration: Winter wheat (230) - 20 sets of two samples Sugar-beet (102) - 10 sets of two samples In situ SPAD measurements (about 30 SPAD measurements per ESU): Winter wheat (230) 2 ESU Sugar-beet (102) 1 ESU Biomass samples FSU 4 observation points á 2 samples TDR 0-15 cm 45 observation points á 5 samples TDR 0-5 cm 45 observation points á 5 samples * organic matter * soil texture (Koehn) * lime content (negative for all samples) * gravimetric soil moisture * soil colour (Munsell) 21 samples TDR LMU locations Dry matter 6 samples Plant water content 6 samples Biomass samples 6 samples Leaf pigments Uni Valencia For chemical analysis of leaf pigments and SPAD- 502 calibration Sugarbeet (102) - 10 sets of two samples Barley (440) - 20 sets of two samples In situ SPAD measurements: Winter wheat (230) 2 ESU Barley (440) 1 ESU Soil moisture (samples for FSU 17 gravimetric soil moisture estimation) 0-5 cm 18/01/2008 Page 62 of 259

64 Soil moisture (samples for 18 gravimetric soil moisture estimation) 5-10 cm TDR 0-15 cm 49 observation points á 5 samples TDR 0-5 cm 22 observation points á 5 samples Leaf area fotos 119 TDR LHWM 246 TDR LMU locations LAI 24 6 locations TDR FSU 57 observation points á 5 samples * organic matter * soil texture (Koehn) * lime content (negative for all samples) * gravimetric soil moisture * soil colour (Munsell) 58 samples Table 8.4: Soil and vegetation parameter from the 2 nd intensive campaign Measurements made by the ASD are presented in section 7. CORNER REFLECTORS In addition, for calibration purposes and for testing the estimation of extinction directly from the SAR images, small CR's of a size 20 cm x20 cm have been positioned at the second intensive campaign in three fields, winter wheat 230, maize 222 and winter rape 140 and stayed there until the end of the third intensive campaign. The CR s have been provided and positioned by the University of Alicante. Figure 8.4: Small corner reflector at the field Intensive Measurements Also the third intensive measurements campaign was focused on the vegetation parameter and atmosphere sampling. In addition also soil parameter sampling was done. The teams involved at this campaign are Uni Valencia, LMU, FSU, LHWM, DTU, Unina and ISSIA see figure /01/2008 Page 63 of 259

65 Date Parameter Institution Samples (no.) Leaf pigments Uni Valencia For chemical analysis of leaf pigments and SPAD- 502 calibration: sugarbeet (102) - 9 sets of two samples (problems with liquid nitrogen) In situ SPAD measurements: Winter wheat (230)-2 ESU Sugar-beet- 3 ESU TDR LMU locations Dry matter 12 samples Plant water content Biomass samples Soil Water Content (FDR) Unina 12 samples 12 samples Field 102: 10 ESUs Field 222: 44 ESUs Soil Water Content (TDR) Field 222: 5 ESUs + Fieldspec measurements LAI ISSIA 60 (Field 391) 35 (Field 230) 41 (Field 460) 36 (Field 102) Leaf pigments Uni Valencia For chemical analysis of leaf pigments and SPAD- 502 calibration Maize (222) - 20 sets of two samples In situ SPAD measurements: Sugar-beet- 1 ESU Maize (222) 1 ESU TDR LMU locations LAI locations Soil Water Content (FDR) Unina Field 102: 30 ESUs Soil Water Content (TDR) + Fieldspec measurements Soil hydraulic parameters Soil texture LAI Field 222: 5 ESUs Undisturbed soil sample: Field ESUs Disturbed soil sample: Field ESUs Field 102: 2 ESUs Field 222: 12 ESUs TDR ISSIA 64 (Field 230) 18/01/2008 Page 64 of 259

66 22 (Field 102) Leaf pigments Uni Valencia For chemical analysis of leaf pigments and SPAD- 502 calibration Maize (222) - 16 sets of two samples In situ SPAD measurements: Maize (222) 3 ESU Sugar-beet- 1 ESU TDR LMU locations Soil Water Content (FDR) Unina Field 460: 60 ESUs Soil hydraulic parameters Undisturbed soil sample: Field ESUS Soil texture Disturbed soil sample: Field ESUS Soil hydraulic parameters Undisturbed soil sample: Field ESUS Soil texture LAI Disturbed soil sample: Field ESUS Field ESUs Field ESUs TDR ISSIA 72 (Field 230) 24 (Field 823) Table 8.5 Soil and vegetation parameter during the 3 rd intensive campaign Measurements made by the ASD are presented in section Intensive Measurements LAND USE MAP Extensive measurements for the generation of a land use map were carried out at the second and third intensive campaign from the Technical University of Denmark. These measurements were performed for about 100 fields including the fields used for the intensive measurements. For almost all these fields the information collected was 1. Crop type 2. Phenological stage 3. Crop height estimation 4. Field photographs 18/01/2008 Page 65 of 259

67 Figure 8.5: Field shapes where the above mentioned parameters were collected. CHLOROPHYLL Leaf chlorophyll content (in digital counts units) was measured for some crops during the second and third intensive campaigns at different ESU (Elementary Sample Unit). Two instruments, CCM-200 (Opti-Sciences) and SPAD-502 DL (Minolta), were used for chlorophyll content determinations. SPAD has a measurement area of 6 mm 2 and its measurement method is based on the optical absorbency at two wavelengths. CCM-200 has a measurement area of 71 mm 2 and its measurement method is based on the optical density difference at two wavelengths. At figure 8.6 we can see both instruments. Figure 8.6: SPAD (Minolta) (left side) and CCM 200 (Opti-Sciences) (right side) 18/01/2008 Page 66 of 259

68 At the end of the second intensive campaign (June 06) CCM-200 chlorophyll meter began to give memory fails and was retired from the measurement campaign. So, only SPAD measurements are available for the third intensive campaign and may be that calibration procedure can be only applied to the SPAD instrument due to the distrust in the measurements made with CCM-200. This question will be analyzed in the future. As described in the Experimental Plan, two procedures have been carried out in order to obtain chlorophyll and pigments determinations: Weekly SPAD-502 (lent from Valencia University) chlorophyll measurements performed by ZALF team during all the AGRISAR campaigns Sample collection for chlorophyll meter calibration and pigment analysis purposes as well as in situ CCM-200 (June 06) and SPAD-502 DL (June and July 06) measurements performed by LEO group from Valencia University. In situ measurement procedure was described in the Experimental Plan. In Table 8.6 it can be seen the GPS coordinates of in situ measurements carried out for the LEO group in the second and third campaign. On account of the weekly chlorophyll measurements carried out by ZALF team, we centred our work on the samples calibration collection and only a few in situ chlorophyll measurements were made. June Campaign Crop (field) ESU E coordinate ESU N coordinate Winter wheat (230) Winter wheat (230) Winter wheat (230) Winter wheat (230) Sugarbeet (102) Barley (440) July Campaign Maize (222) Maize (222) Maize (222) Maize (222) Winter wheat (230) Winter wheat (230) Sugarbeet (102) Sugarbeet (102) Sugarbeet (102) Sugarbeet (102) Sugarbeet (102) Table 8.6: ESU GPS coordinates for in situ chlorophyll measurements during June and July campaigns 18/01/2008 Page 67 of 259

69 Each ESU measured, described by the GPS coordinates, will be characterized by the mean value of about 30 chlorophyll measurements made on different leaves of the ESU (after calibration procedure will be finished and calibration function for SPAD will be obtained and applied to each digital count measured by the chlorophyll meter) and the standard deviation error will be considered together with the mean value. In order to calibrate the SPAD-502 chlorophyll meter and to make the pigment analysis in the laboratory, 105 sets of samples of different crop leaves was collected between the second and third AGRISAR intensive campaigns (60 sets of samples in the second campaign and 45 in the third campaign). The chemical analysis by HPLC method in being performed at the laboratory of the Experimental Station Aula Dei, CSIC, Zaragoza (Spain) and it s not yet concluded. Analysis procedure is described in the Experimental Plan. For calibration purposes, about six measurements with the chlorophyll meter were made in the same part of each sample leaf and two samples were extracted from the leaf fraction measured. When chlorophyll measurement dispersion observed for a leaf was large, measurements continued until dispersion decrease sufficiently. The samples were cut by means of a copper cylinder with a sharp edge, leading to a sample with an area of (0.785 ± 0.016) cm 2. Each couple of samples was located inside a previously numerated aluminium foil to keep them, and preserved in liquid Nitrogen. At figure 8.7 it can be seen the copper cylinder used in the sampling procedure and the liquid Nitrogen container and at figure 8.8 it can be seen the couple of holes made in one maize leaf. Figure 8.7: Copper cylinder and liquid Nitrogen container Figure 8.8: Holes made in maize leaves when collecting calibration samples 18/01/2008 Page 68 of 259

70 At the end of the second campaign (June 06) the 60 samples were collected and transported to a secure freezer at the ZALF centre in Dedelow, where they were maintained at -20ºC. At the end of the third campaign (July 06), all samples collected in both campaigns were transported by car to Valencia (Spain) inside fully liquid Nitrogen containers. Due to the liquid Nitrogen evaporation rate, five containers of 2 L capacity were spent during the travel in order to refill the samples container with liquid Nitrogen. Two days were spent in the travel to Valencia (Spain) and at the arrival; samples were guarded into a freezer at -80 ºC at the Valencia University research centre. Finally in October (HPLC analyzer was duty before) the samples (inside a liquid nitrogen container) were transported by car to the laboratory of the Experimental Station Aula Dei, CSIC, Zaragoza (Spain) where they are being analyzed. When collecting calibration samples an experience has been carried out in order to analyse chlorophyll and pigments distribution along one leaf and determine the vertical distribution on a plant. Three couples of samples were taken from the same leaf at three distances from the centre of the plant (top, middle and bottom of the leaf) as it can be seen at Figure 8.9. This experience was repeated with two leaves of the same plant at different heights: one leaf in the upper side of the plant (leaf number six from the bottom of the plant) and another leaf at the bottom of the plant (leaf number two from the bottom of the plant). These experiences were repeated with six different plants of maize. Figure 8.9: Three couples of holes made on a maize leaf. At Table 8.7 it can be seen the crops considered for calibration and pigment analysis and the number of samples of each crop collected. Winter rape was impossible to measure with the chlorophyll meter due to the development stage when we went to measure. Crop (Field) Number of samples Winter wheat (230) 20 Sugarbeet (102) 29 Barley (440) 20 Maize (222) 36 Table 8.7: Samples of each crop taken for chlorophyll meter calibration 18/01/2008 Page 69 of 259

71 At figure 8.10 it can be seen the aspect of the maize field (222) used to collect maize calibration samples. At this moment we are waiting for the pigment analysis (including chlorophyll) results in order to translate digital counts measured by the chlorophyll meter to absolute chlorophyll value. Figure 8.10: Overview of the maize field. 18/01/2008 Page 70 of 259

72 crop winter wheat winter barley winter rape sugar beet maize (silo) grassl and field # date bare soil 1st Intensive Campaign at 10 locations: soil moisture 0-10cm, crop coverage (estimated ), vegetation height, photos, LMU München at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF/DLR, TDR, Uni Gent at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, UNI Kiel, Uni Gent at 5 locations: soil moisture 0-10cm, vegetation height, photos, phenology, plants per m2, LMU München at 4 locations: soil moisture, number of leaves, crop coverage (estimated), vegetation height, phenology, photos, LMU München surface rouhness measureme nts with a laser profiler of ESA, ISSIA, TDR, Uni Gent at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m; at 2 loc.: surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel, LMU München at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, surface roughness (photogram metric)uni Kiel surface rouhness measuremen ts with a laser profiler of ESA and TDR, ISSIA, Uni Gent, at 3 locations: at 3 locations: soil moisture soil moisture 0-5cm and 5-0-5cm and 5-10cm, crop 10cm, crop coverage coverage (estimated), vegetation height, photos, (estimated), vegetation height, photos, DLR surface DLR roughness (photogramme tric), UNI Kiel surface rouhness measurements with a laser profiler of ESA, ISSIA 18/01/2008 Page 71 of 259

73 at 5 location s: soil moistur e 0-10cm, crop coverag e (estimat ed), vegetati on height, phenol ogy, photos, LMU Münch en at 5 locations: soil moisture 0-10cm, crop coverage (estimated), vegetation height, photos, phenology, plants per m2; surface roughness (photogram metric) LMU München surface rouhness measureme nts with a laser profiler of ESA, ISSIA, TDR, Uni Gent surface rouhness measureme nts with a laser profiler of ESA, ISSIA, TDR, Uni Gent at 3 locations : soil moisture, number of leaves, crop coverag e (estimat ed), vegetati on height, phenolo gy, photos, LMU Münche n at 2 locations: soil moisture, photos, LMU München at 3 locations: soil moisture 0-10cm, crop coverage (estimated), vegetation height, photos, LMU München at 3 locations: soil moisture 0-10cm, crop coverage (estimated ), vegetation height, photos, LMU München at 2 location s: soil moistur e 0-10cm, crop coverag e (estimat ed), vegetati on height, photos, LMU Münch en at 5 locations: soil moisture 0-10cm, crop coverage (estimated), vegetation height, photos, LMU München surface roughness (photogram metric), LMU München surface rouhness measurements with a laser profiler of ESA, ISSIA at 7 locations: soil moisture 0-10cm, crop coverage (estimated), vegetation height, photos, LMU München 18/01/2008 Page 72 of 259

74 at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF/DL, TDR Uni Gent at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m; at 2 loc.: surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m; at 2 loc.: surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm,surfac e roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, photos, surface roughness (photogram metric), UNI Kiel at 3 locations: at 3 locations: soil moisture soil moisture 0-5cm and 5-0-5cm and 5-10cm, crop 10cm, crop coverage coverage (estimated), vegetation height, photos, (estimated), vegetation height, photos, phenology, DLR surface ZALF/DLR roughness at 2 loc.: (photogramme surface tric), UNI Kiel roughness (photogramme tric), UNI Kiel at 3 locations: at 3 locations: soil moisture soil moisture 0-5cm and 5-0-5cm and 5-10cm, crop 10cm, crop coverage coverage (estimated), vegetation height, photos, (estimated), vegetation height, photos, DLR surface DLR surface roughness (photogramme tric), UNI Kiel roughness (photogramme tric), UNI Kiel 18/01/2008 Page 73 of 259

75 at 3 locations: biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m; at 2 loc.: surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, photos, phenology, plants per m2 / shoots per m; at 1 loc.: surface roughness (photogram metric), UNI Kiel at 3 locations: phenology, ZALF at 3 locations: phenology, ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR at 2 loc.: surface roughness (photogramme tric), UNI Kiel at 3 locations: phenology, ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR surface roughness (photogramme tric), UNI Kiel 18/01/2008 Page 74 of 259

76 2nd Intensive Campaign at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF/DLR at 3 locations: biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF Biomass, Leaf area picutre, FSU BT, Uni Valencia ASD, Uni Valencia, FSU at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel TDR, at 6 locations: biomass, LMU Leaf area picutre, FSU BT, Uni Valencia BT, Uni Valenci a at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel Leaf area picutre, FSU at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF/DL R at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF Leaf area picutre, FSU ASD, Uni Valencia at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel Leaf area picutre, FSU at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF/DL R at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF Leaf area picutre, FSU BT, Uni Valencia at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel TDR, Leaf area picutre, FSU at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel at 3 locations: at 3 locations: biomass (wet biomass (wet and dry), and dry), chlorophyll, phenology, plants per m2, ZALF chlorophyll, phenology, plants per m2, ZALF Leaf area picutre, FSU Leaf area picutre, sample analysis (wetness, color, organic, etc.),tdr, FSU BT, Uni Valencia TDR, LMU TDR, LMU BT, Uni Valenci a 18/01/2008 Page 75 of 259

77 at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR land use map (type, height, BBCH, photos), DTU TDR, LMU München at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel BT, Uni Valencia at 5 locations: LAI, LMU München TDR, LMU München at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel, TDR, Uni Gent at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel TDR; soil moisture (0-5 cm), biomass, FSU/ZALF/D LR at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR surface roughness (photogramme tric), UNI Kiel TDR, biomass, FSU/ZALF at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR surface roughness (photogramme tric), UNI Kiel BT, Uni Valencia Emissivity, Uni Valencia ASD, Uni ASD, Uni Valencia Valencia BT, Uni Valencia BT, Emissivity, Uni Valencia ASD, Uni Valencia ASD, Uni Valencia Emissivity, BT IFOV, Uni Valencia BT, Uni Valencia BT, Uni Valenci a Emissiv ity, Uni Valenci a BT, Uni Valenci a 18/01/2008 Page 76 of 259

78 at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF/DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel at 3 locations: surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel 18/01/2008 Page 77 of 259

79 3rd Intensive Campaign at 3 locations: biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, shoots per m, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF at 3 locations: at 3 locations: biomass (wet biomass (wet and dry), LAI, and dry), LAI, chlorophyll, phenology, plants per m2, ZALF chlorophyll, phenology, plants per m2, ZALF Scintillometer, ITC at 3 locations: biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF land use map (type, height, BBCH, photos), DTU at 6 locations: biomass, LAI, LMU at 6 locations: biomass, LAI, LMU at 3 locations: biomass (wet and dry), LAI, phenology, shoots per m, ZALF at 3 locations: biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF chlorophyl l, spectrorad iometry, Uni Valencia at 3 locations: at 3 locations: biomass (wet biomass (wet and dry), LAI, and dry), LAI, chlorophyll, phenology, plants per m2, ZALF chlorophyll, phenology, plants per m2, ZALF soil cores Scintillometer, grid of soil grid of 44 for ITC moisture & points for soil laboratory temp. For moisture & analysis, UNINA top soil layer, soil cores for temp, cores soil for laboratory laboratory analysis, analysis, ASD, UNINA TDR, digital photos, UNINA LAI, ISSIA LAI, ISSIA LAI, ISSIA LAI, ISSIA EMIS, BT, Uni Valencia EMIS, BT, Uni Valenci a BT, Valencia Uni BT, Uni Valenci a ASD, Uni Valencia chlorophyll, ASD, Uni Valencia Scintillometer, ITC BT, Uni Valenci a Goniometer, ITC / FSU 18/01/2008 Page 78 of 259

80 at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR TDR, FDR, Spectroradi ometry, UNINA at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel TDR, at 6 locations: LAI, LMU soil moisture 0- Scintillometer, ITC 5cm, LAI, ISSIA/ LHWM/ FSU/ DLR Goniomete r, ITC / FSU chorophyll, EMIS, BT, Uni ASD, Uni Valencia Valencia TDR, LMU LAI, soil moisture 0-5cm, LAI, ISSIA/ LHWM/ FSU/ DLR BT, Uni Valenci a at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel chlorophyll, spectroradi ometry, Uni Valencia at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel Grid of soil moisture & temp. For 30 points, TDR, FDR, ASD, LAI, UNINA soil moisture 0-5cm, LAI, ISSIA/ LHWM/ FSU/ DLR Goniometer, ITC / FSU chlorophyll, ASD, Uni Valencia ASD, FSU Spectroradio metry, FSU at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR surface roughness (photogramme tric), UNI Kiel TDR, FDR, Spectroradiom eter, LAI, UNINA soil moisture 0-5cm, LAI, ISSIA/ LHWM/ FSU/ DLR at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR surface roughness (photogramme tric), UNI Kiel TDR, at 5 locations: LAI, LMU ASD, Uni Valencia Scintillometer, ITC EMIS, BT, Uni Valenci a BT, Emissiv ity, BT IFOV, Uni Valenci a TDR, LMU Münch en TDR, FDR, UNINA TDR, LMU LAI, TDR, LMU Münche n TDR, LMU LAI, TDR, LMU LAI, TDR, LMU LAI, 18/01/2008 Page 79 of 259

81 Scintillometer, ITC TDR, FDR, ASD, LAI, UNINA grid of 60 points for soil moisture & temp, UNINA TDR, ISSIA Goniomete r, ITC / FSU Goniometer, ITC / FSU ASD, Uni Valencia ASD, Valencia Uni Scintillometer, ITC soil moisture, ISSIA/ FSU/ DLR Scintillometer, ITC Scintillometer, ITC Scintillometer, ITC Scintillometer, ITC at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, shoots per m, ZALF/DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, phenology, shoots per m, ZALF/DL R at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, chlorophyl l, phenology, plants per m2, ZALF/DL R fungicide application fungicide application Scintillometer, ITC Scintillometer, ITC at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel 18/01/2008 Page 80 of 259

82 at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, phenology, shoots per m, ZALF/DLR at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogramme tric), UNI Kiel no data sampled (field harvested), at 3 locations: soil moisture 0-5cm and 5-10cm, photos, UNI Kiel no data sampled (field harvested), at 3 locations: soil moisture 0-5cm and 5-10cm, photos, DLR (field harvested), at 3 locations: soil moisture 0-5cm and 5-10cm, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogra mmetric), UNI Kiel no data sampled at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated ), vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2, ZALF/DL R (field harvested) ; at 3 locations: soil moisture 0-5cm and 5-10cm, photos, DLR at 3 locations: soil moisture 0-5cm and 5-10cm, vegetation height, photos, biomass (wet and dry), LAI, phenology, plants per m2 / shoots per m, surface roughness (photogram metric), UNI Kiel no data sampled at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel at 3 locations: soil moisture 0-5cm and 5-10cm, crop coverage (estimated), vegetation height, photos, biomass (wet and dry), LAI, chlorophyll, phenology, plants per m2, ZALF/DLR surface roughness (photogramme tric), UNI Kiel at 3 locations: at 3 locations: soil moisture soil moisture 0-5cm and 5-0-5cm and 5-10cm, crop 10cm, crop coverage coverage (estimated), vegetation height, photos, DLR (estimated), vegetation height, photos, DLR Figure 8.11: Sampling matrix of the AGRISAR campaign 18/01/2008 Page 81 of 259

83 9 SURFACE ENERGY BUDGET Boundary layer characterisations with thermal radiation and meteorological parameter have been collected with the Bowen-ratio station from the LHWM and a Goniometer and Scintillometer (mobile and stable) from ITC. 9.1 Bowen-ratio station The Bowen-ratio station have been installed at the first intensive ground campaign mid of April 2006 and have been reinstalled at the third intensive ground campaign in the beginning of July The station has been positioned at the field 250 close to the stable soil moisture station from the LMU. The measurements have been hourly recorded and weekly the recording tape has been exchanged. The following parameters have been measured from April until July 2006: Air temperature and humidity at two levels Wind speed and direction Air pressure Net radiation Ground heat flux Air pressure Figure 9.1: Installation of the Bowen-ratio station on field 250 in April Scintillometer (LAS) The regional sensible heat flux characterisation has been done with the Large Aperture Scintillometer (LAS) that has been positioned at the corn field 222 and a fixed station at a wheat field 250. The parameters that has been measured are the following: Radiation flux: Short & long wave incoming and outgoing radiation (Instruments: Kipp and Zoonen CNR1) Conduction: Soil heat fluxes (Instruments: HFP01 soil heat flux plates and soil thermistors) Sensible heat (Instruments: Kipp and Zoonen Large Aperture Scintillometer) 18/01/2008 Page 82 of 259

84 Air-pressure, -temperature, -humidity and wind speed profiles as well as wind direction in the canopy air space (Instruments: A100L2, W200P) Figure 9.2: Mobile LAS on field 222 (left) and stable station on field Goniometer The architecture of vegetation canopies plays an important role in the complex threedimensional exchange of heat, requiring directional reflection and temperature measurements of the different canopy components. Therefore a Goniometer, is employed for these directional measurements over characteristic fields using different sensors. The following parameters have been measured with the goniometer: Directional reflection (Instruments: Canon Powershot digital camera, ASD Spectrometer) Directional temperature (Instruments: Irisys 1010 Thermal Imager, Everest 3000 Radiometer, CIMEL (ASTER) Radiometer) Figure 9.3: Goniometer measurements at the third intensive ground campaign from ITC 18/01/2008 Page 83 of 259

85 On the following dates and fields the Goniometer has been used: Date Field Gonio 4 july 222 ThC, ASD, DCm 5 july 102 ASD 230 ThC, ASD, DCm, Erm 6 july 102 ThC, DCm, Cim, Erm 440 ThC, ASD, DCm, Cim, Erm Table 9.1: Table of Goniometer measurements Abbreviation to the measuring instruments: ThC = Thermal Camera, ASD = Analytical Spectral Device Field spectroradiometer, DCm = Digital Camera, Cim = Cimel radiometer (ASTER bands), Erm = Everest Thermal Radiometer. 18/01/2008 Page 84 of 259

86 10 AGRISAR DATA BASE The web interface for the users is displayed in figure The interface named as EOWEB is an operational system used also for an easy access for satellite and shuttle data and has been developed by the Germany data archive center (DLR-DFD). Each data set of the AGRISAR campaign is displayed and a quick look added for a first look. The data can be directly ordered through this interface and the data are provided over an ftp site. For all AGRISAR partners a login and password have been provided to the main responsible of each institution in order to be able to check the data quality and to analyse the data within the AGRISAR Team. Whereas, all the data which have not a specified standard were made available through a ftp site provided by DLR-HR. This ftp site has no special web interface and is accessible for all AGRISAR participants. The login and the password have been distributed to all AGRISAR participants. The main information layers stored on this ftp site are: ground measured data (vegetation, soil, atmosphere) optical data (AHS, CASI) simulated data (Sentinel-1 and Sentinel-2) Figure 10.1: The DLR s EOWEB portal for AGRISAR radar data downloads 18/01/2008 Page 85 of 259

87 11 DATA QUALITY 11.1 Airborne Data Quality Synthetic Aperture Radar Data The main processing procedure for the AGRISAR data is displayed in a sketch in figure 11.1 Accordingly, the SAR raw signals are recorded on tape during the acquisition flight. Later on the tape is transcribed on a hard disk and first image surveys are produced. The data are then processed to radar geometric images (RGI), where as an input navigation data are used to account for velocity variations and aircraft roll, drift and pitch angles. Tiepoints (corner reflectors) are used to evaluate the radiometric performance. The RGI outputs are radar data in slant range and ground range geometry. In addition also system and processing related information files are generated that are useful for further information extraction. The RGI are stored in a DLR archive named data information and management system (DIMS). For geocoded and terrain corrected data (GTC) an elevation model is introduced in a further processing step. The digital elevation model is derived from the single pass X-band data acquisition. The GTC products are also stored in the DIMS and are available through a web-interface for all AGRISAR partners. All 16 campaigns were stored in the DIMS with a total amount of around 21 GByte of RGI and 11 GByte of GTC data. E-SAR Processing Raw data on tape Transcription and survey processing Raw data on disk Tiepoint positions Aircraft position and attitude DEM Tiepoint positions Processing Processing RGI data GTC data RGI data: GTC data: - SLC_sr, ML_sr, ML_gr data - geocoded data - images and additional information - geotiff images 400MB 1.3GB / product - images and additional information MB / product DIMS data archive at DLR-DFD RGI: Radar Geometry Image GTC: Geocoded Terrain Corrected DIMS: Data Information and Management System EOWEB: Earth Observation on the WEB EOWEB User Interface Figure 11.1: AGRISAR SAR processing chain Agrisar User 18/01/2008 Page 86 of 259

