All-weather land surface temperature estimates from microwave satellite observations, over several decades and real time: methodology and comparison with infrared estimates C. Jimenez, C. Prigent, F. Aires, S. Ermida Estellus, Paris, France Observatoire de Paris, France IPMA, Lisbon, Portugal LSA-SAF, Lisbon, 06/2018
Motivation 2 LST from thermal infrared and passive microwaves LST traditionally measured from thermal infrared - the most direct estimate ~ e IR LST 4 e IR close to 1, varying on a rather limited range - measuring the skin temperature LST can also be measured from passive microwaves, but: - more sensitivity to the emissivity e MW ~ e MW LST e MW changing with vegetation, soil moisture - lower frequencies have larger penetration depth (in that case measuring an integrated subsurface temperature) - large collection of estimates from different sensors and platforms ( e.g. AVHRR, GOES, MODIS, AIRS, ATSR, MSG, AATSR ), with possible high temporal and spatial resolutions - only polar satellites (e.g. SMMR, SSM/I, AMSR-E, ) and low spatial resolution
Motivation IR emissivity 10.8um 01/2004 IR and MW emissivity 3
Motivation 4 What can MW observations offer? IR estimates being limited to clear-sky, MW can penetrate clouds to a large extent and provide an all-weather LST (ISCCP, Rossow and Schiffer, BAMS, 1999) No microwave LST products routinely available today
Methodology 5 Original multi-variable retrieval over land from SSM/I inversion method based on a neural network trained with simulated brightness temperatures (BT) to retrieve simultaneously the LST, e MW, atmospheric water vapor, and cloud liquid water [Aires et al. (2001), JGR] example of retrieved fields from inversion algorithm [03/04/1993]
Methodology 6 New single-variable LST retrieval over land for MW conical scanners [SSM/I, SSMIS, AMSRE-E] Idea: to use the multi-retrieval LST combined with a minimum of ancillary data to derive a simple transfer function between BTs and LST. (a) Training the neural network Observations: 7 MW Brightness Temperatures (from instrument) 19 GHz, 22 GHz V, 36 GHz, 85 GHz First Guess Information: 7 emissivities (climatology) Neural Network Target output Surface Skin Temperature (from multi-retrieval scheme) Surface Skin Temperature training phase (b) Inverting observed MWs Tbs Observations: 7 MW Brightness Temperatures (from instrument) First Guess Information: 7 emissivities (climatology) Neural Network Surface Skin Temperature operational retrieval
Methodology [Jimenez et al. (2017), JGR] 7 AMSR-E retrieval example - retrieved LST from the 1.30 pm AMSR-E 02/08/2008 daytime overpasses. - associated theoretical retrieval uncertainty from an analysis of the retrieval error in the calibration dataset.
Evaluations in situ Complicated by MW spatial resolution AMSR-E overpass at KIT/LSA-SAF sites ~12km - comparisons are challenging du to the mismatch between the footprints of the MW LST (~25km SSM/I, ~12km AMSR-E) and the station (meters) [Jimenez et al. (2017), JGR] 8
Evaluations in situ Complicated by MW spatial resolution 10K 10K - Example of LST, soil T, and soil moisture at 3 stations in Tibet inside the ground footprint of one SSM/I observation 20% [Prigent et al. (2016), JGR] 9
Comparisons SSM/I in situ 10 SSM/I LST at Cabauw [crop/grassland] 51.97N 4.93E ΔT snow Bias 0.65 RMSD 2.53 2003 2004 [Prigent et al. (2016), JGR] - SSM/I LST closely tracking surface temperature (see steep changes), with the largest differences found in winter, possibly associated to snow episodes.
11 Comparisons SSM/I in situ SSM/I LST at Lindenberg [forest] 52.18N 13.95E snow snow Bias 0.09 RMSD 2.65 2003 2004 [Prigent et al. (2016), JGR] - MW emissivity showing little variability and the SSM/I LST closely tracking the station LST, with the largest differences found again in winter for snow episodes.
12 Comparisons MODIS/AMSR-E in situ MODIS & AMSR-E at 10 stations [2010] [Jimenez et al. (2017), JGR] MODIS AMSR-E (clear) AMSR -E(cloudy) dark color: Night light color: Day - MODIS and AMSR-E comparable biases, but larger AMSR-E RMSDs. - A large number of cloudy situations observed by AMSR-E and missed by MODIS
Comparisons SSM/I AATRS - satellite SSM/I & AATSR [2003] SSMI/I warm bias switch in AATSR algorithm to invert snow and snow-free? lower SSM/I LST due to sub-surface sampling? [Prigent et al. (2016), JGR] - Complicated evaluation due to different satellite overpass times (up to 1h), which contributes to the observed large differences. 13
Comparisons MODIS AMSR-E satellite 14 MODIS & AMSR-E [2011] [Ermida et al. (2016), JGR] - Overall AMSR-E warm bias (~2K), with strange large positive bias for bare soils.
