INTRO TO REMOTE SENSING

Transcription

1 The remote sensing revolution: By 2020, the Earth will be imaged from space at a spatial resolution of at least 1 km x 1km every minute. INTRO TO REMOTE SENSING Hadi (hadi.hadi@aalto.fi) From measurements to maps (GIS-E1020) 20 Sept 2017

2 Hi all! I am sorry that I cannot be there today I had to go to Brussels for two days. Hadi is a very gifted researcher in my team and will give the lecture on my behalf! Tekniikka & Talous, Miina For Finnish speakers:

3 A LITTLE BIT ABOUT ME #yssp17 Recent activities: Final year (wish me luck!) phd candidate ( ) supervised by Miina Phd thesis: Satellite remote sensing of forest biophysical variables in boreal and tropical biomes Focus on optical RS data interpretation Methods: statistical learning, physicallybased simulation, long time series Tools: #rstat, SeNtinel Application Platform (SNAP), Google Earth Engine (#FutureEO), ArcGIS #ESALTC17

4 AFTER TODAYS LECTURE + ASSIGNMENT + READING The big picture of remote sensing is clear to you You can list applications of different types of RS data You can explain how RS data are characterized You are familiar with the basic processing steps of RS data (+ You know how continue your RS studies / where to look for further information)

5 THE BIG PICTURE? Discussions in small groups (10-15 min): What is remote sensing? Where does the history of remote sensing start? When was the first satellite launched by man? How many countries operate remote sensing satellites? Where can you get remote sensing data sets?

6 REMOTE SENSING? United Nations (UN) definition: the sensing of the Earth s surface from space by making use of the properties of electromagnetic waves emitted, reflected or diffracted by the sensed objects, for the purpose of improving natural resources management, land use and the protection of the environment Do you agree with this definition? Why or why not? Earth observation vs. remote sensing how do the concepts differ? + How do you separate remote sensing from photogrammetry?

7 A REMOTE SENSING SYSTEM 1. Platform 2. Instrument(s) + Human, UAV, airplane, Passive, Active satellite,.. v Imaging, Non-imaging Panchromatic, Multispectral, Hyperspectral Lukes Spatial resolution Temporal resolution Viewing geometry Swath width Spectral resolution Radiometric resolution Specim ESA Microsoft

8 INSTRUMENT VS. PLATFORM Instrument (Moderate Resolution Imaging Spectrometer) Platform (Terra) Satellite (platform) A body in orbit around another body Can carry several instruments < 10 % of satellites have sensors for mapping of natural resources Instrument (Advanced Spaceborne Thermal Emission and Reflection Radiometer) Instrument (sensor) Measures EM radiation in a given direction and transforms it to a digital signal

9 DISTANCES TO EARTH? 11 km passenger jet planes km Geostationary satellites 215 km Sputnik km International Space Station 595 km Hubble Space Telescope km Sun-synchronous satellites LEO MEO HEO km km km km GPS satellites image: BlueMarble

10 SATELLITE ORBITS Images: ESA Sun synchronous orbits (near)polar orbits e.g. mapping natural resources (same local time of the day) Geostationary orbit e.g. telecommunication, navigation (stationary with regards to Earth s surface) Something extra? How Do Satellites Get & Stay in Orbit? For more detailed info: Satellite orbits (20-min lecture / University of Edinburgh):

11 SATELLITE CONSTELLATIONS A group of satellites with coordinated ground coverage Usually also shared control Synchronized so that they overlap well in coverage (and complement each other) possibilities to combine data in e.g. catastrophe monitoring GPS constellation 24 satellites, distributed equally NASA A-Train Currently 5 satellites All cross the equator approx. 13:30

12

13 @slide from talk by Zoltan Bartalis of ESA ESRIN, Italy

14 Sentinel-2 ENDLESS USES OF RS IN MAPPING THE EARTH + Cost-effective + Unbiased across political borders + Only technique for repetitive measurements of large areas - Challenges in developing interpretation algorithms - Often limited by weather (e.g. clouds, recent rain,..) - High initial costs - Spatial resolution(s) not optimal for all applications Landsat TM, ETM; OLI Jensen, 2014

15 CHARACTERIZING RS DATA 1) By electromagnetic domain (optical, thermal, microwave) 2) By active vs. passive techniques GrindGIS 3) By resolution (spatial, spectral, temporal, rediometric) 4) By platform (satellite, airborne,..)