88 The E-SAR data specification for the RGI products are summarised in Table 11.1 for the different frequencies. Accodingly, the X-, C- and L-band frequencies have a slant range resolution of 2 m for the single look complex data (SLC) and for the multilook data. The azimuth resolution is m for the SLC data and up to 4.5 m for the multilook data. Geocoding has been performed onto a 2x2 m grid in WGS-84 UTM projection, zone 33. Resolution SLC image Multilook image slant range azimuth No. of looks slant range azimuth X-band 2m 0.89m (0.59m) 8 (16) 2m 4m (5m) C-band 2m 0.89m (1.2m) 8 (4) 2m 4m (3m) L-band 2m 1.0m (1.2m) 8 (4) 2m 4.5m (3m) Posting 2m x 2m UTM zone 33 (31) Table 11.1: AGRISAR SAR processing parameters (SLC= single look complex image) Digital Elevation Model The digital elevation model was derived from X-band single pass SAR interferometry. The data were acquired during AGRISAR campaign, Mission 01 on the 18.April The elevation model is covering a wider area of the Görmin test site compared to the follow on continuous and intensive data acquisitions. In total 5 overlapping East-West flight strips and 1 NE-SW strip were flown for the complete coverage. In figure 11.2 the elevation model is displayed with area coverage of 10 km by 14 km. The acquired across track strip has a high squint (> 10 degree) and strong track variation (+/- 10 m). The terrain height is ranging between 19 m and 102 m and can be classified as a pretty flat area. The height is increasing from the river Penne toward the north direction. For the duration of the whole campaign 6 corner reflectors were positioned distributed from near to far range and along track. For the East-West tracks the corners 1 to 4 are deployed and for the NE-SW track additionally the corners 5 and 6 are positioned along azimuth. For the position of the different corner reflectors the height error has been measured. Corner reflectors 1 to 4 are used for processing (tiepointing) and 5 to 6 just for control. 18/01/2008 Page 87 of 259

89 The height errors of the first four corner reflectors are less than 1 m with the exception of corner 1 which shows a variation of 1.9 m. The corners 5 and 6 have higher height errors in the order of 3m. Please note that the corner reflectors 1 to 4 have a size of 1.5 m whereas the corner 5 and 6 are with leg length of only 90 cm. Some uncertainty is also attributed to the fact, that the position of the CR phase center for CR1-4 is located about 0.2m-0.4m above the ground, which means that the DEM value at these positions could be biased. 18/01/2008 Page 88 of 259

90 Figure 11.2: Derived digital elevation model from X-band single pass SAR interferometry for the Görmin region. The digital elevation model is covering more than only the two flight stripes of the continuous and the intensive data acquisition (colour display of the two flight stripes). Corner reflectors positions are displayed in the image. 18/01/2008 Page 89 of 259

91 Acquired SAR Data During the AGRISAR campaign the East_west stripe was acquired for 16 missions and has a time span of 104 days. The time difference is varying between 7 and 14 days. The across stripe covering the north-south part of the Görmin site was only acquired during the intensive ground measurements and the time span between to the first acquisitions is 47 and 77 days. AgriSAR Time Frame flight direction: 271 Along Stripe (E-W): M02: 19. April 06? t M04: 01. May days M05: 11. May days M06: 16. May days M07: 24. May days M09: 07. June days M10: 13. June days M11: 21. June days M12: 05. July days M14: 12. July days M15: 26. July days M16: 02. Aug days flight direction: 204 Intensive campaigns Across Stripe (NNE-SSW): M03: 20. April 06?t M08: 06. June days M13: 06. July days In figure 11.3 all data are displayed that were processed for the AGRISAR campaign in RGI and GTC format for the different frequencies. The X-band frequency on the E-SAR system is only available in single polarisation (hh or vv). The C-band frequency has the capability of acquiring data in a dual polarisation mode (hh and hv or vv and vh). Only the L-band frequency has the fully polarimetric capability at the E-SAR system. 18/01/2008 Page 90 of 259

92 Figure 11.3 SAR data processing for the AGRISAR campaign Radiometric Quality Analysis One important criterium for data quality is the measure of radiometric accuracy by investigation of the radar cross section (RCS) of the deployed corner reflectors. In figure 11.4 the RCS values are shown for all X- and C-band data takes. The accuracy is in all cases within the +/-2dB margin, in C-band even better. 18/01/2008 Page 91 of 259

93 Figure 11.4: Radiometric radar backscattering measured on the corner reflectors in X- and C- band (+ : near range cr, * : far range cr, : theoretical value, : 2 db range) : ΔRCS > 2dB : Squint > 2 : high backgr. wrong estimation 0509_NR 0912_NR 0912_NR 0509_FR 1309_NR 1109_NR+FR 0509_FR 1309_NR+FR 0308_FR 1109_NR+FR Figure 11.5: Radiometric radar backscattering analysis in L-band (+ : near range cr, * : far range cr, : theoretical value, : 2 db range). CR outside the +/- 2dB margin are indicated (see explanation in text). 18/01/2008 Page 92 of 259

94 Figuere 11.5 presents the evaluated RCS for the L-band data takes. Opposite to C- and X-band some of the CR signatures are outside the +/-< 2dB margin. We have evaluated these cases in more detail and found the following explanations: We attribute the cases of CR RCS higher than the theoretical limit + 2dB margin (red circle) to areas of strong background scattering (near range CR). When the CR were deployed the fields were bare, but during the growing season the crops evolved leading to higher backscatter. After harvesting the estimation of RCS was again reliable. For the cases of RCS lower than the theoretical limit -2dB we found significant squint angles for the particular data takes. This means that there is a mis-orientation of the CR with respect to the radar line of sight, which leads to lower estimates for the RCS. Therefore we have a certain confidence that the data themselves are not biased. 06agrsar0902x1_t01, Xvv 06agrsar0904x1_t01, Xhh 06agrsar0909x1_t01, Cvh/vv 06agrsar0910x1_t01, Chv/hh 06agrsar0912x1_t01, Lpol Subsets in Xvv, Cvh/vv, and Lpol Figure 11.6: Geocoded Data from Mission 09 of the 07.June 2006 at different frequencies and polarisation (polarisation displayed in RGB lexicographic) 18/01/2008 Page 93 of 259

95 Synthetic fully polarimetric C-band RGI Product VV-VH RGI Product HH-HV Repeat pass interferometric processing Including residual motion error estimation using HV and VH channels Interferometric phase correction due to the residual baseline Track fusion and Parameter correction Geocoding Synthetic RGI Product HH-HV-VV-VH Synthetic GTC Product HH-HV-VV-VH Figure 11.7: Generation of synthetic quad-pol products from pairs of C-band dual-pol data sets. For the AGRISAR campaign DLR-HR has generated for the first time synthetic quad-pol products in C-band. Two data sets with nominal zero baseline, one vh-vv and one hv-hh, are processed as a repeat-pass interferometric pair, including residual motion compensation. The phase of the hv-vh interferogram is used to eliminate the possible presence of an interferometric phase and CRs are used to calibrate the co-polar phase between hh and vv. The reference track data of the slave are adapted to those of the master in order to suggest data acquisition from a common flight. However, the real tracks keep the information of the individual flights, which could be used e.g. to estimate volume decorrelation. Finally the data are geocoded as quad-pol product. Synthetic quad-pol products were generated only for the intensive campaign dates. A zoom is given in the Figure 11.8 below. 18/01/2008 Page 94 of 259

96 Figure 11.8: Synthetic C-band fully polarimetric image (zoom from a red marked rectangle) Sentinel-1 Simulation Sentinel-1 simulation in interferometric wide swath mode has been performed for the C-band data of second two interferometric campaigns ( ID s 0909 and 0910 acquired June and ID s 1207 and 1208 acquired on July 5, 2006). In addition and for comparison also a simulation for the Sentinel stripmap mode was performed for the June 7 data. The parameters which were used for the simulation are summarized in Table Sentinel-1 simulation Interferometric Wide Swath(**) Stripmap slant range(*) azimuth slant range(*) azimuth Resolution (SLC) 2.1 m 20 m 1.7 m 5 m PSLR 25 db (with spectral weighting function) NESZ DTAR -22 db (insertion of additional noise as function of off-nadir) -22 db (modification of presumming filter) Table 11.2: SAR parameters used for Sentinel-1 simulation. The selected slant-range resolution corresponds to a ground range resolution of 25 deg for the IWS mode and 20 deg for the stripmap mode. 18/01/2008 Page 95 of 259

97 Pauli RGB Image Sentinel Wide-swath mode Resolution: Gr-Range: 5m Azimuth: 20m NESZ: -22dB ESAR Resolution: Gr-Range: 3m Azimuth: 0.75m NESZ: -28dB Figure 11.9: Polarimetric decomposition to Pauli basis: E-SAR high resolution vs. simulated Sentinel-1 data in IWS mode. The comparison of Pauli decomposition (HH-VV, 2*HV, HH+VV) for E-SAR and simulated Sentinel-1 data in IWS mode in shown in Figure The comparison of Sentinel-1 stripmap simulation to E-SAR high resolution and to IWS simulation is given in Fig below. 18/01/2008 Page 96 of 259

98 Figure 11.10: Comparison of high resolution E-SAR data with Sentinel-1 simulation in stripmap and IWS mode. Color coding is RGB: HV-HH-HH Details of build up areas cannot be discriminated any more in the IWS simulation. However the information on the extended agricultural fields is well maintained Optical Data AHS data products and processing methodology The INTA-AHS data distribution protocol delivers, depending on project requirements, ungeoreferenceable raw data (L1a) and georeferenceable at-sensor radiance (L1b) and ground reflectance and temperature (L2b) products which are attached with geographic look-up table (*igm ENVI file) to apply georeference process. For AGRISAR project INTA delivered L1a, L1b and L2b products. Thereafter, different procedures were applied to generate the user s requested products: L1a (raw data, L0R000 in INTA code): no processing, raw data imported to BIL format + ENVI header, 753 values per line (includes maker bit, BB1 and BB2), image and navigation stats computed for quality check L1b (at-sensor radiance, L10020 / L00120): The VIS/NIR/SWIR bands were converted to atsensor radiance applying the absolute calibration coefficients obtained in the laboratory using the integrating sphere. The MIR/TIR bands were converted to at-sensor radiance using the information from the onboard blackbodies and the spectral responsivity curves obtained by the 18/01/2008 Page 97 of 259

99 AHS spectral calibration. The resulting files were converted to BSQ format + ENVI header and scaled to fit an unsigned integer data type, applying the following expression: output_value = fix(input_value * 1000), output units are nw/(cm 2 sr nm). The specifications are: NedL VIS-NIR : <0.2 w/(m 2 srum), NedL SWIR: <0.3 w/(m 2 srum) NedT TIR: <0.33 ºC, L2b (Ground reflectance and temperature L20020 / L00220): ATCOR4 [1] code is used for the atmospheric correction of AHS imagery. ATCOR4 is based on MODTRAN-4 and performs a scan angle atmospheric correction taking into account flight altitude an illumination conditions. Water vapour is estimated directly from the image using 940nm absorption band and aerosol type and visibility estimated by field spectroscopy and meteorological auxiliary data acquired by AGRISAR ground team. Georeferenced process is made using the parametric code PARGE [2] resampling, for the AHS pixel, the digital elevation model delivered by the DLR. PARGE integrates the attitude info: Applanix POSAV-POSEO roll-pitch-heading and position info. Two PARGE optional outputs are used to be appended on each AHS image: *igm file: Geographic Lookup table (for ENVI software) *sca file: which include scan zenith angle (degrees*100), scan azimuth angle (degrees*10) and flight altitude (meters). Each AHS image was appended with metadata files in XML format. Remote Sensing Laboratory created his own metadata profile for the AHS imagery [3], following the standards ISO and using the java tool ISO Metadata Editor (IME) developed by the Laboratory. AHS DATA QUALITY Radiometric and geometric quality check analyses were undertaken to all AHS imagery in order to evaluate the data distributed performance for both projects. A total of 24 AHS images were acquired in AGRISAR Mission-1 and 20 in Mission-2, taking into account all products processed nearly 100 GB of data were delivered. Figure shows example of the AHS NEDL and NEDT representative values for AGRISAR. AHS performance in VNIR region was quite satisfactory in terms of NEDL being all channels below 20 nw / cm 2 sr nm. The SWIR part raised NEDL to 30 nw/cm 2 sr nm with some ending part channels near the specification limits. Furthermore it can be seen zero levels of NEDL in the AHS bands 47, 53, 55 and 61 which are not valid. TIR region performed well below 0.2ºC of Noise Equivalent Difference Temperature except of edge channels 79 and /01/2008 Page 98 of 259

100 Figure 11.11: Examples of the AHS NEDL and NEDT performance for AGRISAR Field spectra acquired by AGRISAR ground team with and ASD FieldSpec [ the field signatures were resampled to AHS spectral configuration using ENVI spectral resampling tool [ and compared with BOA reflectance of AHS for the same surfaces (Figure 11.12). Averaged error of 10% was gathered in the comparison. Figure AHS BOA reflectance comparisons with ASD field spectra. Orthorectified AHS images by parametric method were corrected in a line by line independently way. For flat areas like AGRISAR, the most important factors for uncertainty are the correct Applanix data registration and the GPS reference station for post-processing. The visual comparison with the vector cartography overlaid over the geocoded images shows average geometry accuracy below 2 pixels, getting the error maximum in some off-nadir areas displacements. The superposition of multidate images for each AGRISAR mission can be seen in Figure Perfect match areas between different date composition show grey visualization and on the contrary colour regions indicate areas with no match between images. Mission-2 composition (bottom part Figure 11.13) seems to be quite satisfactory; Mission-1 mutidate image (upper part Figure 11.13) shows a lot of colour areas but with radiometry rather than geometry mismatches. In a general view of the flight line clouds and cloud shadow for the June 6 th and not in June 10 th was the reason. In more detail, that can be seen in the subsets, the colour mismatches are due to shadow houses and trees appeared in different position due to the delay of 2 hours between acquisitions. 18/01/2008 Page 99 of 259

101 AGRISAR Mission-1 P02BD Multidate AHS 15. R (June 6 th 12:57UTC) G (June 6 th 2:57UTC) B (June 10 th 14:57 UTC) AGRISAR Mission-2 P02BD Multidate AHS 15. R (July 4 th 8:58 UTC) R (July 4 th 8:58 UTC ) B (July 5 th 8:19 UTC) Figure 11.13: AGRISAR multidate AHS 15 RGB composition for the same flight line. Mission-1 (upper part) Mission-2 (bottom part) Details are given in [ 48 ], [ 54 ] and [ 18 ] ITRES (ITRES) Standard processing of CASI data produces georeferenced, radiometrically corrected imagery. There are two major steps. The first step is applying radiometric calibrations to convert the raw digital numbers into radiance values. The second step applies measurements from the airborne inertial system and GPS to create a georeferenced mosaic in UTM coordinates using the WGS84 datum. These processes are well-established for the CASI imagery and are described more fully in the following sections. Radiometric Correction Radiometric quality of the imagery is important so that scene brightness remains stable under varying illumination and site conditions. The input parameters for the radiometric corrections include realtime system offset measurement such as dark data, scattered light, electronic offset and calibrated brightness coefficients generated in the lab. Radiometric corrections are applied to the CASI-1500 imagery to convert the raw digital numbers to spectral radiance units (1 SRU = 1.0 µw cm-2 sr-1 nm-1). A traceable light standard is used to determine the appropriate 18/01/2008 Page 100 of 259

102 multiplicative coefficients for the radiometric corrections. ITRES normally quotes an accuracy of ± 2% over the spectral range of 430 to 800 nm. Between 365 and 430 nm and between 800 and 1050 nm, the quoted accuracy is ± 5%. Once radiometrically calibrated, a measurement made from any CASI-1500 instrument will be identical to the same measurement made from another CASI-1500 to within the accuracies quoted above. Figures and demonstrate both the appearance and spectral characteristics of raw vs. calibrated CASI-1500 data, appearing on the left and right in each case, respectively. Of note are the artifacts in the raw image which are removed through calibration. The application of the calibration coefficients allows for the characteristic spectral profile of a tree to be derived from the raw digital data acquired during the flight. Figure 11.14: CASI-1500 data before and after radiometric correction c Figure 11.15: CASI-1500 spectral plots for a tree before and after radiometric correction, left & right, respectively. Notes on accuracy of radiometric correction Over the course processing the flight data from the AgriSAR/Eagle acquisitions, a number of issues related to the radiometric correction of the CASI-1500 data were encountered. 18/01/2008 Page 101 of 259

103 A persistent blue-end ( nm) calibration issue was noted, where the signal to noise ratio (SNR) of the spectral channels In this region were lower than expected, creating some noise effects in the imagery. A second issue which manifested itself in the imagery was the effects of thermal/kinetic shifting of the internal spectrograph of the SHU. Such movement affected the alignment of mapped wavelengths to actual spectral features by up to +/- 1.5 nm, dependent on the nature of the shift. All data was examined for this effect and corrected using methodologies specific to ITRES. Spectral Resampling of CASI-1500 Data to Simulate Sentinel-2 VNIR Channels. CASI-1500 data acquired over the Demmin ROI on July 5th was resampled from 288 bands to match the first 9 channels (VNIR) of the proposed Sentinel-2 imager. Imagery was averaged over rows that best fit bandset using the bandmath function in ENVI. Some non-noise image artifacts were attenuated through spectral averaging, while instrument noise effects in the UV/Blue channels were averaged out due to the increased SNR derived by the resampling process. Geometric Corrections Once radiometrically corrected, the spectral radiance imagery is submitted to geocorrection processing. For this project, imagery is georeferenced to the UTM coordinate system using a Geographic Lookup Table (GLT). The map projection referenced in WGS84 and using horizontal datum UTM zone 33, NAD83 for the AgriSAR data. All EAGLE data is georeferenced using horizontal datum UTM zone 31, NAD83. All units are in meters and the pixel size varies according to data set. This process involves the integration of five separate data streams: 1. Radiometrically corrected CASI image data 2. Coincident airborne GPS data from an integrated GPS receiver 3. Coincident ground-based GPS data from an externally operated GPS base-station 4. Coincident attitude data from an integrated IMU onboard the aircraft 5. Digital elevation model 18/01/2008 Page 102 of 259

104 These data streams are blended to generate a geometrically corrected image map using the following six steps: 1. Raw CASI digital numbers are calibrated. 2. Aircraft GPS data are differentially corrected (DGPS) using data from a nearby GPS base-station. 3. Blended solutions of DGPS positions with aircraft attitude data from the Inertial Measurement Unit (IMU) are generated. 4. Position and attitude data are optimized using sensor misalignments determined in boresite calibration (bundle adjustment process) and applied to the CASI image data. 5. A Digital Elevation Model (DEM) is applied during geocorrection to remove topographic effects and facilitate the final ortho-rectification of the CASI imagery. 6. A north-up image with square fixed-sized pixels is populated using a nearest neighbor algorithm. Figure : CASI-1500 data over Demmin site after geocorrection & resampling to 3x3m pixels for the sake of display (Data aquired July 05, 2007). 18/01/2008 Page 103 of 259

105 There is some blocking or missing pixels visible in the geocorrected imagery. This blocking is a result of rapid (relative to the image scan rate) aircraft pitch changes during data collection. These missing pixels are localized and in this case, primarily due to significant turbulence during acquisition. They make up less than 5% of the block and will not affect subsequent corrections performed on the data. Figure shows samples of CASI imagery over Demmin resampled to 3 x 3m pixels (3 bands approximating true color), after radiometric calibration and after the geocorrection process. Note on the positional accuracy of image data The acquisition of CASI-1500 imagery for the all missions was completed using the same installation on board the CASA-212. ITRES used the concurrent measurements recorded by the INTA IMU mounted atop the CASI-1500 SHU system. The geometric calibration flight for this installation was completed on July 6th, 2007 over Neubrandenburg, Germany. ITRES analysts used the offset parameters between the plane of reference (CCD) and the IMU to derive a solution for georeferencing of the imagery acquired during the AgriSAR and EAGLE missions which allowed for an accuracy of +/-2 pixels RMS. Due to the nature of the long integration times (96ms) between imager scanlines during the AgriSAR mission, some sudden changes in aircraft attitude could not be adequately compensated by the IMU data. In such cases, residual image effects (distortions) can be found in the scan lines related to such events. The incidence of such events is within tolerance vis-àvis the overall image accuracy Sentinel-2 Simulation (Uni Valencia) One objective of the AGRISAR campaign was the preparation of the Sentinel missions, the GMES component for future operational missions (follow-on of Landsat type of missions oriented to operational applications). The Sentinel missions are currently under definition. The baseline is a kind of Landsat-TM type of instrument, and the AGRISAR campaign was used to define enhancements in spectral capabilities for such Sentinel mission over current Landsat capabilities. In order to cover the SWIR and TIR spectral regions we used the INTA-AHS instrument as complement to the CASI-1500 hyper-spectro-radiometer. 18/01/2008 Page 104 of 259

106 Simulation of Sentinel-2 products by using images from CASI and AHS-INTA sensors Within the overall context of AGRISAR-2006 experiment, the available spectral information from the airborne imaging spectrometers (AHS-INTA and CASI) has allowed to simulate of Sentinel-2 products to evaluate their spectral bands capabilities by exploring the possibilities offered on retrieving vegetation properties, such as LAI (leaf area index), fractional vegetation cover, canopy water content and canopy chlorophyll content. For this simulation it has been used IDL programming software. We performed a program emulating the Sentinel-2 bands by adding images of CASI and AHS. In order to this, we had employed the nominal specifications for the future Sentinel-2 sensor, which could be updated in the nearest future, along with CASI images, used in 288 spectral bands mode (it can also work in 144 bands mode) between 371 nm and 1048 nm, and AHS-INTA used in 63 spectral bands mode (PT 1-2) between 455 nm and 2491 nm. For Sentinel-2 nominal band configuration, it was obtained images with 13 spectral bands as it is shown in the Table 11.3 Band Center Wavelength (nm) Spectral width (nm) Purpose Atmospheric correction (aerosol scattering) Sensitive to vegetation sensescing, carotenoid, browning and soil background; atmospheric correction (aerosol scattering) Green peak, sensitive to total chlorophyll in vegetation Max. Chlorophyll absorption Position of red edge; consolidation of atmospheric corrections / fluorescence baseline Position of red edge, atmospheric correction, retrieval of aerosol load LAI, edge of NIR plateau LAI 8a NIR plateau, sensitive to total chlorophyll, biomass, LAI and protein; water vapour absorption reference; retrieval of aerosol load and type Water vapour absorption, atmospheric correction. CASI & AHS for SENTINEL-2 SIM Based on CASI products Detection of thin cirrus for atmospheric correction. Out of range Sensitive to ligning, starch and forest above ground biomass. Snow/ice/cloud separation Assessment of Mediterranean vegetation conditions. Distinction of clay soils for the monitoring of soil erosion. Distintion between live biomass, dead biomass and soil, e.g for burn scars mapping. Table 11.3: Feasible spectral bands of Sentinel-2 Based on AHS products 18/01/2008 Page 105 of 259

107 CASI and AHS-INTA spectral bands were compatible in the different range of Sentinel-2 wavelengths, only the nominal band 10 (at 1350 nm) was not possible to be simulated as it is out of range, and the nominal band 11 (at 1610 nm) could only be related to the AHS band placed at 1622 nm, with a spectral width of 159 nm. The simulation program was divided in two parts, one related to the CASI images treatment and the second one to simulation data based on the AHS- INTA product. The spectral bands of CASI have been integrated to each one of the nominal spectral bands of Sentinel-2 that are placed within the same spectral width. Even if part of the spectral range of CASI is also covered by the AHS product, its band width can be considered larger than the one selected to the Sentinel-2 nominal bands in the VIS and NIR range. Thus, combined to the fact that this interval of AHS bands contains noise effects, points on using the CASI acquired images to simulation data. The Sentinel-2 nominal bands have been simulated by using a Gaussian filter. + = AHS-INTA 63 bands (PT 1-2) [ ,6] nm Sentinel-2 12 bands [ ] nm Figure 11.17: Simulated Sentinel-2 bands The simulation was completed by combining the two products from AHS and CASI (Figure 11.17). Before running the program, it was necessary to re-process and treated the CASI and AHS images selected to the simulation with the ENVI software to transform them to the same spatial resolution (1.5x1.5m/pixel) and projection (WGS-84). The program can be easily manipulated, if it would be needed, in case of futures changes in the spectral widths or the wavelengths configuration. 18/01/2008 Page 106 of 259

108 CASI-1500 Images processed: The CASI-1500 images were processed following the AGRISAR-participants demand and depending on the location of the sampled fields. Considering the large amount of ground measurements carried out on July 2006, and due to the difficulties to process the complete large-sized data, the UNI Valencia team decided to resize and provide to the rest of the participants only those images in which are included the main fields characterized during the 3rd intensive field campaign. The dataset was atmospherically corrected and the geo-rectification performed. Details of the dataset processed are reported intable /01/2008 Page 107 of 259

109 b c d e Figure 11.18: Sentinel simulation products from resized CASI data and different fields: a)sent_casi070506_288_p1_convert_resize_440_refl_geocorrec; b)sent_casi070506_288_p2_resize_450_440_convert_refl_geoc;c)sent_casi070506_288 _P4_Sub_460_REFL_geocorr; d)sent_casi070506_288_p5_resize_101_102_convert_refl_geocorr; e)sent_casi070506_288_p5_resize_140_refl_geocor; f)sent_casi070506_288_p6_sub_250_230_222_refl_geoco f 18/01/2008 Page 108 of 259

110 CASI processed images with 288 spectral band mode and 1,5m spatial resolution Resized flight line Size (GB) Fields in view Number of pixels Area (m2) CASI070506_288_P1_convert_resize_440_ REFL_geocorrec.img number of zones 20x20 m Sentinel simulation based in CASI with 10 spectral bands 443nm- 940nm and 1,5 m spatial resolution P Sent_CASI070506_288_P1_convert_resize_440_ REFL_geocorrec.img CASI070506_288_P2_resize_450_440_convert_ REFL_Geocorrec.img P Sent_CASI070506_288_P2_resize_450_440_convert_ REFL_Geocorrec.img CASI070506_288_P4_Sub_460_ REFL_geocorrec.img CASI070506_288_P5_resize_101_102_convert_ REFL_Geocorrec.img P (partial) 102(partial) 101(partial) P ,5 Sent_CASI070506_288_P4_Sub_460_ REFL_geocorrec.img Sent_CASI070506_288_P5_resize_101_102_convert_ REFL_Geocorrec.img CASI070506_288_P5_resize_140_ REFL_GeoCorrec.img P Sent_CASI070506_288_P5_resize_140_ REFL_GeoCorrec.img CASI070506_288_P5_Sub_102_ REFL_geocorrec.img P Sent_CASI070506_288_P5_Sub_102_ REFL_geocorrec.img CASI070506_288_P6_Sub_250_230_222_ REFL_geocorrec.img P Sent_CASI070506_288_P6_Sub_250_230_222_ REFL_geocorrec.img Table 11.4: Processed and simulated CASI and Sentinel-2 data. 18/01/2008 Page 109 of 259

