Toward “all weather,” long record, and real-time land surface temperature retrievals from microwave satellite observations

摘要

The land surface temperature can be estimated from satellite passive microwave observations, with limited contamination from the clouds as compared to the infrared satellite retrievals. With ～60% cloud cover in average over the globe, there is a need for “all weather,” long record, and real-time estimates of land surface temperature (T_s) from microwaves. A simple yet accurate methodology is developed to derive the land surface temperature from microwave conical scanner observations, with the help of precalculated land surface microwave emissivities. Different tests are conducted to optimize the algorithms. The method is applied to the Special Sensor Microwave/Imagers (SSM/I) observations over 2 years, regardless of the cloud cover. The results are compared to infrared estimates from International Satellite Cloud Climatology Project (ISCCP) and from Advanced Along Track Scanning Radiometer (AATSR), under clear-sky conditions. Limited biases are observed (～0.5 K for both comparisons) with a root-mean-square error (RMSE) of ～5 K, to be compared to the RMSE of ～3.5 K between ISCCP et AATSR. Cloud contamination in the AATSR estimates have been evidenced, and a simple filtering has been proposed. The microwave surface temperatures have also been carefully compared to in situ T_s time series from a collection of more than 20 stations over a large range of environments. Very good agreement is obtained for well-controlled stations in vegetated environments (down to RMSE of ～2.5 K for several stations), but the methodology encounter difficulties under cold conditions due to the large variability of snow and ice surface emissivities.