Land Surface Temperature, Emissivity and Long-Wave Downwlling Fluxes from MSG Observations: current status and way forward Isabel Trigo, Sandra Freitas, Carla Barroso, Isabel Monteiro, Pedro Viterbo 1
Outline Land Surface Temperature & Emissivity Algorithms & Product characteristics Validation Strengths and Weaknesses Developments for CDOP-2 Downwelling Sfc Long-wave Flux Algorithm & Product characteristics Validation against in situ observations and CERES fluxes Developments for CDOP-2 2
SEVIRI/Meteosat IR channels & LST Solar Conamination SEVIRI reponse function Emissivity uncertainty high 3.9 8.7 10.8 12.0 3
Land Surface Temperature - Algorithm Generalised Split-Window 10.8µm and 12.0µm Trained usingclear SKYsynthetic SEVIRI/MSG data (MODTRAN + realistic sfc parameters & atmospheric profiles) 1 ε ε T10.8 + T12.0 1 ε ε T10.8 T12. 0 T s = ( A1 + A2 + A3 ) + ( B1 + B2 + B3 ) + C 2 2 ε ε 2 ε ε 2 GSW parameters depend on: 1. total column water ECMWF fc vapour 2. viewing angle Channel Emissivity Fraction Vegetation Cover 4
Emissivty Pixel MSG ε = ε veg FVC+ ε ground (1-FVC) + δε LSA SAF Product Sfc Reflectances VIS Non-accounted effects (multiple reflections at sfc) Variability of bare ground/ vega within pixel 5
Channel Emissivty Vega/Ground BandEmissivity for VEGETATION/ SOILclasses Emissivity at λ (Spectral Libraries) ε c-vega/soil = λ 2 f λ λ1 λ 2 λ1 f ε λ λ B B λ λ dλ dλ SEVIRI/Meteosat Channels -IR 3.9 -IR 8.7 -IR 10.8 -IR 12.0 Channel response function Broad Band λ1= 3 µm; λ= 14 µm (fλ=1) 6
Emissivty SEVIRI IR3.9 SEVIRI IR10.8 Broad-Band3 14 µm Updated Daily 7
Emissivty uncertainty & impact on LST 1 Desert Areas EMISSIVITY DRY 0.9 0.8 0.7 0.6 0.5 IR3.9 IR8.7 IR10.8 IR12.0 BroadB EverGreen Forests 1 0.9 0.8 0.7 0.6 0.5 IR3.9 IR8.7 IR10.8 IR12.0 BroadB SEVIRI Channels LST Error(K) MOIST 8
Emissivty: Next Phase CDOP-2 Improved Emissivity (EM) maps EM (& LST) combination which minimizes: f = [L obs L RTM (LST, LSE)] i 2 for i = channelsir10.8, IR12.0, 6and 12 UTC, assuming EMunchanged Limited Experience carried out using: LSA SAF LSTand current EMas 1st guess MODTRAN for L RTM ( to be replaced by RTTOV) Trigo et al (2008) in IEEE Trans. Remote Sens. Geosc. Peres et al (2010) in Int J Remote Sens. 9
Emissivty: Next Phase CDOP-2 EMISSIVITY IR10.8 To be explored during CDOP-2: ALBEDO Better EM constraints among channels Temporal (& spatial) filtering 1 Pixel: 10
Land Surface Temperature - SEVIRI LST Generation Frequency- 15 min clear sky pixels... over land... where estimated errors < 4K Available since Europe Feb2005 Full disk Jul2005 11
Land Surface Temperature - SEVIRI LST Error Bars And error bars estimated taking into account: Uncertainty of the GSW regressions Propagation of input uncertainties: Emissivity Sensor noise TCWV ECMWF forecasts 0.0 0.8 1.6 2.4 3.2 4.0 ºC Masked out δlst > 4K Trigo et al (2008) in J. Geophys. Res. Freitas et al (2010) in IEEE Trans. Remote Sens. Geosc. 12
LST - Validation Validation continuous activity main source of info on product compliance with requiremets Comparison against in situ observations: LSA SAF stations: Portugal (Evora, since 2005) Namibia (Gobabeb + Kalahari farm) Field Campaigns (e.g., AMMA) Comparison with similar products derived from other satellites: MODIS AATSR 13
Gobabeb LST - Validation Namibia 14