111 Anomalous peak located at the CASI images spectral profile Once the CASI dataset was processed, a peak was located during a deep spectral analysis, within the [917;969] nm range, (bands 235 and 257). This peak is centred in 939 nm (band number 244) (figure 11.19). Figure 11.19: Anomalous peak located at CASI images As it is shown in figure 11.20, exploration of several spect from different surfaces, presented the same peak within the same range of wavelength and in the same direction. Only the height of the peak seemed to change with the reflectance of each surface. 18/01/2008 Page 110 of 259

112 Figure Different surfaces showing the anomally in the water vapour absorption. This range of the spectrum (939nm) corresponds to the water vapour absorption feature located in the NIR. After a first analysis of the peak, it was found two reasons that could explain it: An error introduced during the atmospherically correction process: due to the heterogeneous variability of the water vapour distribution, as it could be spread irregularly in the atmosphere. An error during the CASI sensor calibration: it could be done considering high water vapour concentrations conditions. Analysis of the peak The analysis of the peak height (h) (Figure 11.21) showed the spatial variation along the image. This variation is related to the water content distribution: when the spatial variability takes the shape of a cloud over the image it means the water vapour is distributed in different concentrations at the atmosphere, and the atm. correction must be repeated. Nevertheless, in the case in which the analysis shows a distribution similar to the original image, it indicates a homogeneous water vapour distribution, pointing on an overestimation of the water vapour absorption during the sensor calibration process. 18/01/2008 Page 111 of 259

113 Figure 11.21: representation of height of the peak. Figure 11.22: Height of the peak versus the reflectance in band 917nm Figure shows h versus the reflectance at 917 nm (band 235, in which the peak starts) taken as a reference. The height of the peak increases with the reflectance and it means the 18/01/2008 Page 112 of 259

114 surfaces with higher reflectance show the higher level for the peak. The plot shows a clear structure that also points to a bad calibration of the sensor. For the simulation of Sentinel-2 products, based on the images from CASI, this peak can be also located. Thus, it is important to see how the peak is affecting the Sentinel-2 simulated bands. Table 11.3 shows the spectral band configuration for Sentinel-2, the purpose of each band and the origin of the simulated products.the peak is located within the interval [917;969] nm, (spectral bands [235;257] of CASI images). The band number 9, at 940 nm with 20 nm of spectral width, is directly affected. The bands 8a (at 865nm with Δλ= 20 nm), 8 (at 842 nm with Δλ=115nm) are out of the peak s interval, so they are not affected after the simulation. Band number 10 (1375nm wavelength with Δλ= 20 nm) can not be simulated with CASI neither with AHS sensor because it is out of range. Bands 9 and 10, located at the wavelengths whose purpose is the detection of thin cirrus, should be simulated by means of tools as for example MODTRAN Atmospheric Data Sunphotometer CIMEL (direct sun irradiance) The raw direct measurements from the CE318 sun-photometer include contaminated data (affected by clouds) and not accurate sun pointing errors, or shadows. For removing these not valid data, the screening algorithm proposed by [ 57 ] is applied at the time of the AOD and water vapour retrievals. This algorithm consists of a set of rules for detecting fast AOD variations, not related to aerosol typical behaviour, but related to clouds. Cut-off values are also imposed for removing shadows produced by buildings and other objects from the surrounding. The raw data is transformed into AOD and water vapour by the application of a calibration. The uncertainty of the retrievals is directly linked to the calibration quality. For the data reported in this document, only a previous calibration has been applied (a pre-calibration), it is, level 1.5. As the instrument attended a calibration campaign in July 2006, a post-calibration is being also applied, so the final uncertainty of data will be considerably reduced. At the moment, comparisons between the UV-CIMEL and the FUB-ASA2 instruments for the AGRISAR campaign give a deviation of about This value is very similar to the estimated degradation of the CIMEL calibration, computed as an annual 2% degradation of the pre-calibration values. In Figure the comparison between both instruments is shown for three different channels. The only common channel is actually the 500 nm filter, but the other filters can give an idea of the uncertainties involved. As a final recommendation, the 1640 nm filter data is not valid, as it 18/01/2008 Page 113 of 259

115 usually presents instabilities and other technical problems, so it is not longer supported for this unit. Figure 11.23: Compared evolution of AOD for the three common wavelengths from the UV- CIMEL and FUB-ASA2 instruments. For the final AGRISAR database, the post-calibration is expected to be applied, and the CIMEL uncertainties should be reduced within 0.02 in optical depth and 0.2 cm in water vapour columnar amount. The post-calibration coefficients (July 2006) have been estimated to be less than 1% for most wavelengths. The exceptions are the ultraviolet channels: 340 nm (10%) and 380 nm (1.1%), as expected Sunphotometer CIMEL (sky radiance) For the radiance measurements of CIMEL, a different set of calibration values was applied. This calibration is usually performed every 6 months, by comparison of radiance measurements against calibrated radiance lamps (in our case, a Bentham SRS8 integrating sphere was used). The estimated uncertainty for the radiance measurements is 5-6% [ 20 ]. 18/01/2008 Page 114 of 259

116 From the sky radiance distribution, other complex aerosol properties (optical and physical) can be derived. For this we use the SKYRAD.PACK version 4 [ 17 ]. An accurate estimation of derived parameters estimation has not been yet published, but it is on progress. The inversion algorithm is in any case, very similar to that used by AERONET [ 19 ]. Although the raw sky data is also contaminated by clouds and shadows, a symmetry check is applied before the inversion code is applied. The almucantar solar plane is the plane defined when the sunphotometer is moved along variable azimuth angles, for a fixed zenital angle. The central point is the sun position, given by a zenital and azimutal angle. Two series of measurements are given, one for the left side from the sun, and the other for the right side. The quality filter consists of comparing, pair by pair, the symmetric measurements. Those pairs whose radiance values are more than 10% apart, are removed from the set. If the almucantar set has not enough data points (less than 22 points), the complete almucantar measurement is removed. Therefore, for cloudy skies, this method can not be applied Sunphotometer Microtops II The Microtops II sunphotometer main product is the columnar ozone, mainly used for correction of AOD of CIMEL. After several field campaigns performed by the UV-GRSV, where this instrument was compared to reference Brewer MKII spectrophotometers, an uncertainty in columnar ozone of 2% was estimated for the ozone value given with the wavelength pair 305/312. The corrected and 312/320 ozone values should not be used, because their uncertainty is higher and also less stable with time [ 21 ]. The AOD (at 1020 nm) has an estimated uncertainty of 0.04, but its uncharacterised and uncorrected temperature dependence makes this parameter not very reliable. The water vapour, on the other hand, has an estimated uncertainty of cm, after comparison with other CIMEL instruments Aureole Sunphotometer FUBISS The FUBISS instrument must be calibrated routinely for an optimum performance. For the sunphotometer measurements, three methods can be applied (Zieger et al., in press). In January 18/01/2008 Page 115 of 259

117 26 th 2006, an airborne calibration was performed. In July 2006, the calibration was done by a ground based Langley plot technique. An aureole calibration technique was also applied, with equivalent results to the Langley plot performed at ground. Due to the date proximity, this calibration is optimum for the analysis of the AGRISAR FUBISS database. The comparison of AOD data has been previously discussed (see Figure 11.23) for the case of CIMEL retrievals. A comparison of Angström wavelength exponent is also possible (see Figure 11.24). The Angström exponent is related to the size of particles, and is obtained by fitting the spectral behaviour of the AOD. Therefore, uncertainties accumulate. The uncertainty estimation is plotted in the figure by error bars for the case of the CE318. Figure Angström wavelength exponent comparison between CIMEL and FUBISS sunphotometric measurements Ground Radiometric Measurements Solar range ground radiometric data Three ASD field spectrometers have been available during two intensive ground measurements campaigns in June and July. One has been provided by the University of Valencia in Spain, one from the Friedrich-Schiller-University in Jena and one from the University of Naples in Italy. The main purpose of these measurements was to obtain sufficient representative spectra from different surface types of the test area. These spectra were used for calibration and validation of the atmospheric correction of airborne and spaceborne hyperspectral images. The instrumentation used was: ASD FS/FR spectro-radiometer, radiometrically calibrated from 350 to 2500 nm. Spectralon white reference panel. 18/01/2008 Page 116 of 259

118 Leaf Transmittance/Reflectance portable dark chamber. Garmin Gecko GPS, connected to the radiometer. Cal/Val measurements were taken following this pattern: first a static surface radiance measurement consisting of five consecutive spectra, with white reference measurements before and after to allow assessment of illumination stability. Then surface radiance was measured continuously while walking to the next stop, in this way each spectrum collected corresponds to the integration over a stripe of surface around 5m long, which would better correlate with airborne measurement of similar footprint. The pattern is repeated until the field is characterized. For validation activities the time span before and after flight overpass had to be not larger than one hour in order to have similar illumination conditions while allowing enough time to cover different reference surfaces. The tight time window makes that the different fields usually cannot be fully covered; therefore the strategy followed was to try to cover the larger variability possible within each field. Later GPS positioning allowed accurate collocation with imaging data. 06/06* 07/06 08/06 10/06* Figure11.25: Spectroradiometric measuring points (from GPS) during the June campaign. 18/01/2008 Page 117 of 259

119 Date 06 June Optical Flight 07 June 08 June 10 June Optical Flight Target Concrete Wheat 230 Grass Corn 222 Sugar Beet 102 Concrete Rape 101 Sugar Beet 102 Fallow Barley 240 Wheat 230 Dry Grass Concrete Type Cross calibration Cal/Val Cal/Val Cal/Val Cal/Val Cal/Val Cal/Val Cal/Val Leaf Refl/Trans Cal/Val Cal/Val Cal/Val Cal/Val Cal/Val Notes at Farm at Soccer Field at Farm GPS failed Chl data Next to 130 with Thermal GPS failed 04/07* 05/07* 06/07 Figure 11.26: Spectroradiometric measuring points (from GPS) during the July campaign. 18/01/2008 Page 118 of 259

120 Date Target Type Notes 04 July Optical Flight Concrete Cross calibration Cal/Val at farm GPS failed Barley 440 Cal/Val GPS failed Corn 222 Cal/Val Scintillometer transect GPS failed 05 July Optical Flight Wheat 250 Cal/Val Scintillometer transect Next to Bowen Corn 222 Cal/Val Leaf Refl/Trans Scintillometer transect 06 July Sugar beet 102 Leaf Refl/Trans Corn 222 Leaf Refl/Trans Soil moisture transect Leaf Reflectance and Transmittance Measurements: A new portable device has been designed and built for field measurements of leaf reflectance and transmittance, without the need to cut the samples for laboratory measurement. The device consists in a dark chamber with a leaf clip to hold the sample; an opening for solar illumination with an alignment mark to keep the sample fully lighted; and a detachable support for the spectrometer s fiber optics that can be positioned above the leaf for reflectance measurement or below for transmittance measurement while the opposite side is tightly closed to act as a dark background. The device is also prepared for direct measurement of fluorescence emission by means of a lowpass 650nm cut-off filter that can be placed at the illumination port in order to remove the incoming light at the fluorescence emission range, while maintaining almost the same illumination level in the PAR region. This device was successfully tested during AgriSAR field campaign, measuring leaf spectral reflectance/transmittance, and upwelling/downwelling fluoerescence emission spectra of two different crops: sugar beet and corn. 18/01/2008 Page 119 of 259

121 Fiber Optic Filter Leaf Clip Dark Background Device for leaf reflectance, transmittance and fluorescence emission. Figure 11.27: Spectra from a corn leaf simple: reflectance and transmítanse (top); upwelling and downwelling fluorescence emission (bottom) Soil and Vegetation Data Continuous data Data quality analysis from the data acquisition of DLR-DFD Table 11.5 gives an overview of all field data sampled by DLR-DFD team in close cooperation with the ZALF. Parameters and the number of corresponding measurements per sampling unit for weekly sampling in the field must be seen as a trade-off in many respects. That is a maximum of ground data to represent field heterogeneity best possible but restricted personal resources and time, as well as physical parameter availability in the course of time for destructive methods (cutting of biomass). The fact that they base on one, two or three measurements per sampling unit always has to be kept in mind when interpreting data. An evaluation of data quality hence confines to a check of plausibility and/ or basic descriptive statistics. 18/01/2008 Page 120 of 259

122 Parameter Specification/ Measuring device Measurements sampling unit per Ground team Gravimetric soil moisture Crop coverage Vegetation height At 2 depths: 0-5cm and 0-10cm Field estimate and picture Photograph in front of gridded backboard; Measurement with ruler 3 at each depth DLR-DFD-NZ 1 DLR-DFD-NZ 1 DLR-DFD-NZ Biomass Wet, dry 2 ZALF Phenology BBCH phase 1 ZALF Vegetation density Plants per m², shoots per m 2 ZALF Chlorophyll SPAD-502 DL 30 ZALF LAI LICOR LAI ZALF Table 11.5: Overview of parameters sampled by DLR-DFD and ZALF during AgriSAR Soil moisture Volumetric soil moisture was averaged out of 3 measurements at each sampling unit and depth. For gravimetric measurements in the lab, a balance of 10mg accuracy was used. Considering the entire sampling period, maximum soil moisture values were detected on 19 April, on field 222_8 (maize). Minimum values were measured on 26 July on field 460_23 (sugar beet), both at 0-5cm depth. Disregarding absolute values, figure displays a similar timedependent pattern of soil moisture for all sampling locations which is particularly distinct for shallow depths. There is a general downward drift of soil moisture values from the start of campaign in April to the end in beginning of August. Although moisture values do not fall constantly but drop down on 10 May and 13 June and jump up the following weeks. This feature coincides with heavy rainfall after these dates saturating the soil with water. Soil moisture contents within the barley, rape and sugar beet field vary only little between sampling points. In contrast, sampling points 8 and 15 on field 222 (maize) and field 230 (wheat) respectively, show perspicuously higher water contents at all dates due to their position in shallow depressions with differing local soil conditions. No data is available for 12 July for sampling points 22, 23 and 24 on the sugar beet field because of fertilizes applications. 18/01/2008 Page 121 of 259

123 Figure displays moisture differences between the two sampled depths for each sampling unit over time. Error bars indicate maximum standard deviations. Moisture contents on field 140 (rape) vary little between depths. Differences are less than 5% and can be neglected considering standard deviations. For field 230 (wheat) and field 450 (barley) differences are up to 10% whereas soil moisture is higher in the upper 5 cm of the soil until 21 June. For subsequent dates, the situation changes which is consistent with increasing temperatures and a dry period. For field 222 (maize) and field 460 (sugar beet) differences are predominantly positive indicating up to 15 % higher water contents at 5-10cm as a result of direct surface exposure to evaporation. Crop coverage is estimated less than 50% until end of July (Fig ). To give a measure of quality for the presented soil moisture data, the variation coefficient was calculated for each sampling unit, depth and date and plotted against mean values over time σ VC = 100 Mean Figure displays mean and maximum values of VC for all samples. Variation between the three samples is on average less than 10% for most of the data sets. An increased mean VC is observed on 5 July and 2 August but there are outliers in the entire data set as pointed out by the maximum curve. To identify these outliers, the variation coefficient is provided along with all soil moisture data files. The following possible sources of error were identified: 1) local inhomogeneities due to soil, shadow and micro relief, 2) disturbed sample tubes due to stones or roots, 3) consistence of soil probe (texture, wetness), 4) human induced errors. Error-proneness of sampling raises the lesser the absolute soil moisture values. 18/01/2008 Page 122 of 259

124 Barley 450_1 450_2 450_ Maize 222_7 222_8 222_ Soil moisture (%) Soil moisture (%) Day of sampling Day of sampling Wheat 230_13 230_14 230_ Sugarbeet 460_22 460_23 460_ Soil moisture (%) Soil moisture (%) Day of sampling Day of sampling Rape 140_25 140_26 140_ Soil moisture (%) Day of sampling Figure 11.28: Soil moisture at 0-5cm. Error bars indicate standard deviations. 18/01/2008 Page 123 of 259

125 Barley 450_1 450_2 450_ Maize 222_7 222_8 222_10 Soil moisture (%) Soil moisture (%) Day of sampling Day of sampling Wheat 230_13 230_14 230_ Sugarbeet 460_22 460_23 460_24 Soil moisture (%) Soil moisture (%) Day of sampling Day of sampling Rape 140_25 140_26 140_ Soil moisture (%) Day of sampling Figure 11.29: Difference between soil moisture at 5-10cm and 0-5cm. Error bars indicate maximum standard deviations. 18/01/2008 Page 124 of 259

126 80 70 Max. VC 60 Mean VC VC (%) Figure 11.30: Mean and maximum variation of coefficient for all soil samples during AgriSAR campaign. Biomass Biomass was assessed by twice cutting one square meter field crop and a gravimetric measurement before and after oven-drying at 70 C. Percentage of vegetation dry mass was calculated as the ratio of dry biomass (kg/m²) and wet biomass (kg/m²) multiplied by 100. Figure shows mean values for vegetation dry mass (%) and dry biomass for each sampling location over time. Dry biomass is constantly growing for all winter crops until end of June beginning of July. Barley and rape exhibit a slight decrease on the following measurements whereas wheat continues to grow. There is no data for barley on 26 July because of harvest. Dry biomass for maize and sugar beet are generally lower due to late sowing and different stage of phenology during the campaign. Both steadily increase until end of sampling. Biomass values for 222_8 indicate a strong increase towards end of July which may be closely related to plantavailable water and soil conditions at this location (see soil moisture). There was no sampling on the sugar beet field on 12 July due to ongoing fertilizer applications. Chlorophyll About 30 chlorophyll measurements with the SPAD-502 were made on different leaves to obtain a good average for each sampling unit. Chlorophyll DC values vary only little for within field sampling units except unit 222_10. Standard deviations constitute about 10% on average with respect to mean values. Measurements for barley on 28 July are subject to higher variation (Fig.11.32). To detect outliers, again variation coefficient will be provided with the data set. The temporal evolution of chlorophyll DC for different crops vary but as a general rule either constantly grow or decrease. On field 230 (wheat) values oscillate which may be explained by plant water stress but has to be dealt with care. Transformation of chlorophyll DC into chlorophyll concentrations (µg/cm²) should be done with the calibration function for SPAD A supplied by Soledad Gandia from University of Valencia. 18/01/2008 Page 125 of 259

127 Leaf Area Index Two LAI measurements were made at each sampling unit where single values are an average out of 4 measurements performed by the instrument. The standard error of LAI (SEL) provided with the instrument is on average less than 10% with respect to mean values. There are some measurements on field 230 (wheat) with higher deviations (Fig.11.33). Absolute LAI values range from 1.1 to 5.3 for winter crops and 0.2 to 3.5 for sugar beet and maize. LAI values at individual sampling dates and units vary only little per field that is crop type. On 26 July there is no LAI measurement due to harvest Barley Biomass, dry (kg/m²) 450_1 450_2 450_3 Veg. Drymass (%) 450_1 450_2 450_ Maize Biomass, dry (kg/m²) 222_7 222_10 222_8 Veg. Drymass (%) 222_7 222_8 222_10 80 Biomass, dry (kg/m²) Veg. Drymass (%) Biomass, dry (kg/m²) Veg. Drymass (%) Biomass, dry (kg/m²) Wheat Biomass, dry (kg/m²) 230_13 230_15 230_14 Veg. Drymass (%) 230_13 230_14 230_ Veg. Drymass (%) Biomass, dry (kg/m²) Sugarbeet Biomass, dry (kg/m²) 460_22 460_23 460_24 Veg. drymass (%) 460_22 460_23 460_ Veg. Drymass (%) Rape Biomass, dry (kg/m²) 140_25 140_27 140_26 Veg. drymass (%) 140_25 140_26 140_ Biomass, dry (kg/m²) Veg. Drymass (%) 20 0 Figure 11.31: Mean vegetation dry mass (%) and dry biomass. 18/01/2008 Page 126 of 259

128 Barley 450_1 450_2 450_ Maize 222_7 222_10 222_8 Chlorophyll SPAD DC Chlorophyll SPAD DC Wheat 230_13 230_15 230_ Sugarbeet 460_22 460_23 460_24 Chlorophyll SPAD DC Chlorophyll SPAD DC Rape 140_25 140_27 140_ Chlorophyll SPAD DC Figure 11.32: Chlorophyll SPAD502 DC values. Error bars indicate standard deviations. 18/01/2008 Page 127 of 259

129 7 7 6 Barley 450_1 450_2 450_3 6 Maize 222_7 222_10 222_ LAI LAI Wheat 230_13 230_15 230_14 6 Sugarbeet 460_22 460_23 460_ LAI LAI Rape 140_25 140_27 140_ LAI Figure 11.33: LAI values with corresponding standard error of LAI. Other parameters measured on ground to describe canopy characteristics were crop coverage, vegetation height, plant phenology and crop density. 18/01/2008 Page 128 of 259

130 Crop coverage was estimated in the field in percentage vegetation cover for each sampling unit. As shown in figure 11.34, estimates are strongly individual-related varying partly over 20% in the beginning of the sampling period. Still coverage estimates deliver a first fast impression of coverage conditions. The data is supported by digital photographs, taken with a wooden frame of 0.5x0.5m for objective documentation and subsequent estimation by means of image analysis. Vegetation height was measured with a rule, additionally supplied with a photograph in front of a gridded backboard. To measure plant phenology, BBCH stages were described once per sampling unit. Finally crop density was measured twice per sampling unit by counting plants per meter _ _2 450_3 222_7 Crop coverage (% _8 222_10 230_13 230_14 230_ _22 460_23 460_ _04_06 03_05_06 10_05_06 17_05_06 24_05_06 07_06_06 13_06_06 21_06_06 05_07_06 12_07_06 26_07_06 02_08_06 140_25 140_26 140_27 Day of Sampling (DD_MM_YY) Figure 11.34: Estimated crop coverage during AgriSAR campaign. Crop IDs indicate: Barley=450, Maize=222, Wheat=230, Sugar beet=460, Rape= Continuous data by CAU Tables 8.1 and 8.2 give an overview of all field data sampled by the teams of Christian- Albrechts-University of Kiel. Samples were collected simultaneously to E-SAR flights (i.e. on a weekly basis), except for May 11 th 2006 and June 13 th 2006, when sampling dates differ by 1 day. The sampling strategy was setup to represent the typical crops in the investigation area and period. In addition, two automatic soil moisture stations (TDR stations) were installed on field 101 (winter rape) and field 102 (sugar beet). Due to the small number of replicates a quality check of the sampled data is limited to basic statistics and/or plausibility test Soil moisture Volumetric soil moisture was calculated out of three measurements at each elementary sampling unit (ESU) in two depths (0-5 cm and 5-10 cm) using standardized 100 cm 3 Kopecky Rings (DIN 18/01/2008 Page 129 of 259

131 ISO 11272). For gravimetric measurements a Kern 572 field weighing machine was utilized with an accuracy of 0.01 g. Samples were weighted direct in the fields and in the lab after oven-drying at 105 C. Soil moisture measurements have been carried out for each ESU successfully during the campaign, except for ESU 19, 20, 21 on the 12 th of July due to pesticide applications. In addition, no soil moisture data is available for ESU 21 on the 06 th of July due to unclarified measurement inaccuracies. Considering the whole investigation period, a decrease in soil moisture can be observed which is more pronounced in the shallow soil layer (0-5 cm) due to the direct exposure to soil evaporation while the deeper measurements show a more buffered decrease (Figure 11.35). This decrease is strongly related to the lack of precipitation. Only a few peaks can be observed shortly after rainfall during the sampling period. For a quantitative evaluation of the delivered data the coefficient of variation (cov), as defined below, was calculated for each ESU, sampling depths and date. cov = σ 100 _ x Figures show the mean and maximum cov for all samples in both depths. Variations between the samples lie within an average of 15 %. An increase in the mean cov for both depths towards the end of the campaign can be observed which is related to the drier soil conditions at the end of the sampling period; it can be stated that the drier the soil the higher the errorproneness of moisture sampling. This is also proved by the fact that even after short rainfalls the mean cov for all samples decreases. There are also some extreme outliers in the entire dataset as indicated in the maximum cov values in figure The following error sources can be identified: 1) spatial inhomogeneities in soil water content within the soil, 2) disturbed sample due to stones and or roots within the Kopecky Rings and 3) human induced errors. 18/01/2008 Page 130 of 259

132 soil moisture [vol.%] Date 35 soil moisture [vol.%] Date Figure 11.35: Temporal evolution of soil moisture in a) 0-5 cm and b) 5-10 cm 18/01/2008 Page 131 of 259

133 mean cov max cov mean cov max cov Figure 11.36: Mean and maximum coefficient of variation in percent for above) 0-5 cm and below) 5-10 cm soil moisture samples Continuous TDR Stations For the whole investigation period two automatic soil moisture stations were installed on field 101 (WR) and field 102 (SB). Soil moisture was sampled at 2 depths (5cm and 10cm ) at three locations per station with a 10 min sampling interval. Figure and display the TDR measurements for each field. As can be seen there are several gaps within the time series. For the winter rape field 101 there was a malfunction of the data logger between the The no data values for the TDR probe F32 (-0.05) between the are due to unspecified vandalism. The no data values from can be addressed to malfunctions of the data logger system. 18/01/2008 Page 132 of 259

134 For a quantitative assessment of the delivered data, the TDR probes have been calibrated in the lab. The TDR probes installed in the winter rape field showed a mean overestimation of 3.8 Vol% water content compared with standardized Kopecky Rings while those within the sugar beet field showed a mean overestimation of 3.45 Vol.%. A qualitative comparison between the TDR measurements and the mean field soil moisture averaged out of the Kopecky measurements for each field showed good agreement. Figure 11.37: Temporal evolution of continuous TDR measurements on field 101 Figure 11.38: Temporale evolution of continuous TDR measurements on field /01/2008 Page 133 of 259

