A Practical Two-Stage Algorithm for Retrieving Land Surface Temperature from AMSR-E Data—A Case Study Over China

Abstract

Land surface temperature (LST) is an important parameter that directly affects the water and heat balance between the Earth surface and atmosphere. Mapping the LST distribution at continuous temporal and wide spatial scales is very helpful for researching many physical and biochemical processes. Remotely sensed instruments are the key players in these studies. Passive microwave remotely sensed data have the advantage of retrieving the land parameters under nearly all weather conditions because of the power to penetrate clouds. In this study, a practical two-stage algorithm, which uses single-frequency and double-polarization passive microwave brightness temperature observations, is presented to retrieve the LST over China. The vertically polarized land surface emissivity (LSE) at 18.7 GHz is first estimated by a parameterization relationship with the polarization ratio (PR), which is defined as the ratio of the horizontal to the vertical brightness temperatures at the same frequency. Subsequently, the LST is retrieved using the estimated LSE by ignoring the atmospheric effect. The evaluation of the simulated data shows an RMSE of 1.45 K, which is very encouraging. The cross validations by the satellite thermal infrared products are from daily to monthly time scales. The daily accuracy is 3.04 K and the bimonthly accuracy is 4.43 K. A high positive bias in arid and semiarid regions and a negative bias in frequently cloudy regions are noticeable in both comparisons. These biases do not reveal the uncertainties of this LST retrieval algorithm; on the contrary, the retrieved the Advanced Microwave Scanning Radiometer − EOS (AMSR-E) LST appears to be more reasonable compared to the thermal infrared LST under certain circumstances.