16 UTILIZING THE ELECTROMAGNETIC SPECTRUM Different mapping applications in optical, thermal and microwave domains Image RSCAL ~ µm visible (optical) ~0.7 3 µm reflected infrared (optical) ~ µm near infrared ~1.4 3 µm shortwave infrared ~3 100 µm emitted infrared (thermal) ~1 mm 1 m microwave (radar)

17 INTERACTION WITH EARTH SURFACE MATERIALS Passive, optical RS: EM reaching the Earth s surface from the Sun is reflected, transmitted or absorbed Basic assumption of RS: specific targets have a characteristic manner of interacting with incident radiation Spectral response depends on: Chemical and physical properties of target Solar azimuth & elevation angles Sensor angle ( look angle ) Status of atmosphere Spectral signature = the spectral response curve of a target Image RSCAL

18 Active, microwave Pottier Also, sensitive to moisture (can be noise) i.e. dielectric properties. Specific wavelength and polarizations important.

19 INTERPRETATION OF SATELLITE DATA Visual interpretation ( a pretty picture ) Checking out your hotel surroundings before a holiday Planning field work in unknown areas Different computational methods for discrete vs. continuous variables Prerequisite: there is a physical relationship between the radiation scattered/emitted by the target and the characteristics we want to retrieve Most common starting point: thematic land cover maps A few examples to showcase different uses of RS data

20 CATCHING TAX EVADERS Finding Swimming Pools with Google Earth: Greek Government Hauls in Billions in Back Taxes (SPIEGEL) the suburbs didn't have 324 swimming pools, as was reported, but rather 16,974. Greece is looking at increasing their tax revenues using remote sensing applications

21 FIGHTING CROP INSURANCE FRAUDS USA: Landsat images used to look for evidence of whether or not a farmer actually planted or harvested 50% of Landsat image analyses support a farmer s insurance claim and 50% indicate fraud. Read more:

22 DETECTING OIL SPILLS FOR MARINE LIFE PRESERVATION ol Maps of oil spills: To delineate the extent and the thickest parts of the oil slicks. Use of optical satellite data: oil changes how the water surface reflects light, either by making the sun s reflection brighter or by dampening the scattering of sunlight (oily area darker) NASA

23 SATELLITE IMAGERY IN MINING ASTER (15 m) satellite image shortwave infrared bands (SWIR) show (in bright pink) rocks associated with copper mineralization Morenci Mine Mineral Mapping - Arizona, USA

24 PREDICTING VOLCANIC ERUPTIONS WITH SATELLITE RADAR DATA Using satellite radar to detect swelling magma near the summit of volcanoes. Synthetic Aperture Radar Interferometry (InSAR) by the Japanese Space Exploration Agency's ALOS satellite Estelle Chaussard

25 MAPPING ANTARCTICA, GREENLAND, AND OTHER INACCESSIBLE AREAS NASA's Ice, Cloud and Land Elevation Satellite (ICESat) ICESat uses a 1064 nm laser to measure at 172-m intervals over ice, oceans, and land Are the Greenland and Antartic ice sheets growing or shrinking? Seasonal dynamics? #LarsenC Video:

26 MAPPING URBAN HEAT ISLANDS True color Landsat Thermal band Landsat Cleantechnica An urban heat island (UHI) is a city or metropolitan area that is significantly warmer than its surrounding rural areas due to human activities. NASA

27 See the time series video (like a time machine ) here: /file/d/0b1htxodgc9wbz0 EwcmNtUndLT3M/view?u sp=sharing Deforestation in Kalimantan, Indonesia ( ) as recorded by long-term continuity of optical Landsat satellites

28 MONITORING SPRING AND CLIMATE CHANGE Spectral vegetation indices to detect automatically onset and offset of growing season. Coarse spatial resolution satellite data (MODIS), emphasis on red and NIR spectral bands (e.g. time series of NDVI). Images Lars Eklundh

29 GLOBAL TOPOGRAPHIC MAPS Los Angeles, SRTM + Landsat data NASA Shuttle Radar Topography Mission (SRTM) collected data on the shape and features of the surface of the Earth.