135 Leaf Area Index LAI Five LAI measurements were performed at each ESU using the LICOR Each single measurement is an average of four single measurements. The mean standard error of LAI (SEL) which is calculated and provided automatically by the instrument lies within 22% with a maximum SEL of 54%. The LAI for the entire data ranges from 0.92 to 6.29 with a mean of For field 102 LAI was only measurable for the last campaign date due to late development of the sugar beets and pesticide application on July 12 th There are no LAI measurements for the winter barley field 440 at the due to harvest. 7 6 LAI Figure 11.39: Temporal evolution of mean LAI and mean SEL for sampled fields Photogrammetric soil surface roughness To capture soil surface roughness dynamics, soil surface roughness was determined using photogrammetric image matching techniques. For image acquisition a calibrated Rollei d7 metric camera with known interior orientation was used. To give a measure of data quality the photogrammetric system allows a quantitative evaluation of quality of the delivered data due to high accurate control points that were installed on the system. The mean standard deviation in elevation (z) related to the control points lie within 1.63 mm and a mean absolute error for z of 1.2 mm. Maximum error is 10.1 mm. The mean positional accuracy is better than a pixel and within the range of 0.37 mm in xy. Soil surface roughness measurements have been carried out on each ESU with two replicates. On the following dates roughness determination was not successful for all ESU: 18/01/2008 Page 134 of 259

136 Date Error sources No data available for ESU 6,7,8,10,16,17, No data available for ESU 4, Camera malfunction Pesticide application on field 102 and Harvest on field 440 Table 11.6: Error source for no data values Biomass Biomass was measured by cutting one square meter field crop and weighing Biomass before and after oven-drying at 70 C with two replicates. For weighing a Kern CB24K1N weighing machine was utilized with an accuracy of 1 g. There is no data available for the sugar beet field 102 until May 24 th 2006 and on July 12 th 2006 due to the late sowing date and ongoing pesticide applications. In addition, no data is available for the on field 440 because of harvest. Figure and show the differences between the two wet biomass replicates that were measured in the fields. Most replicates differ only within 1 kg/m 2 wet biomass for the fields 440 (WB) and 250 (WW). In contrast, the difference between each replicates on field 102 and 101 is much higher, especially for field 101 with a maximum difference in wet biomass of 3.37 kg/m 2. 1, ,00000 [kg/m²] 0, , , , , , , Figure 11.40: Difference between the two wet biomass replicates at each ESU for field 440 and 250 [wet biomass kg/m²] 18/01/2008 Page 135 of 259

137 [kg/m²] 4, , , , , , , , , Figure 11.41: Difference between the two wet biomass replicates at each ESU for fields 102 and 101 [wet biomass kg/m²] Vegetation height, crop coverage and phenology Additional parameters like vegetation height, crop coverage, phenology and vegetation density were measured in the fields to characterize vegetation physiology an structure. To measure vegetation height, two different approaches were utilized. First measuring the heights with a ruler and in addition with digital imaging techniques in front of a gridded board for height estimation and structural characterization. All data has been made available through the ftp- server. Crop coverage was measured using standardized digital photography since estimation by a non objective system (e.g. the field worker) leads to strong data uncertainties. Crop coverage images were provided for each date and ESU trough the ftp- server. Plant phenology was estimated using the BBCH stages. Phenology was estimated on each date once per sampling unit and is provided to the users through the ftp- server Intensive data Intensive campaign sampling by University of Valencia Crop Photosynthetic Pigment Composition and Calibration of an Instrument for Indirect Chlorophyll Content Determination Chlorophylls and carotenoids, the higher photosynthetic pigments in the plant, are essential for plant growth and development. They are responsible of gathering the necessary sunlight for leaf photosynthesis process. Quantifying photosynthetic pigments in agricultural crops and natural vegetation is important to assess their physiological state. Crop and natural vegetation stress detection at an early stage 18/01/2008 Page 136 of 259

138 are important for precision agriculture and forest management. With the development of hyperspectral remote sensing techniques during the 1980s, it has become possible to quantify photosynthetic pigments in extensive agricultural crop areas. Two SPAD-502 (Minolta, Osaka, Japan) devices (named SPAD-A and SPAD-B) were used to obtain in-situ non destructive chlorophyll measurements from several ESUs. In order to convert digital counts, acquired from SPAD devices, to absolute chlorophyll values, it was necessary an instruments calibration. For these purposes, SPAD-A was calibrated during the SEN2FLEX field campaign [2,9] and SPAD-B has been calibrated within AGRISAR-2006 experiment: Weekly measurements were carried out with SPAD-A device on selected crops by the Leibnitz- Zentrum für Agrarlandschaftsforschung (ZALF) team along the crops vegetative cycle. In-situ chlorophyll measurements were also obtained by means of the SPAD-B device and by the Valencia University team along the three intensive AGRISAR-2006 campaigns. Leaves Samples from different crops were collected in order to make high performance liquid chromatography (HPLC) analysis of leaves pigments. Samples from different points of six corn plants were gathered in order to analyse pigments distribution in corn plants: Chlorophylls (Chl a and b), major (lutein and β-carotene) and minor (neoxanthin, violaxanthin, antheraxanthin and zeaxanthin) carotenoids were quantified. Chlorophyll determinations were used to calibrate SPAD-B device. Within this framework, the main objectives are: To calibrate the SPAD-502 handheld chlorophyll meter (SPAD-B device) in order to obtain absolute chlorophyll measurements. To determine representative statistical values of chlorophyll content for different crops from several ESUs. To analyze temporal chlorophyll content variability found at selected crops. To fully characterize accurately the photosynthetic pigment composition of wheat, sugar beet, barley and corn. To study the different relationships found between the photosynthetic pigments values. To analyze the pigments distribution in corn plants. Extraction and Analyses Photosynthetic pigments are located in the thylakoid membrane in a lipidic environment, but thylakoids are surrounded by an aqueous medium. Therefore, chlorophylls and carotenoids are extractable from plant leaves only by organic solvents able to mix with the water contained in the plant tissue. Acetone and methanol, which are water-miscible and easy to handle, are used most often. 18/01/2008 Page 137 of 259

139 Plant tissues have very carefully compartmentalized phs. However, when the plant tissue is ground, the compounds stored previously in cellular compartments (such as the low-ph vacuoles) are incorporated into the extract and may lower significantly the ph. It is therefore desirable to add a buffering agent to the grinding medium. Calcium carbonate and sodium ascorbate are probably the most common chemicals used in the literature for this purpose. Intense light should also be avoided when extracting photosynthetic pigments from plant tissues, although normal laboratory light levels do not create problems. The most important effect of intense light (sunlight) seems to be the conversion of trans-neoxanthin in cis-neoxanthin. Therefore, our extraction procedure was as follows. Leaf samples were taken, wrapped in aluminum foil, dropped in liquid nitrogen, and stored (still wrapped in foil) at 20 ºC. A few ml of bulk acetone, neutralized with calcium carbonate, are put in a mortar and a pinch of sodium ascorbate is added. A small known amount of leaf tissue (0.785 cm2 in our case) is dropped in the mortar and ground. The mixture is then poured into a volumetric flask and acetone is added until the desired volume is attained (5 ml). The mixture is filtered through a 5 µm Millipore filter. The whole extraction process takes approximately 2 min per sample. The filtered extract was stored in foil-wrapped plastic tubes at 20 ºC until analysis. Analyses of HPLC of photosynthetic pigment extracts, obtained with this procedure, indicate that significant decomposition of pigments does not occur (results not shown). Once extractions have been performed, the next step was measuring the concentration of the photosynthetic pigments, which were carried out by using a HPLC separation system. Pigment extracts were thawed on ice, filtered through a 0.45 µm filter, and analyzed by an isocratic HPLC method based on that developed by de las Rivas et al. Instead of three, two steps were used: mobile phase A (acetonitrile:methanol, 7:1 v:v) was pumped for 3.5 min, and then mobile phase B (acetonitrile:methanol:water:ethylacetate, 7:0.96:0.04:8 by volume) was pumped for 4.5 min. To both solvents 0.7% (v:v) of the modifier triethylamine (TEA) was added to improve pigment stability during separation. All chemicals used were HPLC quality. The column was equilibrated before injecting each sample by flushing with mobile phase A for 5 min. The analysis time for each sample was 13 min, including equilibration time. Results in this work have been expressed in an area basis (µmol m -2 ), which is widely used within the photosynthesis research and plant physiology community. Within the remote sensing community, it is frequent to express results as µg cm -2, since (some) remote sensing models require such units for chlorophyll concentration as inputs. For conversion, the following molecular weights must be used: Chl a (893.53), Chl b (907.51), neoxanthin (600.9), violaxanthin (600.9), antheraxanthin (584.9), lutein (568.9), zeaxanthin (568.9), and β-carotene (536.85). Results from data quality analysis: Calibration functions A linear function has provided the best fitting for the experimental calibration points for both of the instruments, SPAD-A (SEN2FLEX-2005) and SPAD-B (AGRISAR-2006). The function has been forced to cross by origin, in order to be sure the SPAD DC measurement becomes zero for 18/01/2008 Page 138 of 259

140 those samples without chlorophyll. SPAD-A calibration from SEN2FLEX-2005 experiment (Gandia et al., 2006) follows the calibration function given by (1) (Figure 11.42A) with the associated error: -2 Chl (μg cm ) = * DC (1) ε ( Chl ) = DC * 2.3 * * ε (DC) (2) The calibration function and the associated error found for SPAD-B (Fig B), are given by (3) and (4): Chl (μg cm -2 ) = * DC (3) ε ( Chl ) = DC * 6.4 * * ε (DC) (4) An instrument calibration by species has been also performed and the results from the calibration functions applied are shown in Figure Better correlations coefficients are found when considering species separately, being their slopes very similar and to the slope obtained from analyse all the species together. Weekly chlorophyll measurements A total amount of 15 ESUs from five crops (three ESUs for crop, selected as representative of the state for the crop chlorophyll content) were measured by ZALF team along their vegetative cycle (SPAD-B). Each ESU was GPS-located considering as its chlorophyll the mean value, obtained from 30 the chlorophyll measurements acquired from different leaves in the same ESU, with the standard deviation error of the measurements. The measurements for each ESU were carried out within an area of about 30 m. The temporal evolution found for the different crops is shown in Figure In this, it can be observed a normal evolution for the chlorophyll content starting to increase from the beginning of the cycle and decreasing after arrive to a maximum value. However, wheat shows a different behaviour due to the plant stress conditions. Figure illustrates the ESUs measured weekly along with the ESUs measured within the 2nd and 3rd intensive field campaigns. 18/01/2008 Page 139 of 259

141 Figure11.42: A) Calibration function for SPAD-A; B) Calibration function for SPAD-B Figure 11.43: Calibration functions results after considering species separately. 18/01/2008 Page 140 of 259

142 In situ chlorophyll measurements within the June and July Intensive Campaigns (SPAD-B) Due to the complex leaf samples collection procedure and also because ZALF team were measuring chlorophyll each week, only a few in-situ chlorophyll content measurements were acquired by UV team. Table 11.7 and Table 11.8 show mean chlorophyll values with their standard deviation error for the ESUs from the 2nd and 3rd intensive field campaigns, respectively. Crop / ESU Code utme June Campaign Mean SPAD ± SE utmn (DC) Mean Chl ± Error (μg cm -2 ) Wheat ± ± 0.6 Wheat ± ± 1.2 Wheat ± ± 1.3 Wheat ± ± 0.8 Sugar beet ± ± 0.4 Barley ± ± 0.5 Table 11.7: 2nd AGRISAR intensive campaign chlorophyll measurements. Figure 11.44: Temporal evolution found for chlorophyll from crops weekly measured. 18/01/2008 Page 141 of 259

143 Figure 11.45: ESUs weekly measured (ZALF) combined to the intensive measurements (UV). July Campaign Mean SPAD ± Crop / ESU Code utme utmn SE (DC) Mean Chl ± Error (μg cm -2 ) Corn ± ± 0.8 Corn ± ± 0.7 Corn ± ± 1.1 Corn ± ± 0.9 Corn ± ± 0.7 Wheat ± ± 0.7 Wheat ± ± 0.7 Sugar beet ± ± 1.0 Sugar beet ± ± 0.9 Sugar beet ± ± 0.9 Sugar beet ± ± 0.5 Table 11.8: Chlorophyll measurements during the July AGRISAR campaign. HPLC data: Chlorophylls Table 11.9 presents values obtained for Chl a, Chl b, and the Chl a/chl b ratio, from the 4 species under study. Chl a values varies from 261 µmolm -2 for sugar-beet to 335 µmolm -2 for wheat, whereas the Chl b range goes from 60 µmol m -2 for sugar-beet to 91 µmol m -2 for wheat. The Chl a/chl b ratio values varied from 3.5 in barley to 4.4 in sugar beet. 18/01/2008 Page 142 of 259

144 HPLC data: Major Carotenoids Major carotenoids found for all species were lutein and β-carotene (Table 11.10). Concentration for lutein from wheat and sugar-beet was higher than β-carotene concentration; getting similar values from barley. Whereas the concentration of β-carotene observed for corn was higher than lutein level (Table 11.10). Wheat merits a special remark, since the concentration of violaxanthin in this case is similar to lutein and β-carotene levels (Tables and 11.11). Species Chl a Chl b Chl a/chl b Wheat 335 ± ± ± 0.3 Sugar beet 261 ± 8 60 ± ± 0.3 Barley 310 ± ± ± 0.3 Corn 294 ± ± ± 0.3 Table 11.9: Chlorophyll a (Chl a, µmol m-2), chlorophyll b (Chl b, µmol m-2), and chlorophyll a/ chlorophyll b ratio (Chl a/chl b) in the 4 species sampled in AGRISAR Data are mean value ± SE. N = samples per species. Species Lutein β-carotene Wheat 66.7 ± 3 62 ± 3 Sugar beet 58.1 ± ± 1.7 Barley 54 ± 3 60 ± 3 Corn 51 ± 2 56 ± 3 Table 11.10: Major carotenoids lutein and β-carotene (µmol m-2) in the 4 species sampled in AGRISAR Data are mean value ± SE. N = samples per species. 18/01/2008 Page 143 of 259

145 HPLC data: Minor Carotenoids Table shows values obtained for the minor carotenoids from the 4 species under study. For most of the cases violaxanthin was the minor carotenoid found in highest concentrations (in an area basis), followed by neoxanthin, antheraxanthin and zeaxanthin.the case of Barley was the exception, because zeaxanthin was found to be in higher concentration than antheraxanthin. Species Neo V A Z Wheat 25.5 ± ± ± ± 0.9 Sugar beet 17.5 ± ± ± ± 1.6 Barley 23.3 ± ± ± ± 1.5 Corn 16.3 ± ± ± ± 0.6 Table 11.11: Minor carotenoids neoxanthin (Neo), violaxanthin (V), antheraxanthin (A), and zeaxanthin (Z) (µmol m-2) in the 4 species sampled in AGRISAR Data are mean value ± SE. N = samples per species. SPARC-2004 and SEN2FLEX-2005 vs. AGRISAR-2006 Comparing data from SPARC-2004 and SEN2FLEX-2005 experiments to those from AGRISAR- 2006, the 2 species present in the three campaigns, sugar-beet and corn, showed the lowest chlorophyll concentration for AGRISAR-2006 (Table 11.11). The A+Z/(V+A+Z) ratios took the lowest values (AGRISAR-2006) for corn, being in the same range to the SPARC-2004 for sugarbeet (Table 11.12). 18/01/2008 Page 144 of 259

146 Total Chl Species SPARC SEN2FLEX AGRISAR Sunflower 461 ± ± Corn 490 ± ± ± 15 Alfalfa 448 ± ± Sugar beet 482 ± ± ± 10 Potato 249 ± ± Onion 380 ± ± Garlic 250 ± ± Vineyard 420 ± ± Quercus ilex ± Table 11.12: Comparison of total chlorophyll (Total Chl) concentrations (in µmol m-2) in the species sampled in Barrax area during the SPARC (2004), SEN2FLEX (2005) and AGRISAR (2006) campaigns. Data are mean value ± SE. N= 15-26, 20-21, and samples per species in 2004, 2005, and 2006 respectively. 18/01/2008 Page 145 of 259

147 Species A+Z/(V+A+Z) SPARC SEN2FLEX AGRISAR Sunflower 0.64 ± ± Corn 0.41 ± ± ± 0.02 Alfalfa 0.52 ± ± Sugar beet 0.45 ± ± ± 0.04 Potato 0.69 ± ± Onion 0.21 ± ± Garlic 0.70 ± ± Vineyard 0.44 ± ± Quercus ilex ± Table 11.13: Comparison of A+Z/(V+A+Z) ratios in the species sampled in Barrax area during the SPARC (2004), Sen2Flex (2005) and AGRISAR (2006) campaigns. Violaxanthin, V; Antheraxanthin, A; Zeaxanthin, Z. Data are mean value ± SE. N = 15-26, 20-2 and samples per species in 2004, 2005, and 2006 respectively. Relationship between Photosynthetic Pigments for Corn Leaves at Different Height Levels Corn samples were taken from 6 plants at 2 different height levels (leaves No. 2 and 6 from the base) from each plant, and at 3 different points along each leaf, looking for analyzes pigment distribution over the plant. Data analysis show similar plant distribution for some photosynthetic pigments. For instance, there were not found remarkable differences for Chl a, Chl b, neoxanthin, lutein and β-carotene (not shown). However, for the case of violaxanthin, antheraxanthin and zeaxanthin, the 3 carotenoids involved in the xanthophylls cycle, it has been observed higher pigment concentrations with increasing distances from the stem. Figure illustrates this question in the case of zeaxanthin. 18/01/2008 Page 146 of 259

148 Each one of the photosynthetic pigments has been also plotted versus the rest of them, in order to find relationships between them. A linear, negative relationship was found for the lutein and for Chl a molar ratio when was plotted versus the Chl a/chl b ratio (leaf No. 6 distance from the stem). 20 y = x R= 0.64 y = x R= Zeaxanthin (µmol m -2 ) 10 5 Leaf No. 2 Leaf No Distance (cm) Figure 11.46: Zeaxanthin concentrations as a function of distance from the main stem at 2 different leaves (leaf No. 2, circles; leaf No. 6, squares).corn canopy growing at Demmin site Intensive campaign sampling by University of Munich (LMU) The University of Munich has collected the following data sets in the within the framework of the AGRISAR 2006 campaign: 1) Installation of a permanent registering soil moisture station 2) Vegetation and soil moisture information for various sample points within the three intense campaigns 3) LAI measurements 4) Biomass measurements 5) Surface roughness measurements 6) photo documentation All data has been integrated into a GIS project (ArcGIS 9.0) and is ready to use. The following Figure shows the sampling points used for the various parameters during the campaign. 18/01/2008 Page 147 of 259

149 AGRISAR 2006 Samples University of Munich Legend \ \ \ \ \ A Biomass samples \ LAI 21 Roughness ( SM-station 1st field campaign 2nd field campaign 3rd campaign Field map A \ \ \ ( A \ A A \ \\ \\\ A \ A Kilometers A A A A A A Contact: Alexander Loew (a.loew@lmu.de) Figure 11.47: Overview of in situ data sampled by the University of Munich Vegetation biomass The vegetation biomass was determined for different fields during the intense campaigns. The fresh biomass was cut on 0.25 square meters and was weighted in the field. Afterwards it was dried in desiccators and the weighted again to obtain the dry matter content. The dataset contains the following numbers: - weight of sample (0.25 m²) - fresh matter (g/m²) - dry matter (g/m²) - water (g/m²) - percent dry matter - percent water - UTM coordinates (zone33, wgs84) 18/01/2008 Page 148 of 259

150 Vegetation biomass 100% 90% 80% 70% sample ID 60% 50% 40% water dry matter 30% 20% 10% 0% Point 250_1 250_2 250_3 250_4 250_5 250_6 391_1 391_2 391_3 391_4 391_5 391_6 250-V1 250-V2 250-V3 250-V4 250-V5 Figure 11.48: Vegetation biomass Fraction of dry matter and water content for different fields during the intense campaigns Field data A comprehensive collection of in situ data was gathered during the 3 intense campaigns. An overview about the samples has been already given in the report. For each sampling points shown in Figure 11.47, surface soil moisture was measured using mobile TDR probes (10 probes at each sampling point), vegetation height and phenology was determined and a photodocumentation was made. Leaf area index Leaf area index has been measured using a LICOR-2000 unit. Measurements were made on field #250 and #222 using 4 independent measurement at each sampling point. The measurements were then averaged and standard deviations for the LAI value were estimated. The following Table summarizes the measured LAI values: Field Mean Stdv 250@ @ @ Table 11.14: Measured LAI values 18/01/2008 Page 149 of 259

151 Roughness Two dimensional, high resolution surface roughness measurements were made using ground based stereoscopic image-pairs which were taken on several fields at the AGRISAR test-site, from The following Figure shows the roughness measurement unit in the field. The images were taken, using a digital camera and a fixed baseline between the two image pairs (Figure 11.50). A digital elevation model was then generated from the image pairs using ERDAS IMAGINE (Leica Photogrammetry Suite). The detailed documentation of the processing steps has been delivered separately together with the data. Figure 11.49: Roughness measurement setup 18/01/2008 Page 150 of 259

152 Figure 11.50: Stereoscopic image pairs (top) and generated elevation model (bottom) Soil moisture station The University of Munich deployed a soil moisture station within the AGRISAR 2006 campaign from 19/04/ /07/2006 in wheat field #250. The system is an IMKO-TDR Soil moisture probe which was installed in five depths (5cm, 9 cm, 15 cm, 47 cm). The soil is a homogeneous sandy Loam. The following Figure shows the installation of the soil moisture station in April 2006 and the measured data sets. The station operated successfully during the entire campaign. No data gaps occurred and the quality of the measured data is high. 18/01/2008 Page 151 of 259

153 Figure 11.51: Installation of soil moisture station AGRISAR soil moisture station 35 soil moisture [vol /04/06 23/04/06 03/05/06 13/05/06 23/05/06 02/06/06 12/06/06 22/06/06 02/07/06 12/07/06 date 4795A-5cm 4797A-9cm 5294A-15cm 5298A-25cm 5283A-47cm Figure 11.52: measured soil moisture during AGRISAR 2006 (wheat field #250) Intensive campaign sampling by University of Naples During the second intensive field campaign (July 4 th -7 th ) volumetric soil water content (θ) and soil temperature (T) were monitored in the superficial soil horizon, simultaneously to the aircrafts overpass. The measurements were acquired with a Frequency-Domain sensor (IMAG-DLO sensor model MCM 101, see picture below). Different parcels were sampled along the AHS and E-SAR airborne flight line. This type of probe, which prototype has been developed by IMAG- DLO, is made of a metallic wave-guide of 7 cm length, which allows an easy insertion in the soil and quick measurements. This feature allows for the acquisition of a set of measurements during an interval of 2-3 hours around the flight time; as such, the influence of the diurnal variation of surface soil water content is minimised. The map of soil water content resulting from the spatial interpolation of the grid of measurements on field no.222 is shown in figure below; the 18/01/2008 Page 152 of 259

154 average soil water content of the surface layer was about 0.100, corresponding to very dry conditions. Below is also shown the spatial variability of surface soil water content in the field 460 on day 6 th. Ground measurements have been carried out in three different plots in order to characterise the spatial variability of Leaf Area Index, by means of the portable canopy analyser Licor LAI The measurement protocol consisted of three consecutive series of 8 readings covering an Elementary Surface Unit (ESU) of approximately 20x20 m. The average value of LAI, resulting from the set of 24 readings, has been considered as representative for the considered ESU. In the figure below the set of LAI measurements is represented; the average development of the canopy in the three fields was similar, with a LAI value of approximately /01/2008 Page 153 of 259

155 3 2.5 LAI data Mean value AGRISAR ± ± ± field 222 field 102 field MAIZE bis (bis) SUGAR-BEET Intensive campaign sampling by ISSIA Surface Roughness measurements From 19 to 21 April 2006, 13 roughness profiles were acquired using the ESA laser profiler over three fields, namely field 102, 391 and 460. The fields 102 and 460 were bare, whereas the field 391 was a wheat field at an early stage. The lengths of the acquired profiles range between 20m and 5m. More precisely, there are 6 profiles 20m-long, 1 profile 15m-long, 2 profiles 10m-long and 4 profiles 5m-long. The profiles were acquired along two directions, i.e. perpendicular and parallel to the row directions. In the following, the acquired data are summarized: 3 profiles for field 102 on April 19; length: 20m, 15m, 10m; directions: 2 perp., 1 par. 6 profiles for field 391 on April 20; length: 3 of 20m & 3 of 5m; directions: 3 perp., 3 par. 4 profiles for field 460 on April 21; length: 2 of 20m, 1 of 10m, 1 of 5m; directions: 2 perp., 2 par. Unfortunately, during the acquisitions several malfunctions of the laser profiler were faced. In particular, there were two main problems: the battery of the laptop, which controls the laser system, had a too limited autonomy (i.e. just a couple of hours). In addition the on-site recharging system did not work; the radio link between the laptop and the laser had several failures which means that there are several gaps in the acquired roughness data. 18/01/2008 Page 154 of 259

156 The presence of data gaps has been ascertained during the pre-processing phase. To ensure a good data quality, only those parts of acquired profiles that were not affected by gaps have been processed and delivered. As a consequence, the 13 profiles here delivered have a reduced length with respect to the original ones. More precisely there are: 8 profiles of lengths ranging between 3 and 5m; 4 profiles of lengths ranging between 9 and 14m; 1 profile of 15m. The delivered roughness profiles are zero mean and have been de-trended using a linear fit. The obtained results in terms of standard deviation are listed in the following table: Field Profile nr Direction Length cm std cm per par per per par per par par per per par per par Table 11.15: Standard deviation of roughness profiles acquired over 3 experimental fields during the 1 st intensive campaign As an example, a picture showing a detail of the soil surface and the plots of the roughness profiles acquired over field 102, are shown. 18/01/2008 Page 155 of 259

157 Figure 11.53: Detail of soil surface and plots of roughness profiles acquired over field 102 TDR measurements TDR measurements were carried out from the ISSIA team during the 1 st and the 3 rd intensive campaign over the fields as shown in table 8.3 and 8.5. Measurements were acquired by using a TDR IMKO equipment with a probe of 10 cm, and for all sampling points, GPS coordinates were recorded by using a portable GPS. For each field, data acquired were checked and plotted. As an example, fig shows the plot of soil moisture (%) values measured in the field 102. In some cases, local trends of soil moisture were observed, and it was checked that occurred in areas of the field surface characterized by local slopes. One of these local slopes is present in the field 102, as it can be observed in the following picture. Moreover, soil moisture mean and standard deviation were computed for each field. Values obtained are reported in table From the table, it can be observed how soil moisture mean values estimated in April for the 6 fields are quite similar. The same occur for mean values estimated in July. Two exceptions are present, i.e. field 843 in April that is drier compared to the other fields, and field 823 in July, that 18/01/2008 Page 156 of 259

