Be aware of the significance of land surface net radiation ($R_{n}$), there is a need for accurate long-term and high spatial resolution global$R_{n}$estimates based on satellite data. Herein, we propose a novel globally applicable, highly effective algorithm for estimating daily$R_{n}$directly from Visible Infrared Imaging Radiometer Suite (VIIRS) top-of-atmosphere (TOA) observations ranging from 2011 to present, using the eXtreme Gradient Boosting (XGBoost) method. This algorithm, named the constraint conditional model (CCM), consists of five conditional models (namely, cases 1–5 model) divided by the combination of the length of daytime (dt), the instantaneous sky condition, and the surface broadband albedo, and the daily downward shortwave radiation (DSR) from ERA5-Land was introduced as a physical constraint when$dt \gt 9$, in which case$R_{n}$is dominated by$R_{\textit {si}}$(incoming solar radiation). The validation accuracy of CCM was satisfactory against the ground measurements, yielding a root-mean-square error (RMSE) of 18.95 Wm−2, a bias of 0.056 Wm−2, and an$R^{2}$of 0.89. The algorithm exhibited superior accuracy and robustness compared to GLASS-MODIS and ERA5-Land under spatiotemporally independent validation samples. This indicates the potential of VIIRS to extent MODIS$R_{n}$products for generating long-term global daily$R_{n}$data.

Downward shortwave radiation (DSR) is greatly affected by rugged terrains, which account for about 24% of the world’s surface. Yet, existing DSR products do not take into account topographical effects. Some topographic correction algorithms have been developed for estimating the clear-sky instantaneous DSR over rugged terrains (DSRins-rugged), but no specific algorithms are available to get the daily average DSR over rugged terrains (DSRdaily-rugged). The objective of this study is to develop an efficient and robust model to retrieve the clear-sky DSRdaily-rugged based on DSR satellite products. After examining ground measurements collected from several mountainous sites over the Chengde Experimental Area in China, we found that the clear-sky DSRins-rugged over a day follows a pseudo-sine curve, depending on aspect, slope, and other terrain factors, which form the foundation of our terrain correction sinusoidal model (TCSM). TCSM also includes a new simple shadow correction method. Validation against ground measurements showed that shadow-corrected clear sky TCSM DSRdaily-rugged estimated from in situ measurements is highly accurate with a root-mean-square error (RMSE) of 9.69 Wm−2, bias of 0.93 Wm−2, and$R^{2}$of 0.99. After applying TCSM to correct the topographic effects of both the Clouds and Earth’s Radiant Energy Systems synoptic Edition4 (CERES-SYN1deg_Ed4A) and MCD18A1 C6 (MCD18) DSR products, the accuracies significantly improved, with the validated RMSE reduced from 63.60 and 64.51 to 14.03 and 12.60 Wm−2, the bias from −38.58 and −36.93 to 5.53 and −7.17 Wm−2, and$R^{2}$from 0.46 and 0.44 to 0.97 and 0.98, respectively. Additionally, the TCSM can be easily applied to other DSR products that do not consider the topographic effects.

The acquisition of a time series of high spatial resolution terrestrial latent heat flux (LE) is crucial for agricultural water resource management. However, the currently and frequently used spatiotemporal data fusion model fails to capture LE spatial and temporal patterns during periods of vigorous vegetation growth due to the limited availability of cloud-free fine-resolution images. In this study, we proposed a multiphase self-adaptive (MSA) spatiotemporal data fusion model to address this issue. Unlike popular spatiotemporal data fusion models that rely heavily on cloud-free images, MSA utilizes all available fine- and coarse-resolution images, including those with partial cloud contamination, as inputs to the model. The proposed MSA method was tested at six sites representing five major land cover types across China. The results demonstrate that the MSA model, utilizing all available Landsat images, was more accurate than that relies solely on cloud-free Landsat images [coefficient of determination (R2): 0.53 versus 0.38; root mean square error (RMSE): 8.12 versus 9.93 W/m2]. We also compared the proposed method with three widely used models, the spatial and temporal adaptive reflectance fusion model (STARFM), flexible spatiotemporal data fusion (FSDAF), and Fit-FC. The results show that MSA performed better than other models at recognizing LE spatial details. Additionally, MSA produced a high spatial resolution daily LE that was the most similar to ground-observed LE (R2 = 0.34 (p < 0.01), RMSE = 27.23 W/m2, bias = −2.75 W/m2). The proposed strategy provides an alternative approach for monitoring the high spatial resolution dynamic flux of heat and water over various land cover types.

