MSI aerosol retrieval algorithm for the Multi- Spectral Imager (MSI) on EarthCare

Transcription

1 MSI aerosol retrieval algorithm for the Multi- Spectral Imager (MSI) on EarthCare Wolfgang von Hoyningen-Huene Huene,, Alexander Kokhanovsky, Vladimir Rozanov,, John P. Burrows,, Gerard Hesselmans 2), Leslie Gale 2), Gerrit de Leeuw 3) Institute of Environmental Physics,, University of Bremen, 2) BMT ARGOSS, 3) FMI Outline:: Desription of task of project MSI conditions Main steps of the retrieval approach Examples, Validation Errors Summary

2 Description of task Tasks according SOW of : = Aerosol Modelling and Retrieval from Multi-Spectral Imagers SOW 3.1: SOW 3.10 ESA AO project development of a prototype algorithm for an AOT product for the future MSI instrument for clean ocean regions give an error characterization of the AOT product containing flagging for clouds land sun glint program for prototype algorithm in Fortran / C, operated under LINUX with description D6 SOW 3.7 validation of the algorithm D7, D5 SOW 3.8 generation of an ATBD D5 for the AOT product

3 MSI Conditions Conditions: MSI single view instrument Swath width 150 km near nadir case 4 spectral channels in red and NIR no information in short wave range MSI MODIS µm m (0.650 µm) µm m (0.860 µm) µm m (1.640 µm) µm m (2.140 µm) solar zenith range 0 70?(80?) deg at low latitudes contamination by sun glint

4 Radiance Requiements for MSI Simulated and measured TOA reflectance for MSI channels Clean aerosol conditions are close to the minimum of full MSI signal TOA Radiance W/(m² sr µm) Channel SZA VZA Remark AOT RTM (small SZA) RTM dl/daot observation MERIS Pacific near Midway Islands (Retrieval 0.66: AOT=0.06) observation AATSR Pacific near Midway Islands RTM (large SZA) RTM MSI full signal range (Min value)

5 Main Steps of the Retrieval Approach Physical background of the AOT retrieval over ocean Radiative transfer equation description of its contributions and its realization for the programming determination of aerosol reflectance LUT for the transfer of aerosol reflectance to AOT Case discrimination, flagging Sun glint Validation Errors

6 Determination of Aerosol Reflectance General Approach TOA-Reflectance, Kaufman et al TOA ρ ( z, z, φ) = ρ ( z, z, φ, δ, δ, p( θ ), ω,0) Black 0 S 0 S Aer Ray 0 No separation between atmospheric constituents, only surface Separation for aerosol effect required: 1. Subtraction of Rayleigh path reflectance Remains the combined effect of aerosol and surface. ttot ( z0) ttot ( zs ) ASurf ( z0, zs ) + 1 A ( z, z ) r ( δ, g) TOA Ray Aer 0 S 0 S 0 S Aer Aer 0 Surf 0 S Hem tot ρ ( z, z, φ) ρ ( z, z, φ, p,0) = ρ ( z, z, φ, δ, p ( θ ), ω,0) ttot ( z0) ttot ( zs ) ASurf ( z0, zs ) + 1 A ( z, z ) r ( δ, g) Surf 0 S Hem tot ttot ( z0) ttot ( zs ) A Aer TOA Surf ρ ( z0, zs, φ, paer ( θ ), ω0,0) = ρ ( z0, zs, φ) ρray ( z0, zs, φ, p,0) 1 A r ( δ, g) Surf hem tot

7 Separation of aerosol reflectance - Ocean Correction for ocean () 0 S Surf 0 S Aer (, z0, zs, ) = TOA(, z0, zs, ) Ray (, z0, zs, ) 1 ρ Surf ( λ, z0, zs, φ, v) ρ Hem ( δ ) ρ λ φ ρ λ φ ρ λ φ Atmospheric disturbance: Rayleigh path reflectance for MSI channels small or negligible effect Ozone transmittance in channel µm t( z ) t( z ) ρ ( λ, z, z, φ, v) total transmittance and hemispheric reflectance Surface effects: Surface reflectance of case 1 water: water leaving reflectance (nearly black ) Fresnell reflection White caps (in the clean ocean case of MSI black ) Fresnell reflection Transfer from the 2.1 µm channel (MODIS)

