A new approach to the retrieval of Land Surface Temperature and Emissivity from multi-channel instruments: an Optimal Estimation methodology for the ASTER instrument

With increasing demand for higher-spatial-resolution land surface temperature (LST) data, there is a need to characterise the thermal land surface in terms of both LST and land surface emissivity (LSE). This paper presents the first optimal estimation retrieval of LST and LSE from high-spatial-resolution, multi-band thermal infrared observations, demonstrated using the five-channel ASTER instrument at 90 m spatial resolution. The method works directly with top-of-atmosphere brightness temperatures and retrieves LST, multispectral LSE, and total column water vapour simultaneously, while providing traceable per-pixel uncertainty estimates. In simulations using laboratory-based emissivity databases and reference geophysical scenarios, the algorithm retrieves LST to 1.5 K or better with a median LSE error of 0.01; uncertainties per pixel are 1.43 K for LST and 0.016 for LSE on average. Validation against in situ measurements results in biases of less than 1 K, and intercomparison with the MODIS MOD21 product results in an overall bias of -0.22 K. LSE absolute biases between MOD21 band 31 and ASTER channel 13 are 0.01 to 0.02, depending on site. The new optimal estimation approach confirms the capability of multi-band sensors to derive LSE with quantified uncertainty and provides improved emissivity information for energy-budget and flux models from local to global scales.