The Chinese GF-1 satellite, the first satellite of the China High-resolution Earth Observation System launched in 2013, can be used to help estimate evapotranspiration (LE), which is important for myriad hydroclimatic and ecosystem science and applications. We propose a novel approach to use the GF-1 visible and near-infrared (VNIR) measurements at 16 m and 4-day resolutions to estimate LE. The NIR (near-infrared)–red spectral-domain (NRSD) model is coupled to a perpendicular soil moisture index (PSI) and a perpendicular vegetation index (PVI). We applied the model to the Huailai agricultural region of China with 55 scenes of GF-1 imagery during 2013–2017 and validated using ground measurements with footprint models for two eddy-covariance (EC) flux tower sites and one large aperture scintillometer (LAS) site. The results illustrate that the terrestrial daily LE can be estimated with squared correlation coefficients ($R^{2}$) of 0.77–0.84 ($p < 0.01$) and root-mean-square error (RMSE) values of 17.9–21.5 W/m2among all three sites. The site-calibrated statistics are improved by 0.14–0.25 for$R^{2}$and decreased by 4.2–8.3 W/m2for RMSE as compared to the commonly used universal PT-JPL model. A satisfactory performance is achieved across all experimental conditions, encouraging the application of the NRSD model to estimate LE for other broad regions.

Surface incident shortwave radiation (Rs) is an important factor of the environment system. Knowledge of the Rs variations is important for studying the hydrological cycle and climate change. Although ground measurements provide accurate Rs data, insufficient observations limit their application for analyzing the variations of the Rs. In this study, long-term trends of the Rs over China were analyzed based on the newly reconstructed Rs data. First, the Random Forest (RF) method was used to develop the Rs estimation model using ground-measured radiation data and meteorological data from the Climate Data Center of the Chinese Meteorological Administration (CDC/CMA); then, the RF-based model was further applied to extend the Rs data at 454 CDC/CMA meteorological stations during 1970-2015; finally, the trends in Rs across China were analyzed using the Mann-Kendall (M-K) test method. The evaluation results of the RF-based model showed that the overall correlation coefficient (R) value was 0.96, the bias value was 0.06 (0.04%) Wm-2, and the root mean square errors (RMSE) value was 23.92 (14.83%) Wm-2. A significant decrease trend of the Rs in China with the average decrease of 1.1 (0.68%) Wm-2per decade showed up between 1970 and 2015. Additionally, a recovery trend of the Rs in China with the average increase of 0.4 (0.25%) Wm-2per decade was observed during 1989-2015.

Surface downward shortwave radiation (DSR) reaching the ocean surface is critical for investigating the Earth's climate and global change issues. General circulation models (GCMs) simulations are one practical approach to obtain long-term global DSR simulations. Previous studies have reported that the GCM DSR simulations exist biases and uncertainties over ocean surface. Thus, this study used the Bayesian model averaging (BMA) method to improve tropical ocean DSR simulations by weighted averaging the 48 GCM simulations. We evaluated the estimated DSR based on the BMA method using the buoy observations from the Global Tropical Moored Buoy Array (GTMBA) from 2000 to 2005. We also compared the BMA results with single GCM simulations, the estimated DSR based on the simple model averaging (SMA) method and the Clouds and the Earth's Radiant Energy System, Energy Balanced and Filled (CERES EBAF) DSR retrievals. The spatial pattern differences of the BMA results and CERES EBAF DSR retrievals were also analyzed. When the buoy observations were used as the validation data, the validation results showed that the root mean squared errors (RMSE), correlation coefficients (R), and bias of the estimated DSR based on the BMA method are 22.51 W m-2(10.27 %), 0.66, 1.79 W m-2(0.82 %), respectively. Moreover, the estimated DSR based on the BMA method were better than any individual GCM DSR simulations and the estimated DSR based on the SMA method.

