Accurate estimation of the satellite-based global terrestrial latent heat flux (LE) at high spatial and temporal scales remains a major challenge. In this study, we introduce a Bayesian model averaging (BMA) method to improve satellite-based global terrestrial LE estimation by merging five process-based algorithms. These are the Moderate Resolution Imaging Spectroradiometer (MODIS) LE product algorithm, the revised remote-sensing-based Penman-Monteith LE algorithm, the Priestley-Taylor-based LE algorithm, the modified satellite-based Priestley-Taylor LE algorithm, and the semi-empirical Penman LE algorithm. We validated the BMA method using data for 2000-2009 and by comparison with a simple model averaging (SA) method and five process-based algorithms. Validation data were collected for 240 globally distributed eddy covariance tower sites provided by FLUXNET projects. The validation results demonstrate that the five process-based algorithms used have variable uncertainty and the BMA method enhances the daily LE estimates, with smaller root mean square errors (RMSEs) than the SA method and the individual algorithms driven by tower-specific meteorology and Modern Era Retrospective Analysis for Research and Applications (MERRA) meteorological data provided by the NASA Global Modeling and Assimilation Office (GMAO), respectively. The average RMSE for the BMA method driven by daily tower-specific meteorology decreased by more than 5 W/m 2 for crop and grass sites, and by more than 6 W/m 2 for forest, shrub, and savanna sites. The average coefficients of determination (R 2 ) increased by approximately 0.05 for most sites. To test the BMA method for regional mapping, we applied it for MODIS data and GMAO-MERRA meteorology to map annual global terrestrial LE averaged over 2001-2004 for spatial resolution of 0.05°. The BMA method provides a basis for generating a long-term global terrestrial LE product for characterizing global energy, hydrological, and carbon cycles.

Clouds and aerosols play essential roles in regulating surface incident solar radiation (Rs). It has been suggested that the increased aerosol loading over China is a key factor for the decadal variability in Rs and can explain the bias in its trend from reanalyses because the reanalyses do not include the interannual variability of aerosols. In this study, we compare the Rs derived from sunshine duration at 2,400 weather stations in China and that from five reanalyses from 1980 to 2014. The determining factors for the biases in the mean values and trends of Rs from the reanalyses are examined, with the help of Rs and the cloud fraction (CF), from satellite and 2,400 weather stations. Our results show that all reanalyses overestimate the multiyear Rs by 24.10–40.00 W/m2 due to their underestimations of CF, which is more obvious in southern China. The biases in the simulated CF in the reanalyses can explain the biases in Rs by 55–41%, and the bias in clear‐sky surface solar radiation (Rc), which is primarily due to biases in aerosol loading, can explain 32–9% of the bias in Rs. The errors in the trend of the simulated CF can explain the errors in the Rs trends in the reanalyses by 73–12%, and the trend errors in the Rc can explain 43–30% of the trend error in Rs. Our study suggests that more work is needed to improve the simulation of aerosols, clouds, and aerosol‐cloud‐radiation interactions in the reanalyses.

Surface incident solar radiation (Rs) is important for providing essential information on climate change. Existing studies have shown that the Rs values from current reanalyses are significantly overestimated throughout China. The European Centre for Medium-Range Weather Forecasts (ECMWF) recently released the fifth-generation of atmospheric reanalysis (i.e. ERA5) with a much higher spatiotemporal resolution and a major upgrade than its predecessor, ERA-Interim. This study is to verify whether ERA5 can improve the Rs simulation using sunshine duration-derived Rs values at ∼2200 stations over China from 1979 to 2014 as reference data. Compared with observed multi-year national mean, the Rs overestimation is reduced from 15.88 W·m-2 in ERA-Interim to 10.07 W·m-2 in ERA5. From 1979 to 1993, ERA-Interim (-1.99 W·m-2/decade, p<0.05) and ERA5 (-2.42 W·m-2/decade, p<0.05) estimated Rs in China continue to decrease and the latter’s decline is closer to the observed. After 1993, they both show a strong brightening, i.e. 2.26 W·m-2/decade in ERA-Interim and 1.49 W·m-2/decade in ERA5, but observations show a nonsignificant increase by 0.30 W·m-2/decade. Due to the improvement of total cloud cover (TCC) simulation by ERA5, its Rs trend bias induced by TCC trend bias is smaller than that in ERA-Interim. In addition, the reason why the simulation trend in ERA5 remains biases might be that ERA5 still ignores aerosol changes on interannual or decadal time scales. Therefore, subsequent reanalysis products still need to improve their simulation of clouds, water vapor and aerosol, especially in aerosol direct and indirect effect on Rs.

Accurate estimation of latent heat flux (LE) is critical in characterizing semiarid ecosystems. Many LE algorithms have been developed during the past few decades. However, the algorithms have not been directly compared, particularly over global semiarid ecosystems. In this paper, we evaluated the performance of five LE models over semiarid ecosystems such as grassland, shrub, and savanna using the Fluxnet dataset of 68 eddy covariance (EC) sites during the period 2000-2009. We also used a modern-era retrospective analysis for research and applications (MERRA) dataset, the Normalized Difference Vegetation Index (NDVI) and Fractional Photosynthetically Active Radiation (FPAR) from the moderate resolution imaging spectroradiometer (MODIS) products; the leaf area index (LAI) from the global land surface satellite (GLASS) products; and the digital elevation model (DEM) from shuttle radar topography mission (SRTM30) dataset to generate LE at region scale during the period [2003][2004][2005][2006]. The models were the moderate resolution imaging spectroradiometer LE (MOD16) algorithm, revised remote sensing based Penman-Monteith LE algorithm (RRS), the Priestley-Taylor LE algorithm of the Jet Propulsion Laboratory (PT-JPL), the modified satellite-based Priestley-Taylor LE algorithm (MS-PT), and the semi-empirical Penman LE algorithm (UMD). Direct comparison with ground measured LE showed the PT-JPL and MS-PT algorithms had relative high performance over semiarid ecosystems with the coefficient of determination (R 2 ) ranging from 0.6 to 0.8 and root mean squared error (RMSE) of approximately 20 W/m 2 . Empirical parameters in the structure algorithms of MOD16 and RRS, and calibrated coefficients of the UMD algorithm may be the cause of the reduced performance of these LE algorithms with R 2 ranging from 0.5 to 0.7 and RMSE ranging from 20 to 35 W/m 2 for MOD16, RRS and UMD. Sensitivity analysis showed that radiation and vegetation terms were the dominating variables affecting LE Fluxes in global semiarid ecosystem.