158 is wetter. To note that in July, for field 230, soil moisture mean value was estimated two times in two consecutive days. It can be observed that the field mean value is quite constant, i.e. around 5.8 %. vol. soil moisture (%) 35,00 33,00 31,00 29,00 27,00 25,00 23,00 21,00 19,00 17,00 15, sampling points Figure 11.54: Plot of soil moisture (%) values measured in the field 102, on 19th of April A picture of field 102, and part of the sampling point path (blue flags) superimposed to the field map, are also shown. Field Date Mv mean Mv std 102 Apr Apr Apr Apr Apr Apr Jul Jul Jul Jul /01/2008 Page 157 of 259

159 Table 11.16: Soil moisture (%) mean and standard deviation measured over experimental fields during the 1 st and 3 rd intensive campaign, LAI measurements LAI measurements were carried out from the ISSIA team during the 3 rd intensive campaign, over the fields as shown in table 11.17, i.e. fields of sugar beet 102 and 460, and fields of winter wheat 230 and 391. Measurements were acquired by using an accupar ceptometer, and for all sampling points, GPS coordinates were recorded. Analogously, from these data, LAI mean and standard deviation were computed for each field. Values obtained are reported in table From the table, it can be observed the differences of LAI mean values estimated for sugar beet and winter wheat fields, due to the different crop structure. Moreover, for the same crop, differences between fields were found. As an example, for winter wheat, LAI mean value measured in the field 230 is higher compared to the value measured in the field 391. Field Crop Date LAI mean LAI std 102 Sugar beet Jul Sugar beet Jul Winter wheat Jul Winter wheat Jul Table 11.17: LAI mean and standard deviation measured over experimental fields during the 3 rd intensive campaign, Intensive campaign sampling by Friedrich-Schiller University Jena Soil moisture The University of Jena has collected the following data sets at intense campaigns. Data quality information is given only for moisture measurements and ASD spectra acquisitions. 1. Soil moisture a. Volumetric (TDR-Probes) (intense campaign 2) b. Gravimetric measurements (Soil sampling) (intense campaign 2 and 3) 2. Laboratory based soil parameter analysis (intense campaign 2) 3. Biomass sampling (intense campaign 2) 4. Vegetation high measurements (intense campaign 2 and 3) 5. Leaf area photographs (intense campaign 2) Regarding soil moisture the details are as follows: 18/01/2008 Page 158 of 259

160 1. Volumetric (TDR-Probes, varied lengths) a cm (470 samples) b. 0-5 cm (620 samples) 2. Gravimetric measurements (Soil sampling, varied sampling depth) a. 0-5 cm (96 samples taken and analysed) b cm (18 samples taken and analysed) c. 0-5 cm (few hundred samples taken by UNI Jena and analysed by DLR) TDR measurements have been carried out with devices (Trime FM 2, Probes P2G and P2D) by IMKO GmbH. See the following Figure for details. The technical specifications including measurement accuracies can be found here: The probes have been calibrated with the handheld devices before measuring. Figure 11.55: Trime FM 2, Probes P2G and P2D At each measuring point between 3 and 5 TDR measurements have been accomplished. Each value is provided via the FTP server. Additionally the mean and the standard deviation are provided. Due to the low values for the standard deviation high quality measures can be assumed. Gravimetric measurements (soil sampling) were carried out by means of cylindrical sampling device (see Figure 11.56). After sampling the cylinders were covered with airproof closure caps. Those caps have been sealed eventually using robust plastic tape. Thus, errors due to soil water loss can be assumed being small. 18/01/2008 Page 159 of 259

161 Figure 11.56: Cylindrical sampling device for gravimetric measurements At all measuring points between 2 and 3 samples were taken. Each value is provided via the FTP server. Additionally the mean and the standard deviation are provided. Due to the low values for the standard deviation high quality measures can be assumed. Most of the samples have been analysed by DLR. Analysis refers to weight reduction estimation due to water loss after drying in the laboratory. The weight of each sampling device has been accounted for. Regarding the accuracy of the laboratory methods the reader is referred to DLR. ASD-Measurements FSU Jena measured transects on different fields with parallel measurements of soil parameters (vegetation height, vegetation density, and soil moisture). GPS points and white reference was measured every 20 m. Transects where continuously measured for spectra with sample averaging of 10 single measurements. Together about 5000 spectra (each averaged with 10 single measurements and saved as reflectance) were collected during the June and July campaigns on the fields 102, 222, 230, /01/2008 Page 160 of 259

162 Figure 11.57: Field spectra collected in the field campaign June Figure 11.58: Field spectra collected during the field campaign July The ASD FS3 is a new generation field spectroradiometer with high spectral resolution (compare with Table 11.18) 18/01/2008 Page 161 of 259

163 Table 11.18: ASD FieldSpec 3 specifications (left table column applies to the used instrument here FS3). The FS3 saves spectra with 3 nm spectral resolution in the VIS are and 10 nm resolution in the SWIR. The sampling interval is however even smaller (1.4 nm and 2 nm). Comparison ASD spectrometer data and AHS data will be conducted see Figure for first result (field 230) 18/01/2008 Page 162 of 259

164 Figure 11.59: Comparison of ASD spectrometer data and AHS data (top) on field 230 (bottom) Preliminary analysis Preliminary analysis focused on the soil moisture effect on backscatter and on spectral data. Figure gives an overview on the investigated fields regarding the soil moisture / backscatter relationship. Altogether 50 scatterplots have been created. Hardly any correlation between soil moisture and backscatter could be detected. Possible reasons are: 1) Vegetation attenuates soil signal, 2) The very low dynamic range of soil moisture: mostly between 2-8 Vol.% (very dry conditions). The best correlations are depicted in Figure Figure 11.60: Investigated fields regarding the soil moisture / backscatter relationship 18/01/2008 Page 163 of 259

165 Figure 11.61: Correlation between soil moisture and backscatter for specific fields and radar frequencies Of particular interest was field 102 (sugar beet) due to the dip perpendicular to the major field extension. The soil moisture pattern is related to its micro relief (see Figures ). Figure 11.62: Field 102 with transverse dip 18/01/2008 Page 164 of 259

166 Figure 11.63: Soil moisture pattern 5 th July 2006, basing on TDR measurements The following Figure presents two cross sections of field 102. Plotted is the backscatter of all E-SAR channels and the volumetric soil moisture. The effect of the dip regarding an increase of soil moisture is visible. However, only a slight effect on SAR backscatter can be detected. Nevertheless, a further question can arise: Does this topography induced soil moisture variation affect the vegetation and the spectra? Thus, the comparison of 3 spectra was conducted: top, slope, dip. See Figure for the results. From this diagram it becomes visible that the vegetation in the dip features the healthiest conditions comparing to the other ones. Thus, obviously water was the limiting factor during the last weeks or months. In a wet year this might be reversed. Figure 11.64: Two cross sections of field 102: Soil moisture and backscatter 18/01/2008 Page 165 of 259

167 Figure 11.65: Comparison of the spectra from top (1), slope(2) and dip(3) Land cover map The land cover map has been first generated during the 2 nd intensive campaign, and hereafter checked for the 3 rd intensive campaign, and hence potential misclassifications have been minimized. Furthermore, the land cover map has been used in connection with classification using the multitemporal SAR data set. In this case, no fields with strange signatures, also indicating potential errors in the reference map, have been found. In conclusion, the extensive land cover map can be considered to have a high quality Surface Energy Budget Bowen-ratio station A Bowen Ratio Energy Balance (BREB) station has been set up in the AgriSAR 2006 test site in Field 250 from April 20 through July 5, The directly measured variables include: Air temperature and humidity at 1 and 2.5 m height. Net radiation. Soil heat flux and temperature. 18/01/2008 Page 166 of 259

168 Air pressure. Wind speed and direction. Using these data, the latent and sensible heat fluxes were calculated. Overall, accurate turbulent fluxes were observed approximately 60% of the time. Unreliable estimates were obtained under the following conditions: Measured Bowen ratios between -0.7 and -1.3, in which case the Bowen ratio method should not be applied. Unreliable measurements of the wet bulb temperature, which are caused by a malfunctioning of the wicks around the wet bulb thermometers. Modelling and Observing the Water and Energy Budget The data from the station above have been compared to scintillometer-derived latent and sensible heat fluxes, measured by ITC. During daytime, the observations from both stations showed a good agreement, but at night-time the scintillometer-derived sensible heat flux was significantly lower than the BREB data. The cause for this discrepancy is the uncertainty in the type of stability functions used at night by the scintillometer. For this reason, the scintillometer data were assumed to be unreliable during night-time. Further, two hydrologic models, TOPLATS and PROMET, have been validated using the BREB data, and soil moisture profiles measured at the same location as the BREB data by LMU. It has been found that PROMET simulates the soil moisture profile slightly better, while TOPLATS simulates the energy balance slightly better. The results of this intercomparison are described in detail in a paper which is currently in press for the Journal of Hydrology [ 45 ] Scintillometer (LAS) Introduction To analyze the energy balance in the area, a Large Aperture Scintillometer (LAS) station including meteorological and radiative instruments has been installed in winter wheat field (Field number 250) and another LAS system has been set-up in corn field (Field number 222). The 5 meter high meteorological tower was equipped with a set of meteorological instruments to measure 4-component net radiation (i.e. incoming and outgoing solar radiation, incoming and outgoing thermal infrared radiation), as well as air temperature, relative humidity and wind speed at two different levels and wind direction was setup in the period 2 10 July 2006 (DOY ). Quality assessment 18/01/2008 Page 167 of 259

169 The meteorological observations (air temperature, relative humidity, wind speed and wind direction) can be considered of good quality during the entire period of installation. In addition, similar recordings were noticed from the nearby Goermin weather station. Evaluation of the radiative measurements (4-component net radiation and soil heat fluxes) showed a high quality for the net radiation measurements. For the soil heat fluxes, however, some notes have to be made. Obligations in another campaign (EAGLE-2006) invoked relatively late installation of the station, which is experienced as problematic for the soil heat fluxes. These require some time (order of weeks, depending on soil characteristics) for relaxation of the soil, which is disturbed during the installation. In addition the canopy cover is influenced by the installation as well, at times yielding direct exposure of the soil to the incoming radiation. The combination of these effects led to higher amplitudes and a shift in time with respect to nearby recorded soil heat fluxes (BREB station in field 250). With respect the quality of the LAS turbulence observations made it should be noted that for days 189 and 190 there were significant rainfall events. This distorts the signal transmitted by the scintillometer; reason for which the recordings made during these two days should not be used in any analysis. In addition, the scintillometer installed in field 222 experienced battery problems after the rain events on DOY 189 and 190; these measurements are also not to be used. Furthermore, with respect to the scintillation measurements, it should be noted that the raw recorded scintillation data are post-processed into values for sensible heat flux, which requires additional data input that was not directly measured. It should be noted that this is usually the case in scintillation measurements (in fact friction velocity measurements are needed, which require eddy correlation measurements). This means that in the current case: 1. Certain assumption regarding the aerodynamic roughness of the observed fields have been made; 2. Required additional meteorological data for field 222 is taken from measurements in field 250 (which are not necessarily representative); 3. The post-processing invokes using Monin-Obukhov stability functions. Which functions to apply during night-time (stable atmospheric conditions) is still under debate. We have applied several of them with variable results, reason for us to neglect the night-time observations and assume the fluxes to be equal to zero (a procedure more often followed, see e.g. [ 43 ] and [ 36 ]. All in all however, for the days without rainfall, the observations can be considered as fairly good, which is confirmed by similar behavior of the nearby installed BREB station in field /01/2008 Page 168 of 259

170 Goniometer Introduction Directional measurements were performed using a goniometric setup. A goniometric setup is able to perform hemispherical measurements of the same target. An ASD and several other spectrometers were used to perform radiative measurements in the spectral range of nm. An Everest 3000 radiometer and an Irisys 1011 thermal imager were used to perform radiative measurements in the spectral range of 8 12 μm. The most stable spectrum acquired over the several VNIR measurements per single angle was used. The BRDF is then produced from this spectrum. As for thermal radiative measurements no unique BRDF can be created, the thermal directional signature is created from the measurements. The post-processing of the data dealt with the directional behaviour of the acquired BRDFs and thermal signatures. A normalized reflectance ratio is introduced to emphasize the directional behaviour of the measured reflectances. Rθ Rnadir Ratio = R + R θ nadir The standard deviation of this ratio per angle then provides a single parameter describing the directional behaviour of the BRDFs. For the thermal images, the ratio of the standard deviation of the temperatures and the mean temperatures are used to acquire a single parameter for the directional behaviour. In the calculation of this thermal directional behaviour of the radiometer the measured values are used without processing. For the determination of the directional behaviour using the thermal imager the average temperature per image are used. The standard deviation of the temperature per image is used to derive a separation parameter. This separation parameter is then used as an indicator whether further processing for retrieving canopy component temperatures would be successful. Final output contains the BRDFs and the thermal signatures. In Table the maximum standard deviation of the reflectance ratio is shown and in Table the thermal directional signatures and the separation parameter of the measured crops are listed. Crop max(std) max(std) max(std) VIS NIR MWIR Mature Corn Wheat Sugarbeet Table 11.19: Soil and vegetation parameter during the 3 rd intensive campaign 18/01/2008 Page 169 of 259

171 Crop Directional signatures Separation parameter Everest Irisys Irisys Mature Corn Wheat Sugarbeet Barley Table 11.20: Thermal directional signatures and component temperature separation parameter Quality assessment With respect to the solar range, or VNIR measurements, the observations over the wheat display the largest directional behaviour in the VIS region; mature corn displays the largest directional behavior in the NIR region, and wheat and sugar beet display the largest directional behavior in the MWIR region. Comparison between the directional signatures of the Everest measurements and those of the Irisys show similar behavior. With both instruments mature corn and sugar beet show the largest directional behavior. Because the standard deviation of the temperatures per image is highest over sugar beet and mature corn, the directionality of these crops types can be used to separate the temperature of the different canopy components. The overall quality of the hyperspectral VNIR measurements is good; 1. The hyperspectral VNIR measurements were acquired with the new ASD spectrometer of the group of the Jena University and a new white reflectance panel. Therefore sensors errors can safely be assumed very low. 2. If cloud conditions change rapidly the acquired reflectance per angle may be changing. However weather conditions during the acquisitions were very stable. In addition no clouds were present at the hemisphere. 3. The spectra are not splice corrected. The ASD has three sensors that each measure a part of the spectrum. At the intersection of these sensors a splice occurs for which several correction schemes exist. Since each of them produced different output results, we have not performed this correction. The quality of the thermal measurements can also be characterized as good; 1. Thermal signatures may change very rapidly. As the weather profile was very stable and each of the transects was carried out within a time span of 5 minutes, the temperature drift per measurement is minimal. In addition, before and after every measurement the temperature of a single target was measured to investigate the temperature drift, which was less than 1 K on average. 18/01/2008 Page 170 of 259

172 2. The thermal directional behaviour agrees with measurements performed during previous field campaigns (SEN2FLEX-05 and EAGLE-2006). As well, the measured temperatures are in limits of expectations according to the air temperatures. 3. The weather conditions were clear which should produce the maximum thermal signature. 4. The Everest thermal directional signature showed the same behaviour as the Irisys thermal directional signature. 18/01/2008 Page 171 of 259

173 12 PRELIMINARY DATA ANAYLSIS 12.1 Synthetic Aperture Radar Derivation of soil surface roughness dynamics using L-Band PolSAR data by the University of Kiel The roughness of natural surfaces, defined as the height deviations from a plain reference in the scale of mm [ 50 ], plays an important role in numerous physical processes. Several investigations showed the impact of soil micro relief on processes such as the shortwave radiation balance, wind and water induced soil erosion or near-surface soil moisture as well as their description in respective models [ 22 ]. The nature of rough surfaces can be described statistically by means of various roughness indices, like the RMS height, the surface correlation length or the Tortuosity index [ 59 ]. Temporal roughness variability is obvious due to wind, agricultural practice or soil sealing and elutriation by precipitation or irrigation. However, little knowledge is available on these effects on from the field or to the small watershed scale [ 49 ], [ 59 ], [ 35 ]. Thus, roughness is often assumed constant in respective modelling efforts, introducing strong simplification and considerable data uncertainty [ 8 ], [ 10 ], [ 5 ]. To bridge this scientific gap, the potential to derive soil surface roughness information on field scale from multi temporal airborne PolSAR data is investigated and evaluated In field roughness measurements Roughness measurements were performed using photogrammetric imaging techniques. For roughness survey a portable tripod was developed were a calibrated Rollei d7 metric digital camera and 12 high accurate (3/10 mm) ground control points (GCP) were fitted. Images were taken from approx. 118 cm height above ground and the height-base ratio was approx The horizontal coverage of the stereo model 18/01/2008 Page 172 of 259

174 Figure 12.1: Self-developed portable tripod for Ground-Truth roughness estimation. is 70x70 cm². The goal to develop the tripod was to reduce processing time and reduce the sampling duration as much as possible. To get information on soil surface roughness of vegetated sample points the vegetation was carefully cut off direct above the ground and removed from the scene. The three dimensional surface reconstruction was done by using Leica Photogrammetry Suite LPS V 9.0 software. Exterior orientation of the two images was established using bundle block adjustment techniques. Therefore, additionally to the 12 known GCPs, tie-points were derived and their three dimensional coordinates were calculated by bundle block adjustment. Best results in bundle block adjustment were achieved by using a additional 12 parameter model (Ebner Model). The achieved accuracy from the exterior orientation is in z=0.8 mm and in xy=0.37 mm referred to the known GCPs. For derivation of the DSMs different strategies had been developed. Thus, LPS works in epipolar lines the strategies vary only in the x-direction depending on the degree of soil sealing. Minimum correlation coefficient was set to Roughness Indices To quantify the soil surface roughness two roughness indices had been calculated from the derived DSMs, the rms-height s and the Tortuosity Index T B after [ 34 ]. The rms-height is defined as the standard deviation of the height values Z: 18/01/2008 Page 173 of 259

175 s = n i= 1 ( Z i 1 n Z ) 2 The Tortuosity Index T B for three dimensional surfaces was first introduced by [ 34 ]. It is defined as the ratio of the 3D-surface area to the 2D-projected area: 3D T B = 2 D For characterization the impact of rainfall to changes in micro relief topography T B is more sensitive as investigated by [ 59 ] Radar data For data analysis geocoded SLC L-Band Data was chosen to derive the roughness information. As shown by [ 61 ] it is feasible to use geocoded SLC L-Band data to perform decomposition algorithms. The Radardata was speckle filtered by applying a 7x7 window enhanced LEE-Filter. To obtain an improved understanding of the involved scattering mechanisms, a Cloude decomposition of the backscatter signal was performed using a 5x5 box-car filter. It distinguishes the dominant scattering process (surface, volume or double-bounce) by means of the alpha angle and allows to determine the proportional fraction of the other scattering components in terms of backscatter entropy and anisotropy. For roughness determination, three well established roughness estimators were calculated: Anisotropy As shown by [ 32 ] and [ 12 ], the Anisotropy defined as: λ2 λ3 A = λ + λ 2 with λ x = second and third eigenvalues, is sensitive to soil surface roughness on bare soil fields. Reference [ 12 ] and [ 13 ] introduced two inverting approaches depending on roughness states to estimate soil surface roughness. Note that the autocorrelation length is unconsidered for estimation. Circular polarization Coherence The circular coherence is defined as: 3 γ RRLL = S S RR 2 RR S * LL S LL 2 with S RR = right-right rotation, S LL = left-left rotation of the electric field vector about the line of sight. Reference [ 38 ] verified first a significant sensitivity of the circular coherence due to 18/01/2008 Page 174 of 259

176 surface roughness while the impact of the dielectric constant is reduced. As roughness increases the circular coherence decreases. The circular coherence was calculated by applying a 5x5 boxcar filter on the SLC data. Real part of the circular polarization coherence The real part of the circular coherence was first introduced by [ 55 ] and is defined as: Re [ RRLL] = S S HH HH S S VV VV S S HV HV 2 2 It is a further development of the circular coherence with the advantage of not using the imaginary part of the circular coherence which is sensitive to asymmetric scattering contributions such as vegetation. Furthermore it is very insensitive to parameters such as the dielectric constant. The real part of the circular coherence was calculated by applying a 5x5 window boxcar filter Results Correlation coefficients between the potential roughness estimators specified in and the ground truth roughness values were calculated for the total investigation area, for areas with dominant surface scattering and for bare soil fields separately. As indicated in Table 12.1, the correlation showed only good results for Re[RRLL] and s. Note, that these results were only achievable by masking out those values with s<1 cm (ks < 0.27 cm). It can be seen, that values of s <1 cm (low roughness) are rather dominated by noise. Parameter r R² s Tb s Tb Anisotropie surface scatt Anisotropie bare soil γ RRLL total γ RRLL surface scatter γ RRLL bare soil Re [RRLL] total Re [RRLL] surface scatter Re [RRLL] bare soil Table 12.1: Correlation coefficients between the radar parameters and the ground truth roughness information for each roughness Index. 18/01/2008 Page 175 of 259

177 3 2.5 y = 0.72x R 2 = s[cm] Re[rrll] Figure 12.2: Correlation between s [cm] and Re [RRLL] (Values s<1 cm are masked out) It is obvious that s and the radar parameters are more correlated outperforming the Turtosity index. The correlation between s and Re [RRLL] is shown in Figure Based on the correlation between s and Re [RRLL] the surface roughness for the whole investigation area was estimated. Figure 12.3 shows as an example the estimated soil surface roughness based on the correlation. Settlements, forest and streets are masked out in light gray tones. As indicated different roughness states between fields and inside some fields are obvious to distinguish. Areas with the same roughness states appear also in the same gray tones. Figure. 12.3: Estimated roughness map for the based on the correlation between s and Re [RRLL] 18/01/2008 Page 176 of 259

178 Multi-temporal roughness analysis For each campaign day, roughness maps were calculated based on the correlation shown in Figure The developing of the roughness states for the whole agri-penological cycle is shown in Figure 12.4 and Figure It is obvious in both figures that the roughness state is changing over time. Under winter vegetation (Figure 12.6) such as Winterrape (101) the roughness decreases slightly. For Winterwheat and Winterbarley a stronger decrease of roughness can be observed until the 17 th of May. Than Winterwheat stays quite low (s= cm) while the roughness state of the Winterbarley field increases slightly again. This causes the suspicion that the roughness estimation for the Winterbarley field is influenced by vegetation, indeed statistical analyses showed no impact. In general it is to note, that the soil surface roughness under winter resistent vegetation is overestimated by means of 0.8 cm without any impact from vegetation The developing of roughness states for soil surface under summer vegetation (102, 460 (SB), 222 (M)) shown in Figure 12.5 is similar to the winter vegetation (Figure 12.4). For the Maize field (222) a decrease can be observed. Indeed the roughness 2 s[cm] /12/2006 4/19/2006 4/26/2006 5/3/2006 5/10/2006 5/17/2006 5/24/2006 5/31/2006 6/7/2006 6/14/2006 6/21/2006 6/28/2006 7/5/2006 7/12/2006 7/19/2006 7/26/2006 8/2/ (WR) 1012 (WR) 1013 (WR) 2501 (WW) 2502 (WW) 2503 (WW) 4401 (WB) 4402 (WB) 4403 (WB) Date Figure 12.4: Roughness developing for the sample points 101 (WR), 250 (WW) and 440 (WB). 18/01/2008 Page 177 of 259

179 2.2 s[cm] (SB) 1022 (SB) 1023 (SB) 2221 (M) 2222 (M) 2223 (M) 4601 (SB) 4602 (SB) 4603 (SB) 0.4 4/12/2006 4/19/2006 4/26/2006 5/3/2006 5/10/2006 5/17/2006 5/24/2006 5/31/2006 6/7/2006 6/14/2006 6/21/2006 6/28/2006 7/5/2006 7/12/2006 7/19/2006 7/26/2006 8/2/2006 date Figure 12.5: Roughness developing for the sample points 102 (SB), 222 (M) and 460 (SB) S modelled S measured Figure 12.6: Modelled vs. measured roughness values measured values for s in the field are 0.2 cm higher than the estimated roughness values. Both of the Sugarbeet fields show first a strong decrease in soil surface roughness until the 7 th of June and then show a continuous increase very similar to the groth of the Sugarbeet plants. However, a multiple regression between the roughness values, vegetation parameters and Re [RRLL] showed only a strong relationship between the values on field 460. On field 102 there was no relationship measurable. For the Sugarbeet fields an overestimation of roughness can be observed in average of 0.21 cm for the field 102 and 0.26 cm for 460. Figure 12.6 shows the deviations between the measured and estimated roughness values for s. As detailed explained above, it can be observed that high roughness values are underestimated and small values for s tend to result in overestimations. 18/01/2008 Page 178 of 259

180 Conclusions As shown in this investigation, only the real part of the circular polarization coherence is sufficient correlated with the rms heigth measured under natural conditions over a wide range of soil surface roughness states, outperforming any other potential correlation or roughness index. These results verify the investigations by [ 55 ] and [ 60 ]. However, this investigation shows, that Re [RRLL] is not only sensitive to soil surface roughness. Even tough an influence on Re [RRLL] from vegetation could not be quantified with the obtained standard vegetation parameters it is obvious that the developing of plants affect the estimation of s through Re [RRLL] Investigation of Soil Moisture by ISSIA Introduction Spatial and temporal distribution of soil moisture content plays a crucial role in hydrology, agriculture and meteorology. The appropriate sampling for the soil moisture fields strongly depends on the characteristic spatial/temporal scale of the specific application, whereas the science requirement for the accuracy of the superficial volumetric soil moisture content is usually set to 4%-5%, allowing the identification of 4-5 moisture classes. Due to the high sensitivity of radar backscatter to the soil water content, SAR systems are well suited to retrieve superficial soil moisture content at the local and regional scales (i.e. from the field to the basin scale). The method most often adopted to retrieve soil moisture from SAR data is the inversion of direct models, either semi-empirical or theoretical, relating SAR observations to the soil moisture. Despite SAR systems have the potential to meet the science requirements for monitoring the superficial soil moisture at high spatial resolution, to date their use has been generally limited and no operational algorithm is available. One of the reasons that have mainly hindered the use of past SAR space missions data in operational applications is the long revisit time (e.g. five weeks), which hampers services requiring information with higher frequency. Moreover, an important part of the limitations to monitor superficial soil moisture is due to the disturbing effect of surface roughness and of vegetation layer that both modulate the radar sensitivity to the soil moisture content thus rendering intricate the retrieval problem. For relatively simple SAR configurations (e.g. single frequency, one/two polarizations, one incidence angle), there generally exist many combinations of surface parameters mapping the same SAR observable, then the retrieved optimal solution (i.e. most probable or minimum rms error) may be characterized by poor accuracy [ 53 ]. This problem may be tackled by introducing a priori information about the surface parameters [ 39 ] and using multi-temporal SAR data. In the framework of this study, a preliminary analysis on the use of multi-temporal C-band SAR data to retrieve soil moisture content was carried out. More precisely, the analysis focused on the development and assessment of a retrieval algorithm for the superficial soil moisture content underlying winter wheat using multi-temporal C-band SAR data. The algorithm basically inverts a semi-empirical model describing wheat backscatter and it has been developed with a view to the 18/01/2008 Page 179 of 259

