Physics of Remote Sensing Gastellu-Etchegorry Jean-Philippe CESBIO Paul Sabatier University, CNES, CNRS, IRD, France 4:21 PM Outline • Major remote sensing configurations and remote sensing (RS) images. • Sun radiation reflexion: - Radiometric variables & Components of TOA radiance - Hot spot, penumbra, 2D display of reflectance, albedo • Thermal emission: Radiometric quantities and Components of TOA radiance • Interpretation of VIS/IR/TIR satellite images: angular anisotropy, spectral wealth,… • Hyperspectral remote sensing: mixels, end-members,… • Satellite / airborne sensors with finite FOV • LiDAR 4:21 PM Very often, analysis of RS signal without any physics understading is dangerous… Approach: - To introduce physics with examples ⇒ simple questions not always so simple… ⇒ thank you for participating - To illustrate RS physics with RS model (DART) (www.cesbio.ups-tlse.fr/dart) ⇒ DART input parameters represent major RS atmosphere / Earth surface parameters Spectrum, atmosphere transmittance and Sun / Earth thermal emission 4:21 PM W/m2/µm TOA irradiance (Kurucz) 2 W/m /µm 2,000 30 Blackbody: 5900K and 6000K (at sun distance) 25 Blackbody 1,500 20 BOA irradiance: H2O (Atmosphere US Standard 1,000 15 H2O - No aerosols) H2O 10 500 Blackbody: 300K H2O O3 5 H2O ) H2O, CO2 (at sun distance O2 0 0 0.200 0.700 1.200 1.700 2.200 2.700 µm 3 13 23 µm Transmittance Total H2O CO2 O3 Modtran - US Standard model 0.4 0.5 0.6 0.7 µm µm Visible X-Rays Ultraviolet Near IR Thermal IR Far IR Microwave Radio 0.01µm 0.1µm 1µm 10µm 100µm 1mm 10mm 100mm 1m Wavelength 1016 1015 1014 1013 1012 1011 1010 109 Frequency (Hz) Remote sensing configurations in the optical domain (λ ∈ [0.25µm 100µm]) 4:21 PM Lidar (WF, PC) Radiometer Lλ(Ωs,Ωv) Ω Ω Ω ρλ(Ωs,Ωv) TB,λ( v) wλ( s, v, t) 3D radiative Any Earth budget landscape Orbits of Earth observation satellites 4:21 PM Equatorial orbit Polar orbit +3h, 2 orbits +6h, 4 orbits +9h, 6 orbits Earth rotation Sun Sun Sun Sun Nearly polar sun-synchronous orbit: every 3 hours: MODIS, Landsat (500km-1500km) Non sun-synchronous orbit: IceSat (LiDAR) Geostationary orbit: Meteosat, etc. (36 000km) Remote sensing raster image 4:21 PM Columns (j) Lines or 11 15 65 71 94 1 rows (i) 111 90 33 21 17 2 83 6 22 9 11 Bands (k) Grey/color 17 27 33 57 48 3 tone 4 X axis DN Multi-spectral image. Images (i.e., DNs) are displayed DNk(i,j) is the digital number of pixel Pk(i,j) with LUTs: grey/color tone = f(DN) High res commercial data galleries - http://www.geoeye.com/CorpSite/gallery/Default.aspx - http://www.digitalglobe.com/ - http://www.digitalglobe.com/index.php/27/Sample+Imagery+Gallery Signatures of Earth surfaces vs. Resolutions in remote sensing images 4:21 PM Along track scanner: pushbroom Cross track scanner: whiskbroom IFOV for each detector: 1mrad Scan rate: 2 10-2 s per scan line IFOV: 1mrad Altitude 90° Altitude 10km 10km FOV X axis Ground resolution Ground resolution cell 10m x 10m cell 10m x 10m Sabin, 1997 Dwell time ⇒ Signal to noise ratio (SNR) Scan rate per line 2 10 Cell dimension 10 m per cell = = 10 s/cell = = 10 2 s per cell ∆t = Number cells per line 2000 cells ∆t = Velocity 200 m / s −2 −5 5 − Spatial resolution: GIFOV Pixel dimension 