Fractional vegetation cover (FVC) is an important land surface parameter for many environmental and climate-related modeling and agricultural applications. Incorporating vegetation growth information into FVC estimation process could effectively improve FVC estimation accuracy. Methods utilizing vegetation growth information from field measurement and coarse resolution FVC product have been developed recently to estimate site-scale and finer spatial resolution FVC, and achieved satisfactory performances. However, the computational efficiency of these methods is not satisfactory and they are only feasible for analyzing historical data containing a complete vegetation growth cycle. This letter developed a time-efficient FVC estimation method at Landsat scale based on temporally rich data from coarse spatial resolution Global LAnd Surface Satellite (GLASS) FVC, which facilitates development of a time-efficient dynamic vegetation growth model, and radiative transfer models linking Landsat 7 reflectance to FVC, and all combined in a probabilistic dynamic Bayesian network (DBN) framework. In addition, the proposed method is also suitable for real-time FVC estimation and has the potential to be applied on a larger scale. Validation results indicate that the performance of the proposed method is satisfactory (R2= 0.889, RMSE = 0.0917) and comparable to previously developed inefficient but well-established FVC estimation method incorporating the vegetation growth model represented by modified Verhulst logistic equation (R2= 0.884, RMSE = 0.0913).

The daytime surface net radiation (Rn) product from version 4.0 Global LAnd Surface Satellite (GLASS) product suite was recently generated from Moderate Resolution Imaging Spectroradiometer data. It is the daytime average product of Rnderived from 2000 to 2015 at a spatial resolution of 0.05°. This letter describes the results of validation of this new Rn product using ground measurements collected from 142 sites distributed worldwide. The overall accuracy of the GLASS daytime Rnproduct was satisfactory, with an R2of 0.80, root-mean-square error of 51.35 Wm-2, and mean bias error of 0.11 Wm-2. Its accuracy and quality were highly consistent for different land cover classes and elevation zones.

The accurate estimation of the terrestrial latent heat flux (LE) from satellite observations at high spatial and temporal scales plays an important role in the assessment of the water and heat exchange between the earth's surface and the atmosphere. Although a variety of data fusion methods have been proposed to merge different LE products for more reliable estimates, most of them have ignored the spatiotemporal consistency of LE products across different resolutions. In this paper, we apply the multiresolution tree (MRT) method to improve the accuracy and reduce the inconsistency between the Moderate Resolution Imaging Spectroradiometer (MODIS) LE (MOD16) product and the Landsat-based LE product at different resolutions. Eddy covariance (EC) ground measurements at five sites, MODIS and Landsat images from January 2005 to December 2005 in the north central USA, are used to evaluate the performance of the MRT method. The results show that the MRT method can improve the accuracy of the original LE products (MOD16 and Landsat), and it has the potential to significantly reduce the uncertainty and inconsistency of these products. The bias decreased by 38.3% on average, and the root-mean-square error (RMSE) decreased by approximately 49.2% after the MRT was applied at each scale. Further studies are still required to make the MRT method more universal on a variety of land cover types for long-time periods.

