Satellite-derived surface upward longwave radiation (SULR) is essential for monitoring the global surface radiation budget, ecological processes and climate change. However, the widely-used SULR products derived from thermal infrared (TIR) remote sensing exhibit spatial discontinuities because TIR signals cannot penetrate cloud cover. Conventional cloud-sky SULR estimation approaches often utilize post-processed reanalysis data as inputs, which could not meet the real-time requirement of operational system. This study proposes a lightweight cloud-sky SULR real-time estimation method for the Fengyun-4A (FY-4A) geostationary satellite using a Light Gradient Boosting Machine (LightGBM) model. The daytime cloud-sky SULR is estimated by applying the established relationship between auxiliary variables and clear-sky SULR to cloudy conditions, while the nighttime cloud-sky SULR values are estimated by applying the determined relationship between input variables and a publicly accessible, gap-filled SULR product. The model inputs include (1) spatial-temporal location record data, (2) multiple surface characteristic parameters generated from previous-year data, and (3) two categories of operational FY-4A radiation products, with both components being available in real-time. Validation against six Heihe Watershed Allied Telemetry Experimental Research (HiWATER) sites demonstrates that the reconstructed cloud-sky SULR achieves acceptable RMSE (MBE) values of 33.4 W/m2(1.5 W/m2) for daytime and 25.2 W/m2(4.7 W/m2) for nighttime conditions. Therefore, the proposed lightweight method could improve the spatial coverage of current FY-4A SULR product and further promote real-time SULR-related applications.

Thermal radiation directionality (TRD) describes the anisotropic signature in the thermal infrared domain, leading to significant uncertainties in current land surface temperature (LST) and surface upward longwave radiation (SULR) products. The kernel-driven model (KDM) is considered as the most potential tool to correct TRD effects due to its good tradeoff between physical accuracy and computational efficiency. However, the application of existing 4-parameter KDMs is limited due to the requirement of simultaneously ≥4 multi-angle observations. By combining a diurnal temperature cycle model, the time-evolving kernel driven model (TEKDM) had achieved significant TRD elimination effect over the overlapping region of two Geostationary Operational Environmental Satellite (i.e., GOES-16 and GOES-17) LST products. However, the performance of TEKDM in SULR TRD elimination is still not clear. In this letter, the TEKDM was extended to the overlapping region of FengYun-4A (FY-4A) and Himawari-8 satellite observations, in order to correct the directional SULR (SULRD) to hemispherical integrated SULR (SULRH). Then, a step-by-step SULR downscaling method based on multiple linear regression was conducted forSULRDandSULRH(from 4 km to 40 m). The downscaled SULRD and SULRH values were validated by the pyrgeometer-observed hemispherical SULR of 14 in-situ sites within a heterogeneous region. The validation result shows that the TEKDM could eliminate the TRD effect of FY-4A and Himawari-8 SULR with an RMSE decrease of 4.47 W/m2(16.5%) and 4.03 W/m2(15.1%), respectively. Therefore, the TEKDM also performs well for the TRD correction of SULR products of two geostationary satellites.

Daily mean, maximum, and minimum land surface temperatures (Tmean, Tmax, and Tmin) are fundamental parameters for geophysical studies. Machine learning methods provide powerful tools to estimate these three parameters from instantaneous land surface temperature (LST) observations. This paper evaluated eight machine learning methods for two Moderate Resolution Imaging Spectroradiometer (MODIS) LST products (MxD11A1 and MxD21A1) and their joint application. The preliminary results indicate the support vector regression and linear regression methods exhibit satisfactory performance and fusion of two MODIS LST products is beneficial for estimating Tmeanand Tmax. Notably, the MxD11A1 nighttime LST product performs best for estimating Tmin.

Land surface temperature (LST) is an essential parameter for geoscience studies, and the 1-km polar-orbiting satellite LST products are widely used due to their global coverage on a daily basis. Aimed to jointly utilize all 1-km LST products to thoroughly quantify the LST temporal tendency within one day, comprehensive evaluation of existing 1-km LST products over the same sites is meaningful. In this study, taking sparsely vegetated sites as an example, ten 1-km polar-orbiting LST products were evaluated against the in situ measurements of ten USCRN sites with high spatial representativeness (from December 2019 to November 2020 covering four seasons). The evaluated LST products included three Moderate Resolution Imaging Spectroradiometer (MODIS) LST products (i.e., MOD11, MYD11, and MYD21), two Sea and LST Radiometer (SLSTR) LST products (i.e., Sentinel-3A and Sentinel-3B), one LST product from Advanced Very High Resolution Radiometer (AVHRR) onboard Metop-B, three Visible Infrared Imaging Radiometer Suite (VIIRS) LST products (i.e., VNP21, S-NPP EDR, and NOAA-20 EDR), and one LST product from Visible and Infra-Red Radiometer (VIRR) onboard FY-3B. The results indicate that: 1) the RMSE is much higher in the daytime (2.5–4.0 K) than in the nighttime (1.7–2.5 K); 2) the daytime MBE is negative (from −0.3 to −2.8 K) and this underestimate trend is significantly slighter for nighttime MBE (from −1.7 to 0.2 K); and 3) the daytime RMSE varies from 2.7 to 4.6 K in summer, 2.7–4.5 K in spring, 2.5–3.7 K in winter, and 2.0–3.4 K in autumn. Furthermore, we found that the severe daytime RMSE is related to the thermal anisotropy amplitude, which is around 4 K in summer, 2 K in spring and autumn, and 1 K in winter. Correcting the thermal anisotropy of daytime LST is an essential step that needs to be taken seriously before conducting high-quality joint applications, such as diurnal temperature cycle (DTC) modeling.