Land surface temperature (LST) is one of the most valuable variables for applications relating to hydrological processes, drought monitoring and climate change. LST from satellite data provides consistent estimates over large scales but is only available for cloud-free pixels, greatly limiting applications over frequently cloud-covered regions. With this study, we propose a method for estimating all-weather 1 km LST by combining passive microwave and thermal infrared data. The product is based on clear-sky LST retrieved from Moderate-resolution Imaging Spectroradiometer (MODIS) thermal infrared measurements complemented by LST estimated from the Advanced Microwave Scanning Radiometer Version 2 (AMSR2) brightness temperature to fill gaps caused by clouds. Terrain, vegetation conditions, and AMSR2 multiband information were selected as the auxiliary variables. The random forest algorithm was used to establish the non-linear relationship between the auxiliary variables and LST over the Tibetan Plateau. To assess the error of this method, we performed a validation experiment using clear-sky MODIS LST and in situ measurements. The estimated all-weather LST approximated MODIS LST with an acceptable error, with a coefficient of correlation (r) between 0.87 and 0.99 and a root mean square error (RMSE) between 2.24 K and 5.35 K during the day. At night-time, r was between 0.89 and 0.99 and the RMSE was between 1.02 K and 3.39 K. The error between the estimated LST and in situ LST was also found to be acceptable, with the RMSE for cloudy pixels between 5.15 K and 6.99 K. This method reveals a significant potential to derive all-weather 1 km LST using AMSR2 and MODIS data at a regional and global scale, which will be explored in the future.

Downscaling of passive microwave derived soil moisture (SM) using thermal-infrared remote sensing data is a common method to obtain higher resolution SM information. However, these thermal-based downscaling methods are limited by their requirement of daily reparameterization because thermal-infrared data are influenced by both the SM content and weather conditions. To overcome these limitations, this study developed a new method called downscaling method based on optical trapezoid model (DespiTe), which improves the spatial resolution of passive microwave derived SM products by using only optical remote sensing data. This method relies on the linear relationship between SM and the shortwave-infrared transformed (STR) reflectance, which is less affected by weather conditions. Thus, the DespiTe method needs to be parameterized only once for each area. By establishing statistical relationships between optical parameters (normalized difference vegetation index (NDVI) and STR) and passive microwave derived SM observations, the DespiTe method enables a higher spatial resolution SM estimation to be derived. The downscaling results from DespiTe were evaluated with aircraft and in situ observed SM data acquired during the SMAPEx-4 and SMAPVEX16 field experiments. Moreover, these results were compared with those from disaggregation based on physical and theoretical scale change (DisPATCh), a commonly used thermal-based downscaling method. A comparison against aircraft SM data showed that the DespiTe method achieved an unbiased root-mean-square error (ubRMSE) of 0.026 m3/m3, which outperformed the DisPATCh method that achieved a ubRMSE of 0.059 m3/m3.

The SM2RAIN (Soil Moisture to Rain) model has been widely used for rainfall estimation worldwide. However, due to the lack of sufficient ground observation, the SM2RAIN model driven by different passive microwave soil moisture products over the Tibetan Plateau has not been fully validated. In this paper, four widely used satellite microwave soil moisture products (including SMAP, ASCAT, SMOS, and AMSR2) were used as input data for rainfall estimation. Rainfall data from eight ground observation stations during 2016-2018 were used to evaluate the overall performance of the SM2RAIN algorