Multi-sensor data base over desert sites for calibration purpose. P. Henry ¹, X. Briottet ², C. Miesch ², F. Cabot ¹ ¹CNES, ²ONERA

Transcription

1 Multi-sensor data base over desert sites for calibration purpose P. Henry ¹, X. Briottet ², C. Miesch ², F. Cabot ¹ ¹CNES, ²ONERA

2 Outline Introduction SADE database Calibration method Some results Desert characterization Conclusion 2/25 CEOS-IVOS Workshop October 2004

3 Introduction (i) Initial objectives : In orbit vicarious calibration to assess : Multiangular calibration (detectors normalization in the f.o.v.) In time calibration monitoring (on-board calibration verification) Intercalibration of sensors Main characteristics of the requested sites : Stable in time : no vegetation Easy to access : low cloud coverage, good atmospheric conditions High reflectance : to reduce the impact of atmospheric effects Low directional effects Choice : Desert sites 3/25 CEOS-IVOS Workshop October 2004

4 Introduction (ii) Sites selection : Spatial uniformity Better than 2% for 100x100 km 2 area Statistics using Meteosat acquisitions Stability over time (seasonal effect) Stability better than 20% (after atmospheric effect filtering) 1 year of Meteosat data (1 per day) Low directional effect Directional effects less than 15% 1 month of AVHRR data completed with Meteosat data Sites characterization : Ground truth measurements in Algeria (1993) 4/25 CEOS-IVOS Workshop October 2004

5 Introduction (iii) 20 sites selected over North Africa and Arabia 5/25 CEOS-IVOS Workshop October 2004

6 SADE Data Base (i) Systematic collect of satellite acquisitions over the 20 sites : POLDER 1 (oct june 1997) POLDER 2 (2003) Since 1990 : some SPOT high resolution Since 1998 : SeaWiFS, VEGETATION 1 & 2, AVHRR 14 & 16 Since 2001 : MERIS, MODIS + MISR, ATSR2, AATSR + Meteo data Storage in a data base : SADE data base : Structure d Accueil de Données d Etalonnage Easy data management Link between satellite measurements and calibration results (traceability) Nota : the SADE data base also includes calibration measurements over ocean, sun glint, clouds and snowy sites. 6/25 CEOS-IVOS Workshop October 2004

7 SADE Data Base (ii) ADEOS2/POLDER AQUA/MODIS ENVISAT/AATSR SPOT5/HMA1 SPOT5/HRS2 SPOT5/HRS1 SPOT5/HRG1 SPOT5/HRG2 SPOT5/VEGETATION ENVISAT/MERIS NOAA16/AVHRR3 TERRA/MODIS TERRA/MISR SPOT4/VEGETATION SPOT4/HRVIR1 SPOT2/HRV2 SeaStar/SeaWiFS SPOT1/HRV1 ERS2/GOME ERS2/ATSR2 ADEOS/POLDER NOAA14/AVHRR 01/01/96 29/06/96 26/12/96 24/06/97 21/12/97 19/06/98 16/12/98 14/06/99 11/12/99 08/06/00 05/12/00 03/06/01 30/11/01 29/05/02 25/11/02 24/05/03 More than multispectral acquisitions available in SADE 7/25 CEOS-IVOS Workshop October 2004

8 SADE Data Base (iii) 8/25 CEOS-IVOS Workshop October 2004 Structure of the SADE data base

9 SADE Data Base (iv) 9/25 CEOS-IVOS Workshop October 2004 ToA reflectance comparison

10 SADE Data Base (v) 10/25 CEOS-IVOS Workshop October 2004 ToA reflectance comparison

11 Calibration method (i) Compare two sensors : One sensor as reference Comparison at TOA level Needs accounting for: Directional effects Atmospheric conditions Spectral discrepancies Reference Sensor Atmospheric correction to surface reflectance Spectral resampling Surface reflectance for sensor 2 Sensor 2 Comparison Atmospheric simulation to ToA reflectance 11/25 CEOS-IVOS Workshop October 2004