Surface shortwave net radiation (SSNR) and surface downward shortwave radiation (DSR) are the two surface shortwave radiation components in earth's radiation budget and the fundamental quantities of energy available at the earth's surface. Although several global radiation products from global circulation models, global reanalyses, and satellite observations have been released, their coarse spatial resolutions and low accuracies limit their application. In this paper, the Global LAnd Surface Satellite (GLASS) DSR product was generated from the Moderate Resolution Imaging Spectroradiometer top-of-atmosphere (TOA) spectral reflectance based on a direct-estimation method. First, the TOA reflectances were derived based on the atmospheric radiative transfer simulations under different solar/view geometries; second, a linear regression relationship between the TOA reflectance and SSNR was developed under various atmospheric conditions and surface properties for different solar/view geometries; third, the coefficients derived from the linear regression were used to compute the SSNR; and finally, the DSR was estimated using the SSNR estimates and broadband albedo at the surface. A 13-year (2003-2015) GLASS DSR product was generated at a 5-km spatial resolution and 1-day temporal resolution. Compared with the ground measurements collected from 525 stations from 2003 to 2005 around the world, the model-computed SSNR (DSR) had an overall bias of 8.82 (3.72) W/m2and a root mean square error of 28.83 (32.84) W/m2at the daily time scale. Moreover, the global land annual mean of the DSR was determined to be 184.8 W/m2with a standard deviation of 0.8 W/m2over a 13-year (2003-2015) period.

Fractional vegetation cover (FVC) is an important parameter for earth surface process simulations, climate modeling, and global change studies. Currently, several FVC products have been generated from coarse resolution (~1 km) remote sensing data, and have been widely used. However, coarse resolution FVC products are not appropriate for precise land surface monitoring at regional scales, and finer spatial resolution FVC products are needed. Time-series coarse spatial resolution FVC products at high temporal resolutions contain vegetation growth information. Incorporating such information into the finer spatial resolution FVC estimation may improve the accuracy of FVC estimation. Therefore, a method for estimating finer spatial resolution FVC from coarse resolution FVC products and finer spatial resolution satellite reflectance data is proposed in this paper. This method relies on the coupled PROSAIL radiative transfer model and a statistical crop growth model built from the coarse resolution FVC product. The performance of the proposed method is investigated using the time-series Global LAnd Surface Satellite FVC product and Landsat-7 Enhanced Thematic Mapper Plus reflectance data in a cropland area of the Heihe River Basin. The direct validation of the FVC estimated using the proposed method with the ground measured FVC data (R2 = 0.6942, RMSE = 0.0884), compared with the widely used dimidiate pixel model (R2 = 0.7034, RMSE = 0.1575), shows that the proposed method is feasible for estimating finer spatial resolution FVC with satisfactory accuracy, and it has the potential to be applied at a large scale.

Incident solar radiation (Rs) over the Earth's surface plays an important role in determining the Earth's climate and environment. Generally, Rscan be obtained from direct measurements, remotely sensed data, or reanalysis and general circulation model (GCM) data. Each type of product has advantages and limitations: the surface direct measurements provide accurate but sparse spatial coverage, whereas other global products may have large uncertainties. Ground measurements have been normally used for validation and occasionally calibration, but transforming their “true values” spatially to improve the satellite products is still a new and challenging topic. In this paper, an improved thin-plate smoothing spline approach is presented to locally “calibrate” the Global LAnd Surface Satellite (GLASS) Rsproduct using the reconstructed Rsdata from surface meteorological measurements. The influence of surface elevation on Rsestimation was also considered in the proposed method. The point-based surface reconstructed Rswas used as the response variable, and the GLASS Rsproduct and the surface elevation data at the corresponding locations as explanatory variables to train the thin-plate spline model. We evaluated the performance of the approach using the cross-validation method at both daily and monthly time scales over China. We also validated the estimated Rsbased on the thin-plate spline method using independent ground measurements and independent satellite estimates of Rs. These validation results indicated that the thin-plate smoothing spline method can be effectively used for calibrating satellite-derived Rsproducts using ground measurements to achieve better accuracy.

The surface downward shortwave radiation (250~3000 nm), also known as insolation, is referred to as total solar irradiance incident at Earth surface, which is an essential parameter in land surface radiation budget and many land surface process models. Currently, the downward shortwave radiation is obtained either from satellite observations based on empirical and physical-based retrieval methods or geostatistical methods using ground-based measurements. Both data type of remote sensing product convey substantial information: the ground-based measurements provide hard (accurate) but scare data, whereas, the radiation images (estimated from remotely sensed data or provided in reanalysis and GCMs data) provide exhaustive but soft (vague) information. In this paper we present a novel approach which takes advantages of both hard and soft data to estimate the surface downward shortwave radiation. This method, which based on a realistic representation of the spatiotemporal domain, can combine rigorously and efficiently various forms of physical knowledge and sources of uncertainty. Cross-validation results using ground measurements indicate that surface downward shortwave radiation estimates from BME are slightly improved.