181 possible use of data acquired by the ASAR system and by the future Sentinel-1 system. In the following, a short summary of the obtained results will be reported. More details can be found in [ 40 ] and [ 41 ] The retrieval algorithm The proposed algorithm transforms a temporal series of SAR images, acquired over wheat fields at C-band, VV polarization and low incidence angles, into maps of volumetric soil moisture (m v ) and wheat water content (F) values. The adopted approach consists of inverting the semiempirical model described in [ 40 ] by using a constrained optimization technique that assimilates a priori information on surface parameters [ 39 ]. The algorithm is expected to be able to estimate soil moisture content of bare and wheat vegetated fields during the entire phenological cycle and its performances have been assessed on the AGRISAR data set. The use of multi-temporal data is beneficial for the accuracy of the retrieved surface parameters under the condition that the surface roughness remains almost constant during the time (T) of the N acquisitions. For instance, given a temporal series of N images, the number of surface parameters to be estimated is 2N+2 (N soil moisture, N wheat water content values and 2 surface roughness parameters, namely s and the correlation length l). For N equal to 1, there is the lowest ratio (i.e. 1/4) between independent measurements and parameters to be estimated (highly inaccurate retrieval expected). Whereas for N large the ratio tends to 1/2 (accurate retrieval expected). In practice, there is a trade off between the maximum possible number N of multi-temporal images and the time span T of the total acquisitions. A number of multi-temporal images between 3 and 5 can be foreseen. Under these circumstances, it is worth noticing that, for the same time span T, the availability of multi-temporal and multi-polarization SAR data (e.g. HH & VV) is expected to significantly improve the algorithm performances. The retrieval algorithm has been assessed on data acquired on the field 221 which is the only wheat field imaged at relatively low incidence angles (i.e. 33 ) on the Demmin site. Since no ground data were systematically acquired on field 221, the accuracy of soil moisture content retrieved on field 221 has been assessed by using results derived in a previous study [ 41 ]. In that study, L-band SAR data were employed to obtain soil moisture maps over all the wheat fields in the Demmin study area. Results were validated by using extensive multi-temporal TDR measurements carried out over several fields in the area (e.g. fields 230, 250, 391, etc) but field 221. The developed algorithm has been run considering two different cases. In the first case, the 12 images acquired by the ESAR system from April 19th to August 2nd, were processed in four independent runs. For each run, 3 ESAR images, acquired at subsequent dates, were employed to retrieve surface parameters over field 221. This means to consider N equal to 3 and a time span T of approximately 3 weeks. In the second case, the impact of increasing the time span T to approximately 5 weeks, while preserving N=3, was investigated. In both cases, a priori information concerning soil moisture content and wheat vegetation water content were obtained 18/01/2008 Page 180 of 259

182 from weekly in situ measurements on field 230, whereas a constant value of 1.0 cm was adopted for the s parameter. No a priori information on correlation length l is used. This is because: 1) it is extremely difficult to provide reliable values of l unless accurate in situ measurements had been carried out; 2) in the inversion procedure, the use of l as a free parameter may allow to better match the observed SAR data with the semi-empirical model. In Fig. 12.7, the scatter plots of the C-band SAR-derived volumetric moisture values versus the L-band SAR-derived values for the 12 acquisitions dates are reported. In Fig a and Fig b, mv values obtained by employing C-band SAR data acquired with a time span of approximately 3 and 5 weeks, respectively, are shown. 9 dates; mean m v values estimated over field 221 (a) (b) Fig Scatter plot of C-band SAR-retrieved volumetric soil moisture values versus L-band SAR-retrieved values [ 41 ]. Soil moisture values refer to field 221 and were estimated by using series of 3 ESAR images acquired within a time span of T 3 weeks (a) and T 5 weeks (b), respectively. The agreement between C- and L-band derived values is quite good in both cases with a rms error equal to approximately 4%. However, an unusually high value for the bias has been found (i.e. bias 6%). This is probably due to the fact that volumetric moisture values retrieved from L- band SAR data are not expected to be exactly equal to those retrieved from C-band SAR data due to the different penetration depth into soil. To better investigate the potential of the algorithm performances, a simulation experiment has been carried out in the next section. 18/01/2008 Page 181 of 259

183 The simulated experiment Table 12.2 reports the parameters used to build a simulated data set of ground data. Three conditions of soil and wheat vegetation surfaces representing three different dates were simulated. In particular it was modelled an increase of vegetation water content and an abrupt increase of soil dielectric constant (ε r ) on the second date (i.e. precipitation event) followed by a soil drying on the third date. The simulated a priori information to be used in the retrieval algorithm was obtained by perturbing the mean parameters. The perturbation was obtained by adding a Gaussian noise with zero mean and a std equal to 20% of total variability range of the selected parameter. The simulated set of σ 0 at VV polarization and 23 incidence was obtained by means of the semi-empirical direct model that was perturbed by superposing a Gaussian noise with zero mean and 1.0 db std. Then the algorithm was run, and Fig shows the scatter plot obtained for the volumetric soil moisture content. The error is approximately 5% and the bias is negligible (i.e. 0.39%). For the vegetation water content an error of 0.8 Kg/m2 has been found. Mean Parameters 1st date 2nd date 3rd date <s> (cm) <l> (cm) <Re(ε r )> <F> (Kg/m2) Table 12.2 surface parameters used in the simulated data set 18/01/2008 Page 182 of 259

184 Fig Simulated data. Scatter plot of C-band SAR-retrieved volumetric soil moisture values versus expected ones. Soil moisture values were estimated by using series of 3 simulated SAR images Conclusions A methodology to retrieve superficial soil moisture content underlying wheat fields, based on multi-temporal C-band SAR data at VV polarization and a priori information, has been presented. The algorithm applies to SAR data acquired at low incidence angles and uses a constrained minimization technique to invert a semi-empirical direct scattering model. The retrieval algorithm has been applied on ESAR data acquired during the Agrisar 2006 campaign on field 221. As a priori information, the in situ measurements collected over field 230 have been employed. Results indicate that when temporal series of 3 ESAR images, acquired within a time span T ranging between 3 and 5 weeks, are used, it is feasible to retrieve soil moisture content with an accuracy better than 5%. A wider assessment of the algorithm conducted by means of a simulation experiment demonstrated that to obtain accuracies of approximately 5% in m v and of 0.8 Kg/m 2 in F, a priori information about surface parameters, with an accuracy within approximately 20% of the total variability range of surface parameters, is required. The algorithm has been developed and assessed for single-polarized SAR data, nevertheless the use of multi-temporal and multipolarimetric SAR data (e.g. HH & VV) would significantly improve its robustness and accuracy Decomposition of different scattering mechanisms for soil moisture estimation under the vegetation: Preliminary Analysis on the polarimetric AGRISAR data by DLR-HR Methodology The estimation of soil moisture by means of SAR has been intensively investigated in the last two decades. Most of the research work was focused on bare surfaces, where satisfactorily 18/01/2008 Page 183 of 259

185 estimation results have been achieved by different theoretical as well as empirical or semiempirical approaches. However, bare fields are only a special case in agriculture. Agricultural fields are over large periods of their growing cycle covered with vegetation of different crop types. Thus the scattering scenario becomes more complex as the wave first interacts with the vegetation layer and then propagates through to interact with the underlying surface. Therefore, vegetation and surface effects are superimposed in the scattering signature. In order to decompose their contribution, model-based or experimental decompositions can be used to separate them within a resolution cell. As a basis prerequisite the balance between the observables and the parameters must be given. This condition is satisfied when enough observables are given through multi-parametric SAR acquisitions. In this paper the extension to multi-parametric SAR is given through the polarimetric and frequency diversity. The well known 3-component model based decomposition of Freeman-Durden [ 23 ] was used for the separation of scattering mechanisms. The measure that is used to quantify the individual scattering contribution is the ground-to-volume ratio (m) [ 14]. m is an expression that describes the relation between the surface and the volume scattering component and hence, provides an quantitative estimate of the surface component. The innovation of this is that for the first time m is estimated from SAR polarimetry and the variation is shown over a whole vegetation period. One problem of the 3-component Freeman-Durden decomposition is the simplification of the vegetation which can lead to erroneous assumptions regarding the vegetation. Therefore, the recently published model-based 2-component Freeman decomposition [ 24 ] is used that allows varying the vegetation shape (ρ) and is more flexible in the interpretation of the vegetation structure. This shape parameter is used for a better characterisation of the vegetation layer. The results of both the 2- and 3-component decomposition of the AGRISAR scene will be presented and used for characterisation of the crop vegetation status Freeman-Durden 3-Component Decomposition Model description The Freeman-Durden model was already developed in 1998 [ 23 ] for vegetation cover modelling using simple electromagentic models describing different scattering contributions to the total vegetation scattering. The model allows decomposing the backscatter signal into three scattering mechanisms: volume scattering, first-order Bragg surface scattering and dihedral reflection. The coherency matrix [T tot ] of the total backscatter is related to the scattering matrix [S] via the scattering vector k: S = S S ρ 1 HH HV S k = [ S ] T HH + SVV, SHH SVV, 2SHV VH SVV 2 18/01/2008 Page 184 of 259

186 [ T ] tot = k k + 1 = 2 S + S 2 VV * ( S + S )( S S ) 2 ( S + S ) * HV * 2 * ( SHH + SVV ) ( SHH SVV ) SHH SVV 2 ( SHH SVV ) SHV 2 ( + ) ( ) * * 2 S S S 2 S S S 4 S HH HH VV HV HH HH VV VV HH HV VV HH HV VV S [T tot ] is modelled as the sum of the three coherency matrices of the three different scattering mechanisms: [ T ] = T 21 T 22 0 = [ T ] + [ T ] + [ T ] tot model T11 T T 33 s d v The first contribution originates from rough surfaces. The backscatter from such a surface is described by Bragg scattering (also known as the small perturbation model (SPM) [ 62]). The two parameters f s and β form the coherency matrix of the surface component: 1 [ T s ] = f s β 0 β β 0 * h m 2 f s = Rh + R and v 2 β = Rh Rs R + R h s f s and β depend on the Bragg coefficients and m s, which accounts for roughness [ 62 ]. The Bragg coefficients are defined for horizontal and vertical polarisation and depend only on the incidence angle Θ and the dielectric constant of the soil ε s : 2 cosθ ε sin θ 2 2 s R = and ( εs 1) ( sin θ εs ( 1+ sin θ ) h R 2 v = 2 cosθ + εs sin θ 2 cos sin εs θ + εs θ The power of the surface component P s is defined as the sum of the diagonal elements of the coherency matrix and is hence: P 1 + β 2 s = f s The second contribution describes the signal that is reflected back to the source from two orthogonal surfaces (dihedral) such as from the ground and a vertical trunk or wall. Its coherency matrix representation has the same form as the surface component but different definitions of the parameters: [ T ] = d f d 2 α α 0 α * iϕ RghRth RgvRtve f d = + and 2 R α = R gh gh R R th th R + R gv gv R R iϕ tve iϕ tve 18/01/2008 Page 185 of 259

187 The two reflections are described with the Fresnel model. The parameters f d and α are hence related to the Fresnel coefficients of the ground R g and of the vertical structure (e.g. trunk) R t for horizontal and vertical polarisation respectively (subscripts h resp. v). φ denotes the phase differences between S hh and S vv incorporated by propagation through vegetation or scattering. The Fresnel coefficients are defined for horizontal and vertical polarisation and depend only on the incidence angle Θ and the dielectric constant of the ground ε g resp. trunk ε t : cosθ ε g sin θ R = and gh 2 cosθ + ε sin θ g 2 R th cosθ = cosθ + 2 ε sin θ t 2 ε sin θ t 2 εg cosθ εg sin θ R = and gv 2 εg cosθ + ε g sin θ R tv εt cosθ = ε cosθ + t 2 ε sin θ t 2 ε sin θ t The power of the dihedral component P d is: P 1 + α 2 d = f d The volume is modelled by a cloud of randomly oriented dipoles. This should approximately describe the radar return from vegetation with dipole-like structures in the order of the wavelength such as prolate leafs, needles and branches. The coherency matrix representation of the volume component requires only one parameter f v : [ T ] v 2 fv = The power of the volume component P v is equal to f v. The total power P tot from all three components is: P tot fs fd 2 = 1 + β α + f v The calculation of the parameters is done by solving the system of equations and is described in more detail in [ 23 ]. The key point here is that there are only four observables (T11, T22, T33 and T12=T21 * ) but five parameters (f s, f d, f v, α, β). To account for this an assumption is made for one of the parameters. This is done by first determining whether surface or dihedral scattering is dominant. α resp. β of the non-dominant component is then set to a predefined value, allowing the remaining parameters to be calculated. 18/01/2008 Page 186 of 259

188 Ground/Volume Ratios m b, m d, m min, m max After the model based decomposition three scattering contributions can be separated that are used to define the ground-to-volume ratio m. The ground-to-volume ratio (GVR) is a quantitative measure and expresses the relation of the ground to the volume contribution. The ground scattering contribution has two main components, the surface and dihedral scattering. Therefore, both are defined by the ratio of the surface (Ps) and dihedral (Pd) power to the volume (Pv) power of the Freeman-Durden 3-component decomposition and are expressed as the following GVRs: P s m s = and Pv P m d = P d v m s and m d are bounded between two values m min and m max that can be obtained by a eigendecomposition of their common coherency matrix T g =T s +T d. Since this matrix is of rank 2 and hermitian, it has two positive eigenvalues λ 1 and λ 2 (λ 1 >λ 2 ) that are different from zero. m min and m max are then given by: λ = and 2 mmin P v λ 1 mmax = P v The difference between m s and m d gives the full polarimetric range for the possible GVR. The range decreases with increasing volume contribution. In Figure 12.9 this is illustrated by assuming a power transfer from surface to volume at a constant total and dihedral power. As the volume component increases, both GVRs m s and m d and their boundaries m min and m max decrease reciprocally. Simultaneously, the difference between m min and m max shrinks from 3 to <1 as P v increases from 1 to 6. In this Figure the maximum of GVR set to 10 since the GVR becomes infinite as P v approaches to zero Ps Pv ms md mmin mmax 6 GVR 4 2 Pd Pv Figure 12.9: Reduction of the GVR polarimetric range with increasing volume power (P tot =10, P d =4, P s =P tot -P d -P v, f d =3.96, α=0.1, β=0.2) 18/01/2008 Page 187 of 259

189 Decomposition results The 3-component decomposition described above has been applied on both C- and L-band data of the west-east track of the AGRISAR flight stripe for three different dates lying in the beginning, the middle and at the end of the vegetation period of From the parameters from the decomposition the powers of the three components have been calculated. The results are shown in Figure and using an RGB presentation where the dihedral power corresponds to red, volume to green and surface to blue. For comparison, a Pauli image of the scene has been added to the left, where two fields are highlighted. For the two exemplarily fields, corn and wheat, a detail investigation and discussion is presented. The decomposed images show a quite strong distinction between the three scattering mechanisms and are mostly uniform over single fields. Qualitatively, the resulting image is similar to the Pauli decomposed image, which is due to the fact that the Pauli components correspond in a first approximation to the same three scattering mechanisms. However, for the Pauli decomposition the separation into the three components is less distinct leading to a much lower contrast than obtained by the 3-component decomposition. The main reason for it is that the SAR data are fitted to the model. In the three Freeman decomposed images changes due to vegetation growth in time can be very well observed. The corn field (222) for instance appears in both C- and L-band blue in May and June which indicates surface scattering, i.e. low or no vegetation. In August, however, (see Figure 12.10) clearly vegetation (green) and dihedral (red) dominates which is expected for corn plants. Indeed, field measurements confirm that the field was still bare at the beginning of May and the growth of corn plants happened mainly in July. dihedral flight direction volume surface near range (25 ) far range (55 ) 06/06/ /05/ /06/ /08/2006 Figure 12.10: RGB image of the Pauli (left) resp. Freeman 3-component decomposition at different dates at C-band 18/01/2008 Page 188 of 259

190 dihedral flight direction volume surface near range (25 ) far range (55 ) 06/06/ /05/ /06/ /08/2006 Figure 12.11: RGB image of the Pauli (left) resp. Freeman 3-component decomposition at different dates at L-band For the wheat field (250) the opposite behaviour can be seen and appears differently in C- and L-band: In May the vegetation (C-band) resp. dihedral component (L-band) dominates changing to dihedral (C-band) resp. surface (L-band) in June and finally in August there is almost no component present. This fits well to the ground observations during this period (see Figure 12.11): In May wheat has already grown to >20 cm, but got considerably dryer (i.e. more transparent for radar) in June. In C-band, the dense vegetation in May only allowed volume scattering, but as the vegetation became dryer in June the signal could penetrate deeper into the vegetation resulting in a predominantly dihedral scattering. In L-band, due to the longer wavelength the signal could even penetrate the dense vegetation in May causing dihedral scattering. The dry vegetation in June became almost transparent to L-band that finally surface scattering from the ground dominated. On the 2nd August, when the last image was taken, field 250 had already been harvested (evtl. ground data) and produced only very little backscatter. Ground-to-Volume Ratio (GVR) For the quantification of how the ground component was changing with vegetation growth the two GVRs as formulated above are derived and presented in Figure and The surface and dihedral GVRs have been derived for the same scene and dates. In August in both C- and L-band the surface GVR m s is much lower for the field 222, when the corn was fully grown. In C-band, m s is significantly higher in field 250 than in the months before due to the harvesting of the corn. In L-band, however, m s is already increased in June due to the dry condition of the crop. The dihedral GVR m d of the wheat field shows in C-band a comparably strong dihedral contribution in the dry month of June, which is to a certain extend still present in August on the harvested field, where short cornstalks are still left. In L-band, due to the deeper 18/01/2008 Page 189 of 259

191 penetration the dihedral GVR is already quite high in May, getting even higher in June, when the volume contribution strongly decreased because of the dry conditions. In August m d is very low indicating that the remaining cornstalks were too short as compared to the wavelength of L-band (λ L-band = 20 cm, λ C-band = 5 cm) to create dihedral scattering. high (10 db) low (-20 db) m m m m m m Figure 12.12: Intensity image of the surface m s (left) and dihedral m d (right) GVR at different dates at C-band high (10 db) low (-20 db) m m m m m m Figure 12.13: Intensity image of the surface m s (left) and dihedral m d (right) GVR at different dates at L-band Finally, the temporal behaviour of the polarimetric range of the GVR given by m min and m max has been determined for selected points in the fields 222 (corn) and 250 (wheat) at which also ground measurements had been taken simultaneously. The results have then been averaged for all selected points within one field and are given in Figure and In C-band, for the corn field a continuous narrowing of the polarimetric range of the GVR can be observed which is consistent with an increase of vegetation. Indeed, the corn grew mainly in June and July (days 18/01/2008 Page 190 of 259

192 151 to 212 of year (see Figure 12.15)). The wheat field on the right side of Figure starts with a rather small polarimetric range of the GVR, which widens around day 170 (19 June) and at the very end (2 August). The narrow region coincides with the time when the wheat was fully grown (see Figure 12.15). The June widening is assumed to be due to the dry conditions during that time (Figures 12.14) making the vegetation more transparent for radar. However, the final widening of the GVR range is clearly related to the harvesting, that took place during that time. L- band seems to be rather insensitive to the above described effects on the polarimetric range of the GVR. A very prominent feature in the lower left plot of Figure is the sharp decrease of both m min and m max at day 130 on the corn field. On this day the field had been ploughed which is presumably the reason for this behaviour. Interestingly, this cannot be observed in C-band. The highest difference between m min and m max could be observed in C-band on the corn field with a difference Δm of more than 20 db for bare surfaces and is decreasing to 10 db for dense vegetated areas. For L-band the strongest separation between m min and m max is Δm 19 db and is decreasing also to Δm 10 db. Though, the interpretation of the m s of both crop types is easier at C-band. Figure Precipitation during the AGRISAR campaign (blue background) estimated with a nearby by weather station. The dashed areas are periods, where intensively ground measurements were collected. crop height [cm] day of year corn day of year Figure 12.15: Crop height presented in time (April to August) for corn and wheat. crop height [cm] wheat 18/01/2008 Page 191 of 259

193 m max m min m max m i Figure12.16: Minimum (m min ) and maximum (m max ) GVR of the corn (left) and the wheat field (right) at C-band m max m min m max m i Figure 12.17: Minimum (m min ) and maximum (m max ) GVR of the corn (left) and the wheat field (right) at L-band Freeman-2-Component Decomposition Model description The Freeman-2-Component decomposition is based on the Freeman-Durden model which was recently developed [ 24 ] for vegetation cover modelling using simple electromagnetic models describing the different scattering contributions to the total vegetation scattering. With the 2- component model the polarimetric SAR backscatter return is decomposed into two scattering contributions: a volume and a ground component, each described by two parameters. The coherency matrix [T tot ] of the total backscatter is related to the scattering matrix [S] via the scattering vector k: S = S S ρ 1 HH HV S k = [ S ] T HH + SVV, SHH SVV, 2SHV VH SVV 2 18/01/2008 Page 192 of 259

194 [ T ] tot = k k + 1 = 2 S + S 2 VV * ( S + S )( S S ) 2 ( S + S ) * HV * 2 * ( SHH + SVV ) ( SHH SVV ) SHH SVV 2 ( SHH SVV ) SHV 2 ( + ) ( ) * * 2 S S S 2 S S S 4 S HH HH VV HV HH HH VV VV HH HV VV HH HV VV S [T tot ] is modelled as the sum of the two coherency matrices of the two different scattering mechanisms: [ T ] = T 21 T 22 0 = [ T ] + [ T ] tot mod el T11 T T 33 g v The first contribution is attributed to ground scattering. This can either originate from rough surfaces (Bragg scattering as described by the small perturbation model (SPM) [ 62 ]) or dihedrals such as ground-trunk reflections. In both cases the coherency matrix representation has the same form: 2 α [ T g ] = f g α 0 α * Whether the ground component is attributed to surface or dihedral scattering is determined by the value of α. According to its definition, α must be less than 1 for dihedral scattering and greater equal 1 for surface scattering (see Figure 12.18). 1000, ,000 dihedral surface 10,000 α 1,000 0,100 0,010 0, incidence angle [ ] Figure 12.18: Incidence angle dependence of the α parameter of the ground component for selected values of the dielectric constant α of ground α g and trunk α t (α g =20, α t =30, α=0). The red curve represents the dihedral and the blue curve the surface case. 18/01/2008 Page 193 of 259

195 The following equations show how f g and α are related to the Fresnel- respectively Braggcoefficients: surface scattering: 2 s m 2 f g = Rh + R and v 2 α = Rh + Rv R R 2 cosθ ε sin θ 2 2 s R = and ( ε s 1) ( sin θ ε s ( 1+ sin θ ) h R 2 v = 2 cosθ + ε s sin θ 2 cos sin ε s θ + ε s θ h v dihedral scattering: 1 2 iϕ RghRth RgvRtve f g = + and 2 R α = R gh gh R R th th R + R gv gv R R iϕ tve iϕ tve cosθ ε g sin θ R = and gh 2 cosθ + ε sin θ g 2 R th cosθ = cosθ + 2 ε sin θ t 2 ε sin θ t 2 ε g cosθ ε g sin θ R = and gv 2 ε g cosθ + ε g sin θ R tv ε t cosθ = ε cosθ + t 2 ε sin θ t 2 ε sin θ t In the case of surface scattering f g also depends on m s, which accounts for the wavelength and surface roughness [ 62 ]. The Fresnel and Bragg coefficients are defined for horizontal and vertical polarisation and depend only on the incidence angle Θ and the dielectric constant ε of scattering media. The power of the ground component P g is defined as the sum of the diagonal elements of the ground coherency matrix [T g ] and is hence: P 1 + α 2 g = f g The second contribution is derived from volume scattering. It is modelled by a cloud of randomly oriented particles. The shape of the particles is described by the so-called shape parameter ρ. It accounts for the ellipticity of the particles ranging from perfect dipoles (ρ=1/3) to a sphere (ρ=1) (see Figure 12.19). ρ=1/3 ρ=1 dipole sphere Figure 12.19: Relation between the shape parameter ρ and the particle shape. 18/01/2008 Page 194 of 259

196 The coherency matrix representation of the volume component is hence given by the two parameters ρ and f v : [ T ] = v f v 1 + ρ ρ ρ The power of the volume component P v is then: P v = f v ( 3 ρ ) The total power P tot from both components is: P tot ( ) 2 = P α g + Pv = fg 1+ + fv 3 ρ The calculation of the parameters is done by solving the system of equations and is described in more detail in [ 67 ]. The main difference between this model and the Freeman-Durden 3- component decomposition is that both ground contributions (surface and dihedral scattering) cannot be determined at the same time, but the volume component can be better described by generalising it using the shape parameter ρ. Moreover, the reduction from three to two components eliminates the requirement of making an assumption for one of the parameters. Negative Powers As can be seen from the physical definitions of the parameters, the powers of all components must be positive. However, since the model only approximates the measured signal to which also other scattering mechanisms may contribute, some f-parameters may evaluate to a negative number leading to a non-physical negative power. In this case the affected component will be set to zero. In order to preserve the total power, the remaining powers are calculated accordingly. 18/01/2008 Page 195 of 259

197 Decomposition results The 2-component decomposition described above has been applied to both C- and L-band data of the west-east track of the AGRISAR flight stripe for three different dates lying at the beginning, the middle and the end of the vegetation period of The powers of the surface, dihedral and volume components have been calculated. The results are shown in Figure and using an RGB presentation where the dihedral power corresponds to red, volume to green and surface to blue. For comparison, a Pauli image of the scene has been added to the left, where two fields are highlighted. For the two exemplarily fields, corn and wheat, a detailed investigation and discussion is presented. The decomposed images show a quite strong distinction between the three scattering mechanisms and are mostly uniform over single fields. Qualitatively, the resulting image is similar to the Pauli decomposed image, which is due to the fact that the Pauli components correspond in a first approximation to the same three scattering mechanisms. However, for the Pauli decomposition the separation into three components is less distinct leading to a much lower contrast than obtained by the 2-component decomposition. The main reason is that the SAR data are fitted to the model. In the three Freeman decomposed images changes due to vegetation growth in time can be very well observed (see Figure and 12.21). The corn field (222) for instance appears in both C- and L-band blue in May and June which indicates surface scattering, i.e. low or no vegetation. In August, however, clearly vegetation (green) and dihedral (red) dominates which is expected for corn plants. Indeed, field measurements confirm that the field was still bare at the beginning of May and that the growth of corn plants happened mainly in July. For the wheat field (250) the opposite behaviour can be seen and appears differently in C- and L-band: in May and June dihedral and surface components dominate whereas in August there is almost no component present. This fits well to the ground observations during this period (see Figures ): In May wheat has already grown to about 20 cm, but got considerably dryer (i.e. more transparent for radar) in June (see Figure 12.20). In C-band, the vegetation in May produced mainly dihedral scattering, but as the vegetation became dryer in June the signal could penetrate to the ground resulting in a predominantly surface scattering. In L-band, due to the longer wavelength the signal could already penetrate the vegetation in May causing surface scattering. However, as the vegetation became higher in June this gave rise to an increased dihedral component. On the 2nd of August, when the last image was taken, field 250 had already been harvested and it produced only very little backscatter. 18/01/2008 Page 196 of 259