12 Calibration method (ii) Directional effects : Direct comparison of measurements in the same geometry Use of reciprocity principle to extend field of matching geometries MERIS POLDER Available geometries for Libya 1 VGT 12/25 CEOS-IVOS Workshop October 2004 SeaWiFS

13 Calibration method (iii) Atmospheric correction : Atmospheric correction performed using SMAC and meteo data : Rayleigh scattering correction Water vapour Ozone Other contributors : CO2, CO, NO2, CH4 (constant values) Problem : aerosol correction Desert model Aerosol optical thickness Tau =0.2 Comparison performed at ToA level Reference sensor ToA data are corrected to obtain ground reflectances ToA Sensor 2 reflectances are simulated using these ground reflectances and meteo data at the date of Sensor 2 acquisition 13/25 CEOS-IVOS Workshop October 2004

14 Calibration method (iv) Spectral resampling : Performed after atmospheric correction Reference sensor measurements used to fit an empirical spectral model (4 parameters) : Ro = A arctan[ a ( λ - λ 0 ) ] + B Arctan model integrated in the Sensor 2 spectral bands 14/25 CEOS-IVOS Workshop October 2004

15 Some results (i) 443P POLDER2 vs POLDER1 443NP /25 CEOS-IVOS Workshop October 2004

16 Some results (ii) POLDER2 vs POLDER /25 CEOS-IVOS Workshop October 2004

17 Some results (iii) 443P POLDER2 vs MODIS 443NP /25 CEOS-IVOS Workshop October 2004

18 Some results (iv) VGT 1 vs VGT 2 18/25 CEOS-IVOS Workshop October 2004

19 Some results (v) SeaWiFS vs POLDER 19/25 CEOS-IVOS Workshop October 2004

20 Desert characterization (i) Objective : Use of SADE data to characterize the desert sites behavior First step : spectral characterization Second step : directional characterization Finally : spectral and directional model How : Global assimilation method Adapted inversion algorithm Initial values of calibration coefficients Top of atmosphere reflectance over a site Ground reflectance of reference site Satellite measurements extracted from SADE Computation of a new set of coefficients Algorithm Outputs: Site characterization Calibration coefficients 20/25 CEOS-IVOS Workshop October 2004

21 Desert characterization (ii) Problem : Characterization of the desert reflectance using a large number of satellite data coming from different instruments (and spectral bands) Method : Initial assumptions : Atmospheric parameters are known (SADE) Sensors are inter-calibrated Determination/choice of a spectral or directional model Development of an adjustment algorithm able to deal with data of various origins (reflectance integrated in the spectral bands) 21/25 CEOS-IVOS Workshop October 2004

22 Desert characterisation (iii) First step : spectral characterization Determination of an ad-hoc spectral model of sandy desert from desert samples characterized with ONERA devices => use of a simple «arctan» model (4 parameters) 22/25 CEOS-IVOS Workshop October 2004 Typical spectral behaviour of desert sand

23 Desert characterisation (iv) Fit of the model with a satellite data set over Libya 1: => use of POLDER, VGT, SPOT and AVHRR data Adjustment: the simulated TOA data using the adjusted ground model are compared to original data Comparison with GOME: the adjusted model is transferred at TOA and is compared to a GOME acquisition 23/25 CEOS-IVOS Workshop October 2004

24 Desert characterisation (v) Second step : directional characterization (study in progress) Use of a Rahman directional model Fit of the model using satellite data (at ToA level) First results are encouraging : Comparison of directional model obtained over Libya 1 using separately POLDER 1 and POLDER 2 data (principal plan for θ s =40 ) 24/25 CEOS-IVOS Workshop October 2004

25 Conclusion and perspectives Site characterization improvement Finalize current studies on desert directional characterization Establish spectral & directional model for each site Atmospheric correction improvement Set up a CIMEL photometer on one site to improve the aerosol correction (probably in Libya) Extension of the data base Continue the collect of satellite data over the sites (long term series + new sensors) Open the data base to other users 25/25 CEOS-IVOS Workshop October 2004