Estimation of clear-sky land surface longwave radiation from MODIS data products by merging multiple models

The surface longwave radiation budget plays an important role in the Earth's climate system. Remote sensing provides the most practical way to map surface longwave radiation on a large scale and at a high spatial resolution. In this paper, we evaluate both surface downward longwave radiation (DLR) and upwelling longwave radiation (ULR) models under clear-sky conditions from MODIS data products. There are multiple DLR models available with variable uncertainties, and the Bayesian Model Averaging (BMA) method is incorporated in this study to combine the predictive distribution of these models for better accuracy. The integrated estimates for DLR based on the BMA method have lower root-mean-square errors (RMSEs) and higher coefficients of determination (R-2) than the best individual model. The RMSEs decreased by approximately 10 W/m(2) at two forest sites and by approximately 4 W/m(2) at other sites. The R-2 value increased at each site by more than 0.05. Two models for calculating the surface upwelling longwave radiation (ULR) are also evaluated at 16 sites. The results show that both the land surface temperature (LST)-emissivity method and the direct method, the Wang-U model underestimate the clear-sky ULR. The validation results show that the surface net longwave radiation (NLR) estimated using DLR estimates based on the BMA method and ULR estimates based on the LST-emissivity method is the most accurate.