LST - Validation LST MSG/SEVIRI versus T sup in situ at Gobabeb Obs In Situ (ºC) 60 40 20 0 May 2008 Bias -0.31ºC RMSE 1.16ºC 0 20 40 60 LST SEVIRI/MSG (ºC) Obs In Situ (ºC) 60 40 20 0 Jul 2008 Bias -0.91ºC RMSE 1.67ºC 0 20 40 60 LST SEVIRI/MSG (ºC) Obs In Situ (ºC) 60 Nov 2008 40 20 Bias +0.24ºC RMSE 1. 44ºC 0 0 20 40 60 LST SEVIRI/MSG (ºC) Obs In Situ (ºC) 60 Mar 2009 40 20 Bias -1.20ºC RMSE 1.74ºC 0 0 20 40 60 LST SEVIRI/MSG (ºC) 15
LST - Validation EVORA L1 (L1, L2, L3) average L2 L RR ( T ) + ( 1 ε ) L = ε RR _ sfc LRR sfc RR _ sfc RR _ atm L3 scene emissivity LST_InSitu downward radiation [( ) ( ) ( ) ( ) ] 2 2 2 2 δ LST + δ LST + δ LST δ 1 2 δ ( LST ) = ε + RotRad InSitu InSituVarT InSituVarSp 16
LST - Validation EVORA ( C) SEVIRI MODIS Daytime BIAS +1.9-1.8 RMSD 2.2 2.6 ( C) SEVIRI MODIS Night-time BIAS -1.7-2.6 RMSD 2.1 2.7 17
LST - Validation SEVIRI vs MODIS LST Iberian Peninsula Morning MODIS passage (~11 UTC): [ C] 2-8Jan06 [ C] 14-20Sep05 T_SEVIRI T_MODIS MODIS Zenith Angle Remotely sensed LST is directional variable Research towards an isotropic LST during CDOP-2 18
Concluding Remarks - LST LST is generated, archived & disseminated (NRT or off-line) on an operational basis using: SEVIRI/Meteosat Generalized Split-Window AVHRR/MetOp Validation exercises show that: LST accuracy depends on retrieval conditions in accordance to information provided by respective error bars. CDOP-2: Improve EMISSIVITY CDOP-2: Revise LST algofor MTG LST is directional variable CDOP-2: model of directional effects for different land covers LST uncertainty; LST correction. 19
Downwelling Long-wave Flux at the Sfc DSLF IRradiation(4-100 µm)emitted by whole atmosphere Bulk Parameterization: F = ε sky σt sky 4 Temperature& Water Vapour Profiles Clouds/ Cloud Type Remote Sensing NWP Trigo et al. (2010) in J Geophys Res., in press 20
DSLF - SEVIRI Generation Frequency 30 min over landpixels available since 2005 with revised algorithm since 2009. 21
DSLF - Validation Ground Stations (BSRN) & CERES Northern Europe Stations CERES SEVIRI DSLF Period Jan 2006 -Apr 2007 clear sky cloudy Stations Cambourne, UK Lerwick, UK Toravere, Estonia Clear Sky All Sky Bias -0.0-2.1 CERES RMSE 12.7 23.5 SEVIRI DSLF Bias -2.2 3.1 RMSE 22.3 25.3 Problems: -Cloud identification at the edge of Meteosat disk -DSLF model for ice clouds -Temperature inversions 22
DSLF - Validation Ground Stations (BSRN) & CERES Central Europe Stations CERES SEVIRI DSLF Period Jan 2006 -Apr 2007 clear sky cloudy Stations Palaiseau, France Payerne, Swtzerland Carpentras, France Clear Sky CERES Bias RMSE -0.9 13.4 SEVIRI DSLF Bias RMSE 0.8 14.5 Problems: -Temperature inversions All Sky -1.4 22.7 1.6 22.5 23
DSLF - Validation Ground Stations (BSRN) & CERES Semi-Arid & Desert Stations CERES SEVIRI DSLF clear sky cloudy Period Jan 2006 -Apr 2007 Stations Tamanrasset, Algeria Sde Boqer, Israel Niamey, Niger Clear Sky CERES Bias RMSE 13.0 26.8 SEVIRI DSLF Bias RMSE -4.4 14.1 Problems: -Impact of high aerosol loads All Sky 17.5 29.6-5.4 16.9 24
Concluding Remarks - DSLF DSLFis generated, archived & disseminated (NRT or off-line) on an operational basis using: SEVIRI/Meteosat Bulk Parameterization AVHRR/MetOp Scheme DSLF validation against in situ data: 60-70%of retrievals meet the target accuracy of 10% low up to mid latitudes accuracy comparable to CERES downward long-wave fluxes. CDOP-2: Algorithm development shared within SAF Network, targeting known deficiencies Aerosol, ice clouds, Review existing dependence on NWP model output. 25
http://landsaf.meteo.pt 26