4:21 PM 10km 1km 1km Satellite track 10km 1km 1km Satellite track Image: 10pixel x 10pixel - 10km x 10km What is the spatial resolution (GIFOV) of the sensor? Remote sensing (RS) of Earth surfaces: sensors that measure radiation • Passive sensors (multi/hyper spectral radiometers): Variety of satellite - sun radiation reflected by Earth surfaces / airborne sensors - radiation that is thermally emitted by Earth surfaces • Active sensors: LiDARs and Radars (not considered here) - Vegetation: biochemistry (Chl, H2O,...), spatial extent,... Very diverse - Soil: humidity, temperature, chemistry, altitude,... information - Water: temperature, salinity, roughness,... Many applications: Biosphere functioning, Cartography & Topography, Agriculture (irrigation, crop forecast,…),Oceanography , Forestry , Environment (fires, floods,…),.. Sea level trend: Disks: 3.14 mm/year Seasonal evolution of vegetation DEM: Digital Elevation Model crops + circular irrigation Signatures of Earth surfaces vs. Resolutions in remote sensing images 4:21 PM Information about Earth surfaces to derive from remote sensing signals: • Spectral signature: signal variation with wavelength ⇔ spectrometers • Angular signature: signal variation with view (sun) direction ⇔ multi-view sensor,… • Spatial signature: signal variation with space ⇔ high / mid spatial resolution • Temporal signature: signal variation with time ⇔ sensor acquisition repetitivity • Architecture / distance signature: backscattering of LiDAR pulse • Fluorescence emission (specific spectral bands), polarization, etc. Several resolutions: - Radiometric resolution - Spatial resolution GIFOV (IFOV + altitude) ≠ pixel dimension - Spectral resolution - Temporal resolution Importance of sensor characteristics and configuration AVIRIS images: forest clearing fire (Cuiaba, Brazil, 25/08/1995) - Radiometric resolution - Sun / Sensor configuration - Spatial resolution, - Spectral domain/resolution, - Time resolution. 2 bit image: 8 bit image: 4 levels / grey tones 256 levels / grey tones Image5 Image4 Image3 Image2 Image1 End Begin imaging • Fire signal low at 1µm and largeimaging at 2.5µm. Why? Line-of-•sight Earth surfaces: very low signal at 2.5µm. Why? of imager Spectrometer 15km CHRIS / Proba Perception vs. spectral domain: GOES 12 Sounder ⇒ Brightness temperature at several altitudes • Signal is null in the visible band (0.65µm). Why? • In thermal infrared (TIR) bands: - The low (-50C) and large (20C) signals (brightness temperature TB) correspond to? - The atmosphere thermal emission changes with spectral band (i.e., lower/higher TB). Why? GOES-12 sounder 7:46 UTC, 29/04/2004 (Menzel, NOAA/NESDIS/ORA) 4:21 PM Perception vs. spectral domain and passive vs. active sensor: G-LIHT images (NASA) Airborne G-LIHT sensor: RGB color Brightness DSM ⇒ DEM, composite temperature tree height,… 3D vegetation structure Imaging Thermal Scanning spectro. camera Lidar LiDAR apparent reflectance • Why brightness temperature is larger for grass than for trees? • How 3D vegetation is obtained? • What is LiDAR apparent ρ? • What is the usefulness of the irradiance spectrometer? 