198 dihedral flight direction volume surface near range (25 ) far range (55 ) 2006/06/ /05/ /06/ /08/02 Figure 12.20: RGB image of the Pauli (left) resp. Freeman 3-component decomposition at different dates at C-band dihedral flight direction volume surface near range (25 ) far range (55 ) 2006/06/ /05/ /06/ /08/02 Figure 12.21: RGB image of the Pauli (left) resp. Freeman 3-component decomposition at different dates at L-band 18/01/2008 Page 197 of 259

199 The shape parameter ρ Figure shows histograms of the shape parameter ρ of the corn and the wheat field. For the histograms not the entire area of the fields has been taken. For computational reasons just the largest rectangle within the field and parallel to the scene borders has been selected. For some pixels a shape parameter outside of the physical interval 1/3-1 was obtained from the Freeman decomposition. In this case the shape parameter ρ was set to 1/3. This is the reason why the area under the histogram curves is not always the same for each field the remaining points are at ρ =1/ /05/ /06/ /08/ /05/ /06/ /08/ /05/ /06/ /08/ /05/ /06/ /08/02 Figure 12.22: Histograms of the shape parameter of the corn (left) and the wheat field (right) for C- (top) and L-band (bottom). For details see text. The distributions for the shape parameter ρ vary significantly peaking at values between 1/3 and 0.8. Distributions in the low ρ range correspond to rather dipole-like particle shapes such as stems, branches, leafs and needles, thus standing for high vegetation. The further the peak of the distribution moves to 1 the more spherical the particle shape is. This corresponds to vegetation without a distinct particle shape such as low vegetation. 18/01/2008 Page 198 of 259

200 high (1) low (1/3) 2006/05/ /06/ /08/ /05/ /06/ /08/02 Figure 12.23: Shape parameter ρ at three different dates (see below images) and at C-band (left) and L-band (right) Focussing again on the corn and the wheat field the ground observations can in general be confirmed by the histograms in Figure and the intensity plots for the shape parameter ρ in Figure Attributing the high ρ-values to bare surface or low vegetation, the strong increase in vegetation height for corn in July (see Figures ) can be seen on the last (August) image in both C- and L-Band. The field (222) appears in blue which indicates the contribution from high vegetation. The harvested wheat field (250), however, shows in C-band rather high ρ -values in August which is in accordance with the low vegetation. The same field appears blue in May but with a significantly higher ρ -value in June. This is presumably due to the dry vegetation during this period leading to an increased surface contribution. In L-band the ρ -values seem in general to be higher than the low ρ -values in C-band. This is probably due to the longer wavelength of L- band (20 cm as compared to 5 cm in C-band) penetrating deeper into the volume showing more contribution from the ground which has in general high ρ -values. Though, it is difficult to understand why in L-band the harvested wheat field (250) has lower ρ-values than in the previous months (see also bottom right histograms in Figure 12.22). It is probably due to the effect that at very dry conditions, as there were in August, on a harvested field with still remaining parts of the wheat on the field the electromagnetic wave sees the soil surface as very rough and has therefore a higher ρ-values. Other interesting areas are the sugar beet fields that appear homogenously yellow to red in the upper right part of the very left image in Figure In May the fields were still bare but became vegetated later on. Indeed, both in C- and L-band the ρ-value changes from high in May to low in June and August. However, as sugar beet does not grow very high and has large leafs; it is surprising that it behaves as a dipole in terms of its ρ-value. The fully grown sugar beet looks very similar to the corn field (222). Hence the shape parameter ρ does not include information to distinguish between crop types. 18/01/2008 Page 199 of 259

201 Summary The Freeman 2- and 3-component decompositions have been applied on the full-polarimetric AGRISAR dataset at C- and L-band and at three representative dates in the vegetation period of It is shown that a separation of different scattering components can be successfully performed with both decompositions. The analysis for both different times and frequencies has delivered in-depth information about the state of vegetation and surface throughout the vegetation period. In general the statement can be made that C-band is less sensitive to ground contribution with vegetation growth as compared to L-band which is sensitive to the ground for much higher vegetation. It requires high and/or dense vegetation such as corn with 2 m height or higher before only the vegetation volume can be seen in L-band. A method how to quantify the ground contribution during the vegetation growth cycle was presented by the introduction of the ground-to-volume ratio m (GVR). Defining the minimum and maximum of GVR the difference is a measure for the detection of the vegetation contribution. Higher differences indicate that bare surface are dominant and lower differences indicate the presence of vegetation. The difference can be expressed in db and varies between 20 db for bare surfaces and 10 db for vegetated surfaces. It has been observed that C-band show a better interpretable variation in time as L-band for Δm, the reason for it could be that the presence of the ground over a vegetation layer is already so strong, that the sensitivity Δm is weaker. Since the 3-component Freeman-Durden decomposition is based on a physical model, it allows in principle determining the soil moisture by using the component parameters that are related to the dielectric constant of the soil as described above. Thus, the determination and verification of the different scattering mechanisms, as it was done in this work, is the decisive precondition for soil moisture estimation. The main problem with the 3-component Freeman decomposition is that the volume contribution is characterised by a fixed volume structure. Agricultural vegetation is known to be one of the most complex and diverse volumes and is therefore not properly described by the simplified vegetation model of the 3-component Freeman-Durden decomposition. In comparison to the 3-component Freeman decomposition in the 2-component decomposition the volume scattering component is described by a shape parameter that allows to identify differences in the vegetation volume and not to fix it to a defined volume. The variation of the shape parameter in time for C- and L-band can be very well observed over the two selected crop types. In summary it can be stated, that the variation of the shape parameter is not due to the variation of the vegetation structure but is better explained as a good filter between vegetated and non vegetated areas. Very clear this can be observed for the first date in both frequencies, where only two/three fields are non-vegetated and have a smooth surface (high shape parameter) from already vegetated area with a very low shape parameter. The shape parameter is not the appropriate parameter to distinguish different crop types, but it represents an important parameter to provide in addition information for the ground-to-volume ratio. That is an essential parameter for the estimation of soil moisture. 18/01/2008 Page 200 of 259

202 Extinction Estimation by Corner Reflectors Located on Crop Fields by University of Alicante In the experiment carried out, 10 corner reflectors (CR s) were deployed on the test site. All of them have a side length of 20 cm, and were approximately oriented to the radar (flights West- East with antennas pointing to the South) by using a compass. The CR s were deployed on fields 140, 222 and 230, corresponding to different crops (rape, corn and wheat) and at different vegetation stages. They were deployed on June 6 (during the 2nd intensive campaign) and recovered on July 11 (after the 3rd intensive campaign). Extinction can be measured by comparing the radar backscattering from a CR before and after vegetation grows. When there is not vegetation, we measure the RCS of the CR alone, but when vegetation grows in the crop field, the electromagnetic travel twice along the same path inside the vegetation volume. The presence of vegetation produces an attenuation of the waves, which is characterized by the extinction coefficient. Extinction, which depends on the wave polarization, is estimated by substracting both responses (before and after vegetation presence), and knowing the vegetation height, which was measured as part of the ground-truth acquisition campaign. In order to estimate extinction from the SAR images, the first step consists in identifying them in the images. After carefully examining all the SLC SAR images acquired at L-, C-, and X-band at the date of deployment, only the CR s located in the maize field were localized (note that it was almost a bare field at that date). For the maize field, the identification of the CR s was possible only at X-band, since at lower frequency bands the response of the CR s was below the clutter level. This problem is a consequence of the small size of the available targets. In addition, the CR s located at the wheat and rape fields were not visible on the images at any frequency band, because the backscattering from the vegetation, which was already tall at the time of deployment, obscured the response from the CR s. Consequently, extinction estimation was only possible for the maize field 222 and at X-band. After the identification of the CR s, time series of the backscattering coefficient at the CR positions were analyzed and combined with the evolution of the height of the maize plants for deriving extinction values. The time series of backscattering coefficient and height values are shown in Figures and The average clutter around each CR is also presented in the same plots. This clutter corresponds to the backscattered power from the surrounding crop, and it exhibits a rather high variability as a function of time. The whole time span of the campaign is shown in these figures, but the CR s were deployed on day 158, as it is noticed by a strong peak at that date. 18/01/2008 Page 201 of 259

203 Figure 12.24: Evolution of backscattering for VV polarization at the positions of CR s #0, #2 and #3, and vegetation height at the closest reference points. Figure 12.25: Evolution of backscattering for HH polarization at the positions of CR s #0, #2 and #3, and vegetation height at the closest reference points. Table 12.3 presents the extinction estimates for both vertical and horizontal polarizations. Results exhibit a high variability when comparing different CR s and different times. In principle, it is expected that the last estimations (for day 186) should be the best ones because vegetation is taller than before and, hence, the signal is more sensitive to attenuation and the estimate is less affected by small height errors. On the other hand, the first estimates (on day 164) should be noisier, as it happens for the horizontal case with CR #0, obtaining a stronger backscattering for this day than for the date of deployment, thus producing a negative extinction. Some exaggerated extinction values appear too. CR #0 CR #2 CR #3 DoY κ V κ H κ V κ H κ V κ H Table 12.3: Extinction estimates in db/m for the maize field at X-band. There are many aspects that evidence the rather low reliability of the results obtained in this experiment. To begin with, since only a few CR s were used, the statistical degree of confidence of the estimates is poor. This is even worse due to the presence of speckle in the SLC images. 18/01/2008 Page 202 of 259

204 In second place, we can observe in Figures that clutter (i.e. backscattering from the surrounding maize field) changes quite importantly between consecutive times. These changes reach +/- 4 db from one image to the next, which is much larger that the precision required for obtaining good extinction estimates by this technique. In third place, there exist a number of external factors that may affect the working principle of this approach. For instance, as time passes the radar response from the CR s may degrade due to: a) orientation changes produced by the vegetation when it grows or by the effect of rain, b) there appear accumulated water and soil due to rain and wind, etc. An additional uncertainty factor is the variability of vegetation height within the same crop field (up to 1 m difference within the same maize field), so the heights employed in the estimation may not correspond to the plants crossed by the waves at each CR position. As preliminary conclusions and suggestions for future experiments on direct extinction estimation by using the SAR images, we can state the following: 1) The temporal window covered by the presence of CR s should include the whole development period, from showing to just before harvest. 2) Larger CR s are required for increasing both identification and estimation accuracy. The main limitation is that the CR s have to be deployed in the field by minimizing their influence of the development of the plants. CR s with a 40 cm side would guarantee right measurements at C and X-band. Evolution of Polarimetric Signatures as a Function of Time A temporal polarimetric analysis on maize, winter rape and winter wheat crops of Demmin test site has been performed in order to complement the study carried out in [ 33 ]. Fully-polarimetric measurements at L-band enable the computation of polarimetric ratios, correlation magnitudes and phase differences for the linear polarization basis and also for the circular polarizations. In case of C-band, dual-polarized data were collected and, therefore, the phase information is only maintained for the combinations HV/HH and VH/VV for the whole growing season. An analysis from the qualitative viewpoint of the correlation between those indicators and the biophysical parameters as a function of plant development has been carried out. Additionally, a comparison with other results reported in the literature has been also considered, in particular regarding the potential of polarimetric C-band data for maize [ 4 ] and wheat [ 7 ] crops. Next, the most representative results are shown here. At left side of Figure the correlation magnitude between HV and HH at L-band for the maize crop follows a clear agreement with plant height, whereas circular polarizations can be used to detect maize plant emergence, as shown in right plot of Figure 12.26, where the RL/LL ratio shows a 6 db decrease from DoY 164, i.e. the moment when the leaves start to appear and stems height changes from 20 cm to 40 cm, to DoY 172, which corresponds to a more advanced development stage of stems and leaves. In case of winter rape (plots not shown here) a 5 db decrease appears in both circular crosspolar ratios when the plant grows from 20 cm to 1 m. After that moment (DoY 136), the ratios vary no more than 2 db. 18/01/2008 Page 203 of 259

205 Figure 12.26: Plots of (left) HV vs HH correlation magnitude and (right) RL/LL ratio of maize crop at L-band. The dashed line represents the crop height. Figure represents the correlation magnitude and phase difference between RL and RR channels for rape crop at L-band. Note that RL-RR correlation magnitude is above 0.4 and the phase difference shows a quasi-linear trend to π radians from DoY 155. On the other hand, the RR-LL phase difference (not shown here) is constant at -π radians for the whole period, with an increasing correlation magnitude as a function of plant development. Additionally, it is observed that the temporal evolution of correlation magnitude for both cases (see left plot in Figure 12.27) follows a very similar trend when compared with ground-truth measurements of biomass [ 66 ]. Figure 12.27: Plots of RL vs RR (left) correlation magnitude and (right) phase difference of rape crop at L-band. The dashed line represents the crop height. Results at C-band are partially in agreement with previous works reported in the literature for maize and wheat crops. The left plot of Figure shows that the dynamic range for VH/VV ratio for maize crop is about 2 db from DoY 164 (where the leaves start to appear) to the end of 18/01/2008 Page 204 of 259

206 the growing season, and this corresponds to a LAI range of 2 m 2 / m 2, which is in agreement with the simulated values in [ 4 ]. Additionally, it seems that there exists a correlation with biomass up to kg/ m 2, depending upon the measurement point in the field. The maximum biomass value provided in [ 4 ] is 6.5 kg/m 2. On the other hand, the variation of copolar ratio VV/HH for the winter wheat, shown in right plot of Figure 12.28, partially agrees with the results reported in [ 7 ]. In that case, a linear relationship was found between copolar ratio with a 5.5 db dynamic range and biomass up to about 3.25 kg/m 2. In our case the copolar ratio is about 3 db for a maximum biomass about kg/m 2, which corresponds to DoY 136. Figure 12.28: Plots of (left) VH/VV ratio for maize crop and (right) VV/HH ratio for winter wheat crop at C-band. The dashed line represents the crop height Land cover classification by TUB The multi-temporal SAR data set acquired during the AgriSAR experiment is well-suited to assess the potential for land cover classification of a SAR mission with high repetition frequency, like the future Sentinel-1 mission. Below is shown preliminary results of using the data set to compare the results of a multi-temporal classification with a classification based only on a single acquisition. The results are shown for C-band, i.e. the frequency planned for the Sentinel-1 mission. The multi-temporal SAR data along the East-West track were analysed. The scattering matrix data in the form of SLC products were coregistered, converted to covariance matrix elements and multilooked. A pixel spacing of 3 m and an equivalent number of looks of approximately 10 were obtained. With a number of looks of 10, a Gaussian assumption for the probability density function for the backscatter coefficients for the individual polarisations is valid. The classification method used for the single and dual polarisation cases is therefore the standard Baysian classification method for multivariate Gaussian statistics. One training polygon for each of the agricultural land cover classes, i.e. beets, maize, winter barley, winter wheat and winter rape, has been selected. All other polygons have been used for an independent assessment of the classification accuracy. Figure shows the polygons used for the classification assessment. 18/01/2008 Page 205 of 259

207 Figure 12.29: Polygons used for the classification assessment. The classification potential for the different polarizations and acquisition times is evaluated by computing the average classification error for each case. So for a specific combination of these parameters the data are classified, and as explained above, the classification method used for the single and dual polarisation cases is the standard Baysian classification method for multivariate Gaussian statistics. The number of wrongly classified pixels is then determined for each class, and an average error is computed. Figures show the classification errors for the single polarization backscatter 0 coefficients γ vv and γ 0 hv, and the dual-polarisation backscatter coefficients (γ 0 vv,γ 0 xp ) for C-band. For each of the backscatter coefficients, two sets of classification errors are shown, i.e. singleacquisitions errors, and multi-temporal errors. The white columns show the classification error if only the single acquisition indicated with the Julian day is used for classification, whereas the dashed columns show the classification error if all acquisitions up to and including the one indicated with the Julian day is used for classification. These classification results clearly demonstrate the advantage of multi-temporal SAR acquisitions for land cover classification. The classification errors for single date acquisitions are fairly high and constant for all acquisitions. When the multitemporal acquisitions are taken into account, a decreasing classification error with acquisition time is observed for all three sets of parameters. The multi-temporal classification results using all acquisitions for all single and dual polarization cases are shown in Table The best overall performance is obtained for the cross-polarized channel, and for the dual-polarization case, the (γ 0 vv,γ 0 xp ) combination is the best. These results clearly support the idea of a SAR mission with a high repetition frequency, like the planned Sentinel-1 mission. 18/01/2008 Page 206 of 259

208 Classification error VV_C_SNGL VV_C_MTMP Figure 12.30: Classification errors for the VV backscatter coefficient at C-band Classification error XP_C_SNGL XP_C_MTMP Figure 12.31: Classification errors for the XP backscatter coefficient at C-band Classification error VX_C_SNLG VX_C_MTMP Figure 12.32: Classification errors for the VV and XP backscatter coefficients at C-band. 18/01/2008 Page 207 of 259

209 C HH 22,8 VV 18,5 XP 6 HHVV 16,1 HHXP 10,5 VVXP 9,5 Table 12.4 Classification accuracies for the single and dual polarisation results for the multitemporal combinations LAI and Soil Water content by University of Naples Multi-look geocoded E-SAR images in bands C, L and X acquired during the flight of July 5th over the Demmin site have been considered for the present analysis. Images resolution has been degraded to 20 m with pixel value corresponding to the mean. In order to investigate on the relationship between canopy development and radar backscattering, we have considered the LAI map derived from the inversion of a canopyradiative transfer model applied to the image acquired over the Demmin site on July 5th by the Compact Airborne Spectral Imager (see paragraph 12.2). Due to the quality of image data and the elaboration performed to derive the LAI map, this map has been considered has the groundtruth for our subsequent analysis. In analogy with the radar images, the spatial resolution of the LAI map has been degraded to 20 m; in addition, we have considered the value of NDVI from the same CASI image at the resolution of 20x20 m. The correlation analysis has been carried out by calculating the Pearson coefficient for several fields with a range of different crops. As it might have been expected, the correlation results for LAI and NDVI were similar, with slightly higher values of the Pearson coefficient for NDVI compared to LAI. We have noticed that the best results were those corresponding to L-band at VH and VV polarisations. While in the case of maize (field 222) the correlation values are similar for all the three bands, the highest correlation for the sugarbeet field has been found in band L- VV. In the case of winter wheat fields, we obtained the best correlation in band L with HH polarisation. Only in the case of field 460 (sugarbeet) X-band data performed better other bands, probably due to the high moisture content of the plant leaves. Overall, these results suggest that L-band, especially in vertical polarisation, is more suitable than other SAR configurations to monitor canopy development. The second part of our analysis has been focused on the surface soil water content. To this end, we have applied the semi-empirical model of Oh to estimate the soil apparent dielectric permittivity. This model can be applied without a-priori information on soil roughness, which is a significant advantage compared to other methods. Although the model has been conceived for bare soils, we have tested its application to fields 102 and 222, having similar LAI values for vegetation cover. The summary statistics resulting from this test are presented in Table /01/2008 Page 208 of 259

210 Field 102 measur. estim. C - band estim. L - band mean st.dev field 222 measur. estim. C - band estim. L - band mean st.dev Table 12.5: Summary of statistics for estimation of apparent dielectric permittivity by using Oh s model on Agrisar data. L-band data give a better estimation of the field mean value compared to C-band for both crops; however, the single values of ε are significantly scattered around the mean. This large scatter, as indicated by the standard deviation values, is similar in both bands, and it is likely to be related to the water content of vegetation cover. However, the preliminary test carried out in this study has confirmed that L-band data applied to the semi-empirical model of Oh may provide an estimation of the average value at field scale even in presence of a vegetation cover in different crop types. The limitations of the present study mainly rely on the small variability of field conditions during the July 2006 Agrisar campaign in Demmin, either for the canopy development, i.e. range of LAI values, either for the soil water content Optics Analysis of Chlorophyll by University of Valencia Within the overall context of ESA AGRISAR experiment, the available spectral information from the airborne imaging spectrometers (AHS-INTA and CASI-1500) has allowed the simulation of Sentinel-2 products. A second campaign objective was the preparation of the Sentinel missions, the GMES component of future operational missions (follow-on of Landsat type of missions oriented to operational applications). The baseline is a kind of Landsat-TM type of instrument, and the campaign was used to define enhancements in spectral capabilities for such Sentinel mission over current Landsat capabilities. In this framework, Sentinel-2 products was simulated (see section in this report) to evaluate their spectral bands capabilities exploring the possibilities offered on retrieving vegetation properties such as LAI (leaf area index), fractional vegetation cover, canopy water content and canopy chlorophyll content. Chlorophyll content can be consider one of the main inputs in the vegetation models development, when photosynthetic processes at leaf level are described by means of the different pigments involved on them. Thus, it is considered as indicator of the photosynthetic efficiency of the plant. Current methods for estimating chlorophyll content of vegetation canopies are based on the relationship between biophysical properties and spectral measurements [ 3 ] in some few bands 18/01/2008 Page 209 of 259

211 acquired by multi-spectral systems. A large number of spectral indices have been tested with the available spectral information to explore the implications on retrieving Chlorophyll through the use of the Sentinel-2 spectral bands. From the analysis, capabilities of Sentinel-2 spectral bands to retrieve chlorophyll have been evaluated, suggesting potential ways to improve the possibilities offered by Sentinel-2 products. Furthermore, a sensitivity analysis of the spectral bands has been performed comparing the results to the CASI-1500 full spectral data. Biophysical parameters Chlorophyll content, LAI and other biophysical parameters where weekly characterized (by ZALF) and geo-located, at different Elementary Sampling Units (ESUs), within the AGRISAR experiment. Furthermore, Uni Valencia team sampled a set of ESUs during the intensive field campaigns (Table 12.6). Each ESU is considered as the mean value from the 30 measurements carried out in an area of 25m inside the ESU. Those measurements were performed with the two SPAD-502 calibrated chlorophyll meters. The calibration functions applied and the analysis of the results from calibration procedure is detailed at section in this report. The LAI data acquired by ZALF has been used for the sensitivity analysis of the spectral bands. ZALF ESU X Y coord UV ESU X Y coord Rape 140-P Corn Rape 140-P Corn Rape 140-P Corn Corn 222-P Corn Corn 222-P Corn Corn 222-P Wheat Wheat 230-P Wheat Wheat 230-P Sugar beet Wheat 230-P Sugar beet Sugar beet Sugar beet Sugar beet Sugar beet Sugar beet Table 12.6: GPS coordinates of the different ESUs sampled for chlorophyll measurement 18/01/2008 Page 210 of 259

212 Figure 12.33: Elementary sampling units for different crops where leaf chlorophyll content was measured during the July AGRISAR-intensive field campaign. CASI and Sentinel-2 simulation images As it is described at , Sentinel-2 simulation products (S2S) were derived from CASI imaging data acquired during the AGRISAR intensive campaigns. Sentinel-2 spectral bands configuration is shown at Table A set of 23 vegetation regions of interest (ROI), each one of them three pixels sized, have been selected on CASI and S2S images (acquired on 5th July, 2006), focused on the ESUs characterized for chlorophyll in-situ data. The selected dataset included: 3 rape ROIs, 8 corn ROIs, 5 wheat ROIs and 7 sugar beet ROIs (Figure 12.33). The spectral profile of each one of the ROIs was obtained from the image, as well as, the mean value with the standard deviation was calculated for the spectral data retrieved from the pixels included inside each area (ROI). Spectral band B1 B2 B3 B4 B5 B6 B7 B8 B8a B9 B10 B11 B12 λ centre (nm) Width band (nm) Table 12.7: SENTINEL-2 bands configuration. Red colour for bands used for atmospheric corrections. 18/01/2008 Page 211 of 259

213 Figure 12.34: Difference between CASI and S2S spectra Chlorophyll Indices Several spectral indices commonly used to retrieve chlorophyll [ 6 ] have been applied to CASI and S2S data [ 25 ] (RN means reflectance at N nm): - Integral Index: The area intersected between a vegetation spectrum and its corresponding soil spectrum is an indicator of vegetation chlorophyll content. Consequently, the area under vegetation reflectance curve is related to chlorophyll content when soil spectrum can be considered constant. Figure illustrates the integral calculated in this analysis. The limits of the integral have been placed between 600 nm and 700 nm. Figure 12.35: Area between soil and vegetation spectra and area under vegetation spectrum as chlorophyll content indices - Quotient Index: as an example it has been calculated the index defined by [ 28 ] as: R750 GMI= R700 For CASI and S2S the respective index becomes: R749.1 CASI GMI= R699.8 B6 S2S GMI= B5 - Modified Chlorophyll Absorption in Reflectance Index (MCARI), defined as [ 16 ]: 18/01/2008 Page 212 of 259

214 R700 MCARI = ( R700 R670) 0.2* ( R700 R550 ) * R670 For CASI and S2S the respective index becomes: R700 CASI MCARI= ( R699.8-R669.2) -0.2* ( R699.8-R549.4 ) * R669.2 B5 S2S MCARI= ( B5-B4) -0.2* ( B5-B3 ) * B4 - Transformed CARI (TCARI), defined as [ 31 ]: R700 TCARI = 3* ( R700 R670) 0.2* ( R700 R550 )* R670 For CASI and S2S the respective index becomes: R699.8 CASI TCARI=3* ( R699.8-R669.2) -0.2* ( R699.8-R549.4 )* R669.2 B5 S2S TCARI=3* ( B5-B4) -0.2* ( B5-B3 )* B4 - MERIS Terrestrial Chlorophyll Index, defined as [ 15 ]: MTCI = ( R R708.75) ( R R ) For CASI and S2S the respective index became: ( R R709.2) ( ) CASI MTCI= R R681 ( B6 - B5) ( ) S2S MTCI= B5 -B4 - Optimized Soil - Adjusted Vegetation Index (OSAVI), defined as [ 51 ]: 1.16* ( R800 - R670) ( ) OSAVI= R800 - R For CASI and S2S the respective index became: 1.16* ( R R669.2) ( ) CASI OSAVI= R R * ( B8 - B4) ( ) S2S OSAVI= B8 - B /01/2008 Page 213 of 259

215 - TCARI/OSAVI [ 31 ] CASI and Sentinel-2 Simulation Correlations In order to test the accuracy in the retrievals by the different methods, CASI spectral information is taken as optimum reference, and Sentinel-2 simulation products are compared with CASI dataset as a performance criterion. Figure shows the correlations found between CASI and S2S from applying the integral index, GMI, MCARI, TCARI, MTCI and OSAVI. 18/01/2008 Page 214 of 259