Night Day Night Day [Ermida et al. (2016), JGR] 15 Comparisons multi-ir AMSR-E satellite GOES, SEVIRI, MTSAT, MODIS & AMSR-E [2011] GOES-MODIS SEVIRI-MODIS MTSAT-MODIS +15K -15K GOES-AMSR-E SEVIRI-AMSR-E MTSAT-AMSR-E - Comparable biases between GEOS & MODIS and AMSR-E & MODIS at some regions, but overall larger biases and RMSDs for AMSR-E as noticed in the local analyses.
Further work 16 ESA CCI LST - Building a first long data record of MW LST by processing the SSM/I (F11, F13) and SSMIS (F17) from the CM SAF FCDR of inter-calibrated radiances
Further work 17 Starting some preliminary F13 processing at a few sites KIT site Farm Heimat, [Kalahari, semi-desert] - Comparing with ERA-Interim sampled at same location and times.
Further work Starting some preliminary F13 processing at a few sites FLUXNET site BR-Sa1, [Amazonia, tropical forest] - Comparing with ERA-Interim sampled at same location and times. 18
Summary 19 Microwave wavelengths, being much less affected by clouds than the infrared, are an attractive alternative in cloudy regions. A simple and fast microwave LST retrieval intended for production of long time series and near-real time operations has been developed. Comparisons with IR in situ and satellite IR LSTs shows reasonable bias and RMSDs for different environments, but difficulties noticed for specific conditions (e.g., snow, arid) Long time series of calibrated microwave observations now available from CM SAF, (www.cmsaf.eu), making possible to derive consistent long time series of microwave LST for climate applications. A first ~30 year data record of microwave all weather LST will be produced in the framework of the ESA CCI initiative by processing DMSP F11-F13-F17 SSMI observations Additional work related to algorithm improvements for difficult inversions (e.g., snow, sub-surface sampling, cloud contamination), joint infraredmicrowave LST estimation, downscaling of the MW estimates,...
Publications 20 Original multi-variable atmosphere+surface methodology Aires, F., C. Prigent, W. B. Rossow, M. Rothstein, A new neural network approach including first-guess for retrieval of atmospheric water vapor, cloud liquid water path, surface temperature and emissivities over land from satellite microwave observations, J. Geophys. Res., doi: 10.1029/2001JD900085. [RD-1]. Evaluation with near-surface air temperature Prigent, C., F. Aires, and W. B. Rossow, Land surface skin temperatures from a combined analysis of microwave and infrared satellite observations for an all-weather evaluation of the differences between air and skin temperatures, J. Geophys. Res., 10.1029/2002JD002301. Evaluation with ground point-based observations Catherinot J., C. Prigent, R. Maurer, F. Papa, C. Jiménez, F. Aires, and W. Rossow, Evaluation of 'all weather' microwave-derived land surface temperatures with in situ CEOP measurements, J. Geophys. Res., 116, D2310. doi:10.1029/2011jd016439. Development and evaluation of a simplified methodology targeting only LST Prigent, C., C. Jimenez, and F. Aires, F., Towards all weather, long record, and real-time land surface temperature retrievals from microwave satellite observations, J. Geophys. Res., doi:10.1002/2015jd024402. Jimenez, C., C. Prigent, S.L. Ermida, and J-L. Moncet, Inversion of AMSR-E observations for land surface temperature estimation - Part 1: Methodology and evaluation with station temperature, J. Geophys. Res., doi:10.1002/2016jd026144 [RD-2]. Ermida, S.L., C. Jimenez, C., C. Prigent, I.F. Trigo, and C.C DaCamara, Inversion of AMSR-E observations for land surface temperature estimation - Part 2: Global comparison with infrared satellite temperature, J. Geophys. Res., doi: doi:10.1002/2016jd026148.
All-weather land surface temperature estimates from microwave satellite observations, over several decades and real time: methodology and comparison with infrared estimates C. Jimenez, C. Prigent, F. Aires, Carlos Jimenez, S. Ermida Estellus, Paris, France Observatoire de Paris, France IPMA, Lisbon, Portugal LSA-SAF, Lisbon, 06/2018