4:21 PM Radiometry: definitions - Solid angle dΩ of direction Ω(θ,φ): dΩ = sinθ.cosφ.dθ.dφ with θ = zenith and φ = azimuth 2 2 2 - Radiance LΣ(Ω) of Σ along (Ω): LΣ,λ(Ω) W/m /sr/µm (/ effective m !), LΣ,∆λ(Ω): W/m /sr 2 2 EΣ,λ = ⌠LΣ,λ(Ω).|cosθ|.dΩ W/m /µm EΣ,∆λ = ⌠LΣ,∆λ(Ω).|cosθ|.dΩ: W/m - Irradiance EΣ of Σ (in flux): ⌡ ⌡ 2π- 2π- 2 2 - Exitance MΣ of Σ (out flux): MΣ,λ = ⌡⌠LΣ,λ(Ω).cosθ.dΩ W/m /µm MΣ,∆λ = ⌡⌠LΣ,∆λ(Ω).cosθ.dΩ W/m 2π+ 2π+ π. LΣ,λ(Ω) - Reflectance factor ρΣ(Ω) of Σ along (Ω): ρ Σ,λ ( Ω ) = if Σ is lambertian: ρΣ,λ(Ω) = cst EΣ,λ z r.s i n θ.d φ Ω(θ, φ) r. d θ d A dΩ - Solid angle dΩ θ LΣ(Ω): Ω ∈ 2π L (Ω): Ω ∈ 2π+ Σ L (Ω) along direction EΣ Σ r Ω(θ,φ) from Σ y Σ Σ Σ Σ φ Exitance MΣ Irradiance EΣ Reflectance ρΣ(Ω) x What is the physical quantity that is measured by a satellite radiometer? 4:21 PM Different reflectance factors (direct d, hemispherical h, conical c) - In or Out 'direct' flux (index d): flux along a unique direction (i.e., dΩ=0). - In or Out 'conical' flux (index c): flux within a cone (Ω, dΩ≠0). - In or Out 'hemispherical' flux (index h): flux within an hemisphere (i.e., dΩ=2π). Which reflectance is derived from the measurement of a satellite radiometer? Which reflectance is used for computing the albedo of our planet? Satellite image acquisition and display 4:21 PM Oz Landscape: grass, house θ Atmosphere Oy House Sun (upward): θs=30°, φs=330° Satellite: - nadir view direction Ox φ Grass - 3 bands: B, G, R Blue Green Red Sun at nadir: what are the zenith and azimuth angles? RGB Color Composite Satellite image acquisition and display 4:21 PM Landscape: grass, house Atmosphere House Sun: θs=30°, φs=330° Satellite: - nadir view direction Grass - 3 bands: G, R, NIR Near infrared Green Red False Color Composite Spatial resolution and radiometry 4:21 PM 0.18 0.5m resolution 1m resolution 2m resolution Red spectral band 0.15 Spatial sampling ⇒ Change of DNs 5m resolution 10m resolution TOA (Top of Atmosphere) and BOA (Bottom of Atmosphere) acquisitions 4:21 PM Nadir view direction (θv=22°, φv=30°) view direction (θv=22°, φv=330°) view direction y BOA Proportion of observed shadows x can vary a lot with view direction ⇒ Anisotropy of RS signal TOA Why TOA images are bluish? Atmosphere characteristics (atmosphere model), sun direction, etc. 4:21 PM Satellite signal (TOA radiance) LTOA,λ(Ω) - short wavelengths LTOA,λ(Ω) Lmult Latm LEarth TOA: Top of the Atmosphere Atmosphere BOA: Bottom of the Atmosphere LTOA,λ(Ω) = "Radiance LEarth,TOA,λ(Ω) due to scattering of sun flux by the Earth, only" + "Radiance Latm,TOA,λ(Ω) due to scattering of sun flux by the Atmosphere, only + "Radiance Lmult,TOA,λ(Ω) due to scattering of sun flux by {Earth + Atmosphere}" The hot spot (backscattering effect): examples 4:21 PM Hot spot: sensor view direction = sun direction ⇒ maximal reflectance MODIS. 10/27/2002 http://academic.emporia.edu/abe http://earthobservatory.nasa.g rjame/remote/lec10/hotspot.jpg ov/IOTD/view.php?id=83048 From Donald Deering 4:21 PM Reflectance of lambertian and natural surfaces How can there be hot spot? Indeed, Sun direction ρ a sensor with the sun behind it Ωs(θs, φs) dd ρdd(Ωs,Ωv) θv π−θ should see its own shadow… s ρ=1: white Ω Σ lambertian ρ=0.5: grey Lambertian surface: ρdd = ρdh = ρhd = ρhh = cst.