216 Figure 12.36: Results from applying different chlorophyll indices for CASI and SENTINEL-2. Due to the Sentinel-2 spectra bands configuration, band 600 is lost and an interpolation was necessary to calculate the integral between 600 nm and 700 nm. Errors about 10% has been assumed for CASI and Sentinel-2 indices. Good correlations were found for all the indices (r > 0.95). Chlorophyll Retrievals In order to analyze how many those calculated indices correlates to the chlorophyll content at crop level, ROI mean values from chlorophyll data were compared to the Integral Index, GMI, MTCI, MCARI, TCARI and OSAVI results obtained for CASI and S2S (Figure 12.37). Mainly, low correlations coefficients can be found for all indices (r <0.5) for both of the sensors. The reason of this can be found in the little interval of variation considered for the chlorophyll data, which only ranges from 25 to 45 µg cm -2 due to the small number of different crops included in the analysis. When a wider range of chlorophyll content was considered, better correlations was obtained (r = 0.8) for the cases in which chlorophyll in-situ values are represented versus the integral index, applied to CASI spectral data, when the data from SPARC-2004 and AGRISAR-2006 experiments are included (Figure 12.38). 18/01/2008 Page 215 of 259

217 Figure 12.37: Correlations between ground truth chlorophyll data and chlorophyll indices applied to CASI and Sentinel-2. 18/01/2008 Page 216 of 259

218 Figure 12.38: Correlation found between ground truth chlorophyll data and integral index from CASI spectral data from SPARC-2004 and AGRISAR-2006 field campaigns Figure shows the correlations found between chlorophyll in-situ data and the ratio TCARI/OSAVI from CASI spectral data. When it was compared to the relation proposed by [ 31 ] the AGRISAR-2006 data approaches it. Figure 12.39: Correlation between ground truth chlorophyll data and TCARI/OSAVI from the AGRISAR data. Crop Chlorophyll vs. Indices Furthermore, the Chlorophyll content-lai (Chl*LAI) product is considered as indicator to determine the total chlorophyll in a crop, and results from this product were correlated to the Integral Index obtained from S2S. Values of LAI applied to the Chl*LAI product were measured at two different time intervals (early in the morning and during the evening, between 18:30 and 18/01/2008 Page 217 of 259

219 20:30 h-lt). Low correlations are found for both of the cases, but also some improvements can be observed for the results from the LAI measured during the evening. Figure shows the correlations found between Chl*LAI and the rest of the indices. The correlation coefficients obtained provided better results than when considering chlorophyll in-situ data versus the indices (Table 12.8). Due to the strong correlation obtained for the indices from CASI and S2S images, only S2S correlations have been considered in these results Chl*LAI (morning) Chl*LAI (evening) Chl*LAI r = r = S2S Integral Index Figure 12.40: Chl*LAI correlated to integral index from S2S. 18/01/2008 Page 218 of 259

220 Figure 12.41: Correlations between Chl*LAI and S2S chlorophyll indices Chlorophyll indices r (Chl vs. index) r (Chl*LAI vs. index) Integral GMI MCARI TCARI MTCI OSAVI Table 12.8: Correlation coefficients obtained from fitting chlorophyll and Chl*LAI data to the spectral indices Using CASI for LAI estimation in Sentinel-2 configuration by University of Naples The preliminary analysis carried out has been focused on the evaluation of models developed for the estimation of the Leaf Area Index (LAI) with the spectral sampling proposed for the Sentinel- 2 multi-spectral mission, developed by ESA in the framework of GMES. To this aim the hyperspectral image acquired with the Compact Airborne Spectrographic Imager (CASI 1500, ITRES Research Limited) on July 5th 2006 was analysed, with spectral calibration and atmospheric correction provided by the Laboratory for Earth Observation, University of Valencia. The widespread SAIL (canopy) and PROSPECT (leaf) models were selected. These combined models have been inverted by using two different algorithms: an iterative optimization technique and a simple and fast look-up table (LUT) approach. Two different band combinations were used. Set 1 ( B1 ) contained of 7 out of the 220 CASI bands, chosen according to the literature results, as the best band selection to characterize vegetation. The second waveband set ( B2 ) used for the calculations corresponds to the proposed spectral sampling for the Sentinel-2 sensor (bandwidth): 490 nm (65 nm), 560 nm (35 nm), 665 nm (30 nm), 705 nm (15 nm), 740 nm (15 nm), 775 nm (20), 842 nm (115 nm) and 865 nm (20 nm). CASI bands were averaged in a bandwidth of 10 nm, and, in order to take into account of the planned spatial resolution for Sentinel-2 (4 bands 10 m, 4 bands 20 m), high spatial resolution CASI data were degraded to the coarser resolution of 20 m for both band sets. 18/01/2008 Page 219 of 259

221 The measurement configuration used for the model simulations presented the actual condition during the sensor overpass with a solar zenith angle of 35 (field 102) and 34 (field 222) and a view zenith angle of 0 according to the almost-nadir position of the plane in respect to the field. The resulting LAI map is shown in the figure below. The validation of the proposed methods has been carried out by using LAI field measurements of two different crop types: sugar beet field (id 102) with the size of 17.5 ha and maize field (id: 222) of ha. Since the results obtained from model inversion may differ depending on the implemented algorithm, two different criteria were applied. Firstly, a simple cost function to identify the results from the LUT approach has been chosen; secondly, an optimization algorithm, namely Sequential Quadratic Programming (SQP), has been implemented by using the standard MATLAB function. This function solves a quadratic programming sub-problem iteratively. In addition to the previous approaches, based on the radiative transfer models SAIL and PROSPECT, the results obtained by means of a simple semi-empirical logarithmic relationship between LAI and the Weighted Differences Vegetation Index (WDVI) have been compared. The figure below shows the root mean square error (RMSE) between simulated and measured LAI for the sugar beet and maize fields (102 and 222). 18/01/2008 Page 220 of 259

222 Absolute RMSE values between measured and estimated LAI values for sugar beet (102) and maize (222) fields using the different algorithms and band combinations: LUT, SQP algorithm, WDVI LAI relationship, B1: selected bands combination proposed by literature studies; and B2: band combination proposed for Sentinel-2 multi-spectral mission. These results indicate firstly that using the physical approach, LAI estimation can be performed more accurately than with a statistical-empirical model. Secondly, the choice of the optimal spectral sampling seems more essential than the inversion algorithm itself. The use of too many or unsuitable bands seems to introduce noise instead of adding important information about canopy characteristics, confirming the results obtained by previous studies. Some remarks have to be made concerning the inversion techniques: despite the use of the same merit function and range of input variables, the LUT method achieves a more accurate solution than the SQP approach. This may be due to drawback of the iterative optimization methods to converge in a local minimum. Considering the focus of Sentinel-2 developers to improve surface monitoring for various applications and to assure data continuity of the SPOT and Landsat missions, the choice of the Sentinel-2 proposed spectral bands revealed a quite high LAI estimation accuracy compared to best spectral sampling ( B2 RMSE: vs. B1 RMSE: 0.308) for vegetation and crops applications as proposed in previous literature. As such, the results of this analysis may provide the basis for developing an operational physical based model for the retrieval of LAI from Sentinel-2 sensor data Water and Energy Budget Analysis by ITC The observations from the airborne surveys conducted by the Spanish Instituto Nacional de Tecnica Aerospacial (INTA) on 5 th of July 2006, collecting high resolution hyperspectral optical imagery in the visible, near infrared, and thermal infrared are used to estimate the water and energy budget over several fields using an adapted version of the TSEB model [ 44 ]. The computation of the surface energy flux components using TSEB were performed for two overpasses during July 5 th, 2006 around midday. The first overpass over the two fields where 18/01/2008 Page 221 of 259

223 ground observations of surface fluxes were made took place at 10:20 and the second one was at 10:31 utc at an elevation of 9050 ft, resulting in a spatial resolution of 5.7 m. Thermal infrared imagery from the Airborne Hyperspectral Scanner (AHS) was employed to extract Land Surface Temperature (LST) maps and horizontal vegetation density (Fractional cover, Fc), whereas imagery from the Compact Airborne Spectrographic Imager (CASI-1500) of the Canadian ITRES was used to derive vertical (Leaf Area Index, LAI) vegetation density maps. Data processing The Temperature and Emissivity Separation (TES) method [ 26 ] originally designed for the ASTER sensor on board the TERRA platform has been applied on the AHS thermal data to derive land surface temperature estimates (supplied by Global Change Unit-University of Valencia). The method uses atmospherically corrected data and a semi-empirical relation determined form laboratory spectra, between the minimum emissivity (ε min ) and spectral contrast (maximum-minimum difference, MMD). Since this relationship is originally obtained for the ASTER sensor it needs to be recalculated for using it on the AHS data. The relation developed by [ 58 ] is then used on the thermal infrared AHS bands, which are located in the 8 to 12 μm region. In this study AHS data is used to derive NDVI following: NDVI = ( R ( R nir nir R R red red ) ) where R nir and R red are the reflectances measured in the near infrared and red wavelength regions respectively (channels 12 and 9 of the AHS sensor). Following [ 27 ] and [ 9 ] this NDVI is then scaled to derive fractional vegetation cover following: ( NDVI NDVI FVC = ( NDVIveg NDVI soil 2 ) ) soil where NDVI soil and NDVI veg correspond to the values of NDVI for bare soil (LAI equals 0) and a surface with a fractional vegetation cover of 100%, respectively. Assuming a random canopy with a spherical leaf angle distribution following [ 44 ], the desired fraction of the radiometer view that is occupied by canopy, f c (θ), is given by: f c ln (1 FVC) ( θ ) = 1 exp cos( θ ) 2 where θ represents the radiometer viewing angle, readily obtained from the products delivered by INTA. 18/01/2008 Page 222 of 259

224 The so-called MCARI-1 index has been applied in this study to derive the LAI maps (supplied by Remote Sensing Unit-University of Valencia), using the CASI acquired image from the 5th of July. The proposed MCARI 1 index is an improved new vegetation index modified version of MCARI for green LAI predictions. It is developed by [ 30 ] to render this index less sensitive to chlorophyll effects, more responsive to green LAI variations, and more resistant to soil and atmosphere effects and it provided the best results when LAI was retrieved from the tested images dataset. The MCARI-1 index follows the next expression: MCARI1 = 1.2 (1.5 ( R800 R670) 1.3 ( R800 R550)) where R 800, R 670 and R 550 are respectively the reflectance measured for wavelengths values of 800nm, 670nm and 550 nm. In order to apply this expression to derive maps of LAI from CASI data, those spectral bands were selected that corresponded closest to the wavelength values proposed by [ 30 ]. Two additional steps were necessary to derive the maps of LAI from CASI airborne data. First the MCARI-1 index was applied to the data acquired on the 5th of July. The CASI spectral bands selected to obtain R 800, R 670 and R 550 were the bands located at 800 nm, 663 nm and 551 nm respectively. As a second step, the LAI values were derived from the MCARI-1 product by fitting to the function: LAI = a exp b MCARI1 where LAI represents field observations of LAI and MCARI1 is the CASI derived MCARI1 product, whereas a and b originate from the fitted regression. Results Model output for net radiation and soil heat fluxes was rather satisfying with slightly overestimated soil heat fluxes by the model. The two sites covered showed very high vegetation cover and consequently very low soil heat fluxes were observed. The semi-empirical relations used in the two-source model are slightly over-estimating the soil heat fluxes over such dense vegetated areas. However, emphasis was on the turbulent fluxes, which are plotted versus ground observations in the left panel of Figure 12.42, together with the net radiation and soil heat fluxes. 18/01/2008 Page 223 of 259

225 E N N E M odelled Rn G H LE Observed E N N E H [Wm-2] Figure 12.42: Model output versus ground observations for all energy balance components in the left panel, spatially distributed sensible heat flux output on the right. Note the lower sensible heat fluxes in the middle of wheat field 250, where a local depression occurs. Sensible heat flux output showed remarkably low values considering the relatively dry and hot conditions, in parts even stable conditions occur over seemingly wet areas in between fields 250 and 222, see right panel of Figure The high values occurring in a patchy pattern stem from built up areas, whereas in addition, small bare spots inside the fields also show relatively high sensible heat fluxes. Sensible heat fluxes over the corn field are slightly higher than for the wheat field, but both are still rather low (slightly over 100 W m -2 ), which is confirmed by the ground observations. Latent heat fluxes showed remarkably high values considering the dry conditions, however, also these are confirmed by the ground observations as can be seen in the left panel of Figure Obviously the crops transpire at a potential rate, indicating that water is mainly being used from the deeper soil layers. This is confirmed by the model that is able to discriminate between soil evaporation and vegetation transpiration; in these areas soil evaporation was negligible as compared to canopy transpiration. Temporal integration of ground observations made over winter wheat field 250 for the 5 th of July yielded an actual evapotranspiration rate of 5.3 mm. The approach suggested by [ 56 ] offered the possibility to convert instantaneous model output to daily amounts of actual evapotranspiration using the so-called evaporative fraction. In the current case, this also yielded a rate of 5.3 mm per day (which exact coincidence might be considered slightly fortunate). 18/01/2008 Page 224 of 259

226 12.3 Land Processes Integration of RS-Derived Information into Hydrological Models by LHWM Introduction During the last decades the benefit of Synthetic Aperture Radar (SAR) data for the modeling of land surface processes has become widely accepted. The advantages of SAR data in this respect are twofold. On the one hand, a number of parameters needed for this type of modeling, for example land cover parameters, can be retrieved directly, and used as model input. On the other hand, remotely sensed observations of model outputs, more specifically surface soil moisture values, can be used to validate model outputs. Further, these data can also be assimilated into the models, reducing the error in the model predictions. Ever since the pilot study of [ 37 ] a large number of studies have put hydrologic data assimilation in practice. It is clear that up till now, the focus of SAR remote sensing has been on the retrieval of land cover parameters and on the estimation of model variables, more specifically surface soil moisture values. One relatively unexplored issue is the retrieval of soil hydraulic parameters, such as for example hydraulic conductivity values, through remote sensing. This is due to the fact that no direct relationships between the remote sensing observations, more specifically the radar backscatter values, and the parameter values can be derived. However, land surface models can provide these relationships. Attempts have already been performed to estimate saturated hydraulic conductivity values bypassing the use of a land-surface model, for example through the extrapolation of a regression between changes in soil water content and the hydraulic conductivity [ 42 ], or the application of neural networks to multiple drying cycle brightness temperature data [ 11 ]. However, these studies suffer from the drawback of the requirement of in-situ measurements of the remotely sensed soil parameter in order to establish the regression relationships. Further, since it is well-known that soil parameter values are scaledependent, a discrepancy between the scale of observation and the scale at which the parameter values are required will always exist. [ 52 ] showed that soil moisture observations can be used to obtain the soil textural composition, but did not apply their methodology in order to retrieve spatially distributed parameter values. The objective of this paper is to develop a methodology the estimate soil hydraulic parameters at the spatial scale at which they are required, bypassing the need to collect in-situ soil moisture data or parameter values. The wealth of data from the AgriSAR 2006 campaign render this data set very useful for this purpose. The Parameter Estimation Procedure The methodology to estimate the soil hydraulic parameters is the following. First, the hydrologic model, TOPLATS, was applied to the E-W track of the Demmin test site, using the available soil and land cover maps. A comparison of the modeled soil moisture fields to remotely sensed 18/01/2008 Page 225 of 259

227 observations is then performed. These model runs are performed at a 25 m resolution. Then, for each individual pixel, a separate soil class has been defined. For each of these pixels, the most important soil parameters are adjusted, so the modeled surface soil moisture values match the observations as closely as possible. A sensitivity analysis revealed that, for this test site, the saturated hydraulic conductivity (Ks), the pore size distribution index (λ), and the bubbling pressure (γc) were the most important soil parameters. For this reason, these three parameters were calibrated. The Kalman-filter based algorithm as outlined in [ 29 ] was used for this purpose. An analysis of the retrieved parameters is then performed, including the search for a relationship between the parameter values and the original soil texture class. Finally, the improvement in the modeled soil moisture values using the new parameter values was quantified. Application of the parameter estimation algorithm in a full three-dimensional manner would require a large parameter vector, with in the order to entries. This would make the matrix operations (especially the matrix inversion) extremely difficult. Therefore, a simplified method was applied. Since measurements indicate that the water table was rather deep below the surface throughout the experiment, and thus had a very limited impact on the soil moisture contents of the top soil, the calibration procedure was applied to each pixel separately, instead of to the fully distributed simulations. The resulting parameter values were then stored, and a fully distributed model run was then performed using the new parameters. Further, the soil moisture observations of two different pixels were assumed to be the same if, for every time step for which remotely sensed soil moisture values were available, the differences between the observations for the two pixels were lower than 2.5%. The parameter estimation procedure thus had to be applied to only one pixel instead of to both. These simplifications resulted in a strong reduction of the required computational effort. Also, the parameter estimation procedure was only applied to those pixels for which at least six observations were available. The same initial parameter values were used for each pixel. These were 0.5 for λ, 0.3 m for γc, and 10 mm h 1 for Ks. Results The Baseline Run The meteorological data needed for the model application were provided by the meteorological station at Görmin. The soil parameters were determined based on the texture class following [ 47 ]. The land cover parameters for each vegetation class were determined following [ 46 ]. For each land cover class Leaf Area Index (LAI) values were observed a number of times throughout the study period. These LAI values were assumed to be representative for all fields with the same land cover class and were used as model input. Model simulations were performed with 1a spatial resolution of 25 m and a time step of one hour. Figure shows the comparison of the modeled soil moisture to the remote sensing observations for the first three overpasses. Figure shows the same comparison for the second three overpasses, and Figure shows this comparison for the final three overpasses. All these three plots show that, while the observed values are well distributed, the simulated values are strongly centered around two values. This can be explained on the one hand by the fact that the soil moisture observations are 18/01/2008 Page 226 of 259

228 only available for winter wheat fields, which cancels the land cover type as a potential source of variability in the model results. Further, only four different soil texture classes were available for all pixels for which remotely sensed soil moisture values were available. Table 12.9 shows that the two dominant soil texture classes are loamy sand and strong loamy, while the other two soil texture classes are only marginally represented. For all soil classes uniform parameters are used throughout the study site. This fact, combined with the limited effect of topography in the study site, will cause the modeled soil moisture values to be centered around two different values. From the results described in this section, it is clear that the use of soil texture data, combined with a lookup-table, is not sufficient to provide acceptable modeled soil moisture values. Soil Class Slightly Loamy Loamy Sand Strong Loamy Sandy Loam N λ γc Ks RBS μ σ RBS μ σ RBS μ σ Table 12.9: Spatial mean and standard deviation for all parameters throughout the study area. N stands for the number of pixels within the class, RBS for the [ 47 ] values, μ for the average of the retrieved parameter values, and σ for the standard deviation. 18/01/2008 Page 227 of 259

229 Figure 12.43: Comparison of the modeled soil moisture to the observations before and after the calibration procedure for the first three overpasses. 18/01/2008 Page 228 of 259

230 Figure 12.44: Comparison of the modeled soil moisture to the observations before and after the calibration procedure for the second three overpasses. 18/01/2008 Page 229 of 259

231 Figure 12.45: Comparison of the modeled soil moisture to the observations before and after the calibration procedure for the final three overpasses. 18/01/2008 Page 230 of 259

232 The Calibrated Parameters In order to bypass the above described problem, soil parameter values were determined for each pixel, using the algorithm outlined in Section 5.2. Figure shows the spatial distribution of the resulting parameter values and the original soil texture classes. Examining the top and bottom panels of Figure 12.46, one can notice that the values for K s are consistently lower at the locations for which the original soil texture class is strong loamy. The same observation can be made for λ, and to a lesser extent for γ c. Table 12.9 shows the spatial averages of these parameters for the entire study area, the standard deviation, and the values from [ 47 ] originally used in the model application. Since the texture class slightly loamy contains 81% sand, the value for sand from [ 47 ] was used for this class. For the class strong loamy, the value for loam was used. A number of conclusions can be drawn from this table. First, for all soil texture classes except sandy loam (which is only marginally present), the standard deviations of all parameter values are relatively similar. Further, for λ, if the values for λ are sorted in decreasing order, the resulting order of the soil texture classes is similar for the values from [ 47 ] and for the remotely sensed values. However, the remotely sensed values show less variability. These differences in parameter values can be attributed to the scale at which they were obtained: the values in [ 47 ] were obtained at the laboratory scale, while the retrieved parameters in this study are valid for a pixel resolution of 25 m. As stated in the introduction, the scale at which parameter values were obtained can have a strong influence on the parameter values themselves. The same conclusion can be drawn for the K s values, but not for γ c, for which the remotely sensed values for the strong loamy texture class are an exception. Figure shows the relationship between the texture class and the remotely sensed parameter value for each individual field. For all three parameters, the values for the class strong loamy are almost consistently the lowest. For the hydraulic conductivity and the pore size distribution index this is in agreement with the values from [ 47 ], but not for the air entry pressure head. For the other classes the behavior is not as clear, but this can be explained by their textural composition. Table shows that for all classes except strong loamy the textural composition is relatively similar. It can thus be expected that the retrieved parameter values among these three classes will be more similar than the values of the strong loamy class. If the original soil texture classes are lumped into sandy or loamy, a clear distinction between the retrieved values for these classes can be observed. A general conclusion from Table 12.9 and Figure is thus that a relationship exists between the texture class and the remotely sensed soil parameter values. For the hydraulic conductivity and the pore size distribution index this relationship is similar as for laboratory values, but for the air pressure entry head an inverse relationship has been found. 18/01/2008 Page 231 of 259

233 Figure 12.46: Resulting soil parameters over the test site. 18/01/2008 Page 232 of 259

234 Figure 12.47: The soil parameters per field and per original texture class.s 18/01/2008 Page 233 of 259

235 Texture Class % Clay % Silt % Sand Slightly Loamy Loamy Sand Strong Loamy Sandy Loam Table 12.10: The textural composition of the four soil texture classes in the study area. The Calibration Run The calibrated parameters were then used in a final model application. Figures through show the comparison of the modeled soil moisture using the calibrated soil parameters to the remote sensing observations. As can be expected, a strong improvement in the modeled soil moisture values can be observed. Conclusions A methodology has been developed, using remotely sensed soil moisture values and hydrologic modeling, to estimate soil hydraulic parameters in a spatially distributed manner. When the hydrologic model uses soil parameters obtained from in-situ observed soil texture data, a lack of spatial variability in the model results has been obtained. When remotely sensed soil moisture values are used to estimate the three most important soil parameters (the hydraulic conductivity, the bubbling pressure, and the pore size distribution index), a relationship between the original texture classes and the retrieved parameter values can be observed at the level of the study area. This relationship is in agreement with the relationships found in [ 47 ], except for the bubbling pressure. The variation of the parameter values obtained through remote sensing between the texture classes is not as strong as for the values obtained by [ 47 ]. This can be attributed to the scale at which the parameters have been obtained. While the values from [ 47 ] have been obtained at the laboratory scale, the values obtained in this work are obtained at a pixel resolution of 25 m. At the field level, the same relationship between the texture class and the retrieved parameters has been found when the soils are subdivided into classes predominantly consisting of loam or sand. Overall, this research leads to the conclusion that the possibility exists to retrieve soil hydraulic parameters through a combination of remote sensing and hydrologic modeling. Although the parameter values may be different than values found in the literature, this difference can be attributed to the difference in scale at which the parameters are measured. Since hydrologic models generally need to be applied at a spatial scale that is much larger than the laboratory scale, the methodology described in this paper could serve as a basis to derive soil physical parameters at multiple spatial scales. 18/01/2008 Page 234 of 259

236 Large Aperture Scintillometers (LAS) by ITC The observations made by the Large Aperture Scintillometers (LAS) installed in fields 222 and 250 are used to derive turbulent sensible heat fluxes. An LAS provides the possibility to derive turbulent sensible heat fluxes by measuring the structure parameter for the refractive index C N 2 [m -2/3 ] of air, which in turn is derived from the intensity fluctuations of an optical beam between a transmitter and a receiver following [ 63 ]: C N / 3 3 = 1.12 σ D P ln I where the brackets on the left hand side of the equation indicate a spatial average of the measured refractive index. The measured variance of the natural logarithm of intensity fluctuations is represented by σ 2 lni, [-] whereas the D [m] is the aperture diameter of the LAS and P [m] is the distance, or path-length, between the transmitter and receiver of the LAS. Data processing The average value of the structure parameter as obtained form the LAS results from an integration of the structure parameter along the path of the LAS, following: C N 2 1 = 0 C N 2 ( u) W ( u) du where u [-] is the normalized path distance from the transmitter, equal to x/p, with x [m] the actual distance and P [m] the pathlength and W(u) a non-uniform, bell-shaped and symmetrical weighing function, see Figure Terrain elevation [masl] Weighting function Terrain elevation [masl] Weighting function Distance Distance Figure 12.48: LAS line of sight, terrain elevation and weighting function along the transects over winter wheat field 250 (left panel) and corn field 222 (right panel). In the current case the measurements were conducted below the so-called blending height, i.e. a level where the influence of the surface perturbations gradually decays. This invokes that a portion of the surface upstream, called the source area, influences the sensors. Generally this source area is determined using a footprint model which yields a so-called source weight 18/01/2008 Page 235 of 259

237 function, f, that relates the measured flux at height z m to the spatial distribution of surface fluxes. In the case of a LAS one has to combine the source function with the spatial weighting function, W(u), of the LAS in order to estimate the relative contribution of the source area, of which an example is provided in Figure for the 5 th of July at 10:30 utc, where the arrows indicate the actual wind direction. Naturally this is an important issue when using the LAS observations for validating remote sensing-based flux estimates. July 5th, 2006, 10:30 utc E N N E E N N Figure 12.49: LAS footprints for 5th of July 2006, where a more intense red color indicates higher surface contribution to the measured flux.. In the optical domain, the structure parameter for temperature, C T 2 [K 2 m -2/3 ], can be derived from C N 2 as measured by a scintillometer following [ 65 ]: E 1 Görmin weather station 2 BREB station 3 LAS Receiver- field LAS Transmitter - field LAS Receiver - field LAS Transmitter - field 222 C T γ P a = C 1 T N + β 2 in which T a represents air temperature [K], γ is a refractive index for air ( K Pa -1 ), P [Pa] indicates atmospheric pressure and ß [-] is the Bowen ratio, here used as a correction term for humidity related scintillations. Similarity relationships based on Monin-Obukhov Similarity Theory, provide the possibility to derive sensible heat flux, H [W/m 2 ], through the use of the temperature scale, T * [K], following: C T 2 = T * 2 2 / 3 0 ( z d ) f 0 T z d L 18/01/2008 Page 236 of 259