Quantifying Uncertainties in Land Surface Temperature (LST) and Emissivity Retrievals from ASTER and MODIS Thermal Infrared Data
Land surface temperature and emissivity (LST and E) data are essential for a wide variety of surface-atmosphere studies, from calculating the evapotranspiration of the Earth's land surface to retrieving atmospheric water vapor. LST and E products are generated from thermal infrared data acquired from sensors such as ASTER and MODIS on NASA's EOS platforms. NASA has identified a major need to develop long-term, consistent products valid across multiple missions, with well-defined uncertainty statistics addressing specific Earth science questions. These products are termed Earth System Data Records (ESDRs) and LST and E have been identified as an important ESDR. Currently a lack of understanding in LST and E uncertainties limits their usefulness in land surface and climate models. To address this issue, a LST and E uncertainty simulator has been developed to quantify and model uncertainties for a variety of TIR sensors and LST algorithms. Using the simulator, uncertainties were estimated for the MODIS and ASTER TES algorithm, including water vapor scaling (WVS). These uncertainties were parameterized according to view angle and estimated total column water vapor for application to real data. The standard ASTER TES algorithm had a RMSE of 3.1 K (1.2 K with WVS), while the MODIS TES algorithm had a RMSE of 4.5 K (1.5 K with WVS). Accuracies in retrieved spectral emissivity for both sensors degraded with higher atmospheric water content, however, with WVS the emissivity uncertainties were reduced to <0.015. Accurately quantifying uncertainties in LST and E products not only improves their utility and understanding but will also enable the data to be fused into long-term, well characterized ESDRs.