30 DIGITAL ELEVATION MODELS FROM OPTICAL SATELLITE IMAGERY Example: Commercial SPOT 6/7 and Pleaides in the same constellation can revisit the same target Frequently rapid response tasks 5 m Digital Surface Model (DSM) of Mexico City extracted from 1.5m SPOT 6 stereo imagery A DEM of Rapa, French Polynesia from IKONOS images

31 MONITORING GLOBAL GREEN BIOMASS Valerio Avitabile Decision support information: Crop biomass and food security? Droughts is major agricultural areas effects on annual crop prices? Which forests to preserve, which to log?

32 GLOBAL AIR POLLUTION MAPS Global mean tropospheric nitrogen dioxide (NO 2 ) vertical column density (VCD), as measured by the SCIAMACHY instrument on ESA's Envisat SCIAMACHY: SCanning Imaging Absorption SpectroMeter for Atmospheric CHartographY

33 @Gordon Campbell, Anna Burzykowska

34 RESOLUTIONS OF RS DATA Resolutions: spatial, spectral, temporal, radiometric Spatial resolution: The smallest angular or linear separation between two objects that can be resolved by the sensor. GSD ground sampling distance FOV field of view swath width IFOV instantaneous FOV GIFOV ground projected IFOV Note! GSD ( pixel size ) GIFOV Radiometric resolution: The sensitivity of the sensor to incoming radiance: How much change in radiance is needed before a change in recorded brightness value occurs?

35 HCRF SPECTRAL RESOLUTION 0,6 0,5 B4 B4 NEMO 0,4 0,3 B5 0,2 0,1 B1 B2 B3 B Wavelength [nm] Hyperspectral - Dozens / hundreds of bands - high spectral resolution - (Near)continuous cover Multispectral - Several bands - medium spectral resolution - Non-continuous cover of sampled interval Panchromatic - Single band - low spectral resolution - Non-continuous cover of sampled interval

36 TEMPORAL RESOLUTION How often the RS system records imagery of a particular area ( revisit time ). Latitude influences the revisit time. Pointability improved revisit time by altering satellite viewing geometry ( pointing on target ) High temporal resolution: < 24 hours 3 days Medium temporal resolution: 4 16 days Low temporal resolution: > 16 days Satellite imaging corporation

37 PRE-PROCESSING OF RS DATA Raw data As distributed by provider 1. Geometric corrections Correction for systematic errors in image geometry caused by platform and rectification of image in selected coordinate system 2. Radiometric corrections From DN values to callibrated radiance values. Noise removal and correction for different sun and view geometries. 3. Atmospheric corrections Removal of confounding effects of atmosphere on the image 4. Image enhacement and interpretation New product / information derived from original data MAP! You will go through this entire chain in the next course (GIS-E1040).

38 COMMONLY USED DATA PROCESSING LEVELS Data are increasingly provided as preprocessed products for end-users. Level 0 Level 1 Level 2 Level 3 Most commonly used product is surface reflectance factor. Raw data from sensor with correction for systematic errors Level 1a radiometrically corrected, no geometrical corrections Level 1 G radiometrical and basic geometrical corrections Derived geophysical variables at same resolution and location as the Level 1 data Temporal series of level 2 products Level 1 T radiometrical and terrain corrections using DEM

39 IMAGE ENHANCEMENT AND INTERPRETATION Colour composites Histogram stretching Image enhancement Spatial filtering Spectral indices Image classification Physical models Regression models Image interpretation

40 IMAGE CLASSIFICATION Waser et al., 2011, Semi-automatic classification of tree species in different forest ecosystems by spectral and geometric variables derived from Airborne Digital Sensor (ADS40) and RC30 data, Remote Sensing of Environment Each pixel is assigned to one of the selected class (e.g. land cover type) Supervised classification Unsupervised classification