8 LUT AOT will be derived from aerosol reflectance Aerosol reflectance is obtained by RTM of TOA reflectance, using AOT, phase function, single scattering albedo and surface albedo for given geometry conditions. δ Aer ( λ) = f ( ρ Aer ( λ), z0, zs, ϕ) ρ = ρ ρ ρ Aer TOA Ray Surf δ Aer = 0 RTM calculation with δ Aer (λ), p Aer (λ, θ), ω o (λ) δ Ray (λ), p Ray (λ, θ), ρ Surf (λ), geometry Air mass correction for LUT Normalization of geometry effects to air mass dependence of single scattering ρ = ρ * 0 Aer Aer M z0 M ( z ) + M ( zs ) ( ) M ( z ) Reduction of number of LUT cases for MSI (MERIS) geometry conditions Up to AOT > 1 S TOA reflectance is corrected for Rayleigh path reflectance and surface in the same way as in the retrieval consistency with retrieval Coupling effects of aerosol and molecules and surface are included in ρ Aer

9 Cloud Screening BAER uses now 3 empirical criteria for cloud screening: 1. Increased TOA reflectance in several channels ρ TOA > ρ Cloud_Min (Kokhanovsky) flag value Reduced spectral slope of ρ TOA flag value -0.2 ρtoa( λ1 ) 1.0X ρtoa( λ 2) > ρtoa( λ1 ) 1.04 ρ ( λ ) > (clouds have 1) (for MSI λ 1 = nm, λ 2 = 865 nm, TOA 2 Frequency Cloudy Sub-pixel Clouds Clear Sky Increased heterogenity δ σ 5x5 Aer _5x5 < flag value Rho_1 / Rho_2

10 Land Screening Empirical criteria for land detection: 1. Increased NDVI ρtoa(0.659) ρtoa(0.865) NDVI = 0.2 ρ (0.659) + ρ (0.865) 2. Increase of ρ TOA with wavelength ρ TOA ( λ ) TOA 1 ρtoa( λ 2) > 1. X TOA flag value -0.4 flag value -0.5 MSI λ 1 = nm, λ 2 = 865 nm,

11 Sun Glint Correction Use of empirical correlation between the MSI channel 2.2 µm and the other channels Minimum reflectance correlation with 2.2 µm channel for cloud less conditions need to be established 2.2 µm channel aerosol effect is weak Rayleigh scattering negligible no water leaving reflectance Main effect in channel 2.2 µm is sun glint from surface Empirical linear relationship can give a sufficient estimation for the surface contribution in the other channels ρ ( λ) c ρ (2.2 µm) SunGl λ TOA

12 Some other scenes Canary Islands 14. May 2006 Mediterranean 03.July 2005 Trop. Atlantic 10. July 2002 Morth Sea, 13. June 2006

13 Validation 1. Test with synthetic data Simulations of TOA reflectance with the forward model for given AOT and geometry are used Results are compared with inputs 2. MODIS data used for test retrievals, compared against AERONET For the use of MODIS data a reading program has been developed to select MSI channels from MODIS as input for the MSI retrieval program

14 Validation results Relative Error (retrieved - true)/true * 100 Channel 1: µm Channel 2: µm Channel 3: µm Channel 4: µm Synthetic data rel. error / % / AOT Against AERONET for MSI channel 1 and 2

15 Error estimation Error is composed from: Off-set of the linear fits (AERONET validation) Channel 1: 0.04 Channel 2: 0.04 Relative error of synthetic data for the range of AOT > % for lower AOT >10 % 30 % Combined error for AOT in channel 1 and 2 δ = δ Aer Aer

16 MSI - Lidar MSI Retrieval - CALIPSO , 00:10 Intercomparison of: AOT - CALIPSO clouds 1.00 AOT AOD(0.659 µm) with MSI retrieval from MODIS-Aqua scene, :10 of Pacific region near Hawaii AOT(1) - MSI volcanic plume CALIPSO trace, integration of backscatter profiles to AOD(0.553 µm), LR = AOD comparable Volcano plume from Hawaii Latitude

17 Conclusions AOT over clean ocean AOT retrieval over clean ocean subject of project (WP4000) Algorithm has been derived from present BAER approach is existent in form of a Fortran program Algorithm in FORTRAN for LINUX and WINDOWS Deliverables to ESA: ATBD has been written first validation with two different LUT Manual for the use of FORTRAN program Summary of validation: AOT within the error range in comparable values with AERONET for channel 1 and 2 Angström α is either overestimated or underestimated depending on LUT

18 Future Plans (proposed part to IRMA) Extension to AOT retrieval over land ρ λ φ ρ λ φ ρ λ φ 0 S Surf 0 S Aer (, z0, zs, ) = TOA(, z0, zs, ) Ray (, z0, zs, ) 1 ρ Surf ( λ, z0, zs, φ, v) ρ Hem ( δ ) Consideration of variable albedo and BRDF In LUT and surface model Inter-correlation with 2.1 µm for correction of surface contribution Only MSI channel 1 fulfills conditions for retrieval of AOT Characterization of surface is planned by surface model, using NDVI and 2.1 µm reflectance Synergy with ATLID aerosol type, Angström α t( z ) t( z ) ρ ( λ, z, z, φ, v)

19 Thank you