Surface broadband emissivity (BBE) in the thermal infrared spectrum is essential for calculating the surface total longwave net radiation in land surface models. However, almost all narrowband emissivities estimated from satellite observations are in the 3-14-μm spectral region. Previous studies converted these narrowband emissivities to BBE over different spectral ranges, such as 3-14, 8-12, 8-13.5, and 8-14 μm . Errors in the calculated total longwave net radiation must be quantified systematically using these BBEs. Moreover, the best spectral range for longwave net radiation must be determined. The key to addressing these issues is the use of the realistic emissivity spectra. By applying modern radiative transfer tools, we derived the emissivity spectra of water, snow, and minerals at 1-200 μm . Using these emissivity spectra, we first investigated the accuracy of replacing all-wavelength surface longwave net radiation with the surface longwave net radiation in the 3-100-, 4-100-, 2.5-100-, 2.5-200-, and 1-200-μm spectral domains. Surface longwave net radiation at 2.5-200 μm was found to be optimal, with a bias and root mean square (rms) of less than 0.928 and 0.993 W/m2, respectively. We calculated the errors when estimating surface longwave net radiation at 2.5-200 μm with BBE in different spectral ranges. The results show that BBE at 8-13.5 μm had the lowest error and the corresponding bias and rms were less than 0.002 and 1.453 W/m2, respectively. When the 2.5-200-μm surface longwave net radiation calculated by the 8-13.5-μm BBE was used to replace the all-wavelength surface longwave net radiation, the average bias and rms were 1.473 and 2.746 W/m2, respectively. Using the most representative emissivity spectra, we derived the conversion formulas for calculating BBE at 8-13.5 μm from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Moderate Resolution Imaging Spectrometer (MODIS) narrowband emissivity products.

This letter assesses three satellite-estimated surface downwelling shortwave irradiance data sets: 1) GEWEX-SRB; 2) ISCCP-FD; and 3) CERES-FSW, using ground measurements collected at 36 globally distributed sites from 2000–2002. SRB and FD solar irradiance are available at three hourly intervals during daytime and FSW hourly solar irradiance is available at late morning. The data are compared to ground measurements at the temporal resolutions of the satellite measurements. Results indicate that the downwelling solar irradiances of the three products show good overall agreement with ground measurements. FSW has an overall coefficient of determination of$R^{2} = \hbox{0.69}$, a bias of 29.7$\hbox{Wm}^{-2}$(6.0% in relative value), and a standard deviation (STD) of 123.2$\hbox{Wm}^{-2}$(25.1% in relative value). The values are 0.83,$-5.5\ \hbox{Wm}^{-2}$($-$1.9%), 101.3$ \hbox{Wm}^{-2}$(35.0%) for SRB, and 0.83, 2.8$\hbox{Wm}^{-2}$(0.3%), 101.7$\hbox{W m}^{-2}$(35.0%) for FD. However, there are substantial uncertainties in these products in some regions. For example, large biases ranging from$-90.2\ \hbox{Wm}^{-2}$to 45.8$\hbox{Wm}^{-2}$are found in SRB in Southeast Asia. FD has large biases in Southeast Asia (58.9$\hbox{Wm}^{-2}$) and Greenland$\hbox{(-33.2}\ \hbox{Wm}^{-2}\hbox{)}$. FSW has substantial biases in Southeast Asia (72.6$\hbox{Wm}^{-2}$) and Japan (66.5$\hbox{Wm}^{-2}$); while$R^{2}$is 0.35 in Tibetan Plateau and 0.47 in Southeast Asia.