41 Band 2 UNSUPERVISED CLASSIFICATION Pixels are assigned to class automatically based on their spectral behavior D ab = n i=1 (a i b i ) 2 ISODATA kmeans Band 1

42 SUPERVISED CLASSIFICATION Training pixels are used to obtain signature for each land class Polygon areas for each land class, homogeneous and evenly distributed across image Examples of methods: Minimum distance to the mean Maximum likelihood (mean and variance) probabilities Considered a soft classification method

43 OBJECT-ORIENTED CLASSIFICATION Based on objects rather than on individual pixels High spatial resolution data required

44 ESTIMATING CONTINUOUS VARIABLES Spectral indices (i.e. condensing the info content of several bands into one index) Mathematical manipulation of different spectral bands +, -, x, /, goniometric functions etc. Applied on each pixel to create a new image layer Most commonly used to separate vegetation from other land covers: NDVI (normalized difference vegetation index) Physically-based (radiative transfer) models Especially to model the atmosphere and vegetation layers Look-up tables, merit functions,.. Regression techniques, neural networks, Hands-on experience in estimating continuous variables from satellite data in the next courses (GIS-E1040, GIS-E3050).

45 NORMALIZED DIFFERENCE VEGETATION INDEX NDVI = NIR RED NIR + RED <-1;1> < 0 no vegetation > 0 increasing greenness of vegetation

46 WHAT S HOT IN REMOTE SENSING Applications: climate change monitoring, public security & hazard monitoring, mapping natural resources and land use changes NDVI(780,688) Techniques: hyperspectral, hypertemporal, thermal Data fusion: active & passive data sets Open software movement Ready end-user products Hourly data from DSCOVR EPIC Yuri Knyazikhin

47 A gamechanger optical satellite: unprecedented combination of spatial, spectral, temporal resolutions in its class! Planned date 13 Oct 2017

48 HOW TO KEEP UPDATED? Remote sensing is a very rapidly developing field. Datasets, software, web portals, will be different in 5 years! Most important is to understand the physical basics, and to apply them to the constantly new types of datasets. For general info, follow e.g. NASA and ESA websites and If interested in scientific results, read online the most highly ranked journal Remote Sensing of Environment (Elsevier): And open source Remote Sensing (MDPI):

49 RS STUDIES IN AALTO In the GIS MSc Program Photogrammetry, laser scanning and remote sensing, II period Advanced remote sensing, V period For all AALTO MSc students, a minor in Earth Observation (25 cr) Some good online resources for independent study: European Space Agency s LearnEO! Natural Resources Canada s Fundamentals of remote sensing ESA online light course: Monitoring climate from space

50 DURING THIS COURSE Hands-on assignment: Interpret evergreen areas in Otaniemi from orthophotos using simple classification Instructions in MyCourses. DL for reports: 10 Oct 2017 More info from Petri Rönnholm Read more: Chapter 1 (pp. 1-27): Remote Sensing: Basic Principles. In Computer Processing of Remotely-Sensed Images : An Introduction (4 th Edition). By Mather & Koch. Available as en e-book in Aalto Library: Focus on pp (subsections ) This reading is especially for those who feel that they need to refresh their knowledge of electromagnetic radiation and its properties.

51 EXTRA SLIDES

52 HOW DID WE GET HERE? 1858: first aerial photo taken from a balloon 1868: electromagnetic theory published : aerial photos from airplanes 1930s: radar developed 1957: Sputnik launched 1959: first photo of the Earth from space (Explorer-6) 1960s: satellite images of daily cloud cover (Tiros-1) 1966: first geostationary weather satellite (ATS) 1972: first satellite sensor for monitoring natural resources (Landsat) 1970s: digital image processing develops 1980s: processing of satellite images done on desktop computers 1999: first satellite sensor (Ikonos) with high spatial resolution (in civil use) 2000: first hyperspectral satellite sensor (Hyperion) 2000s: global monitoring becomes operational, lidar techniques develop 2005: Google Earth brings satellite images to everybody 2008: All NASA remote sensing data sets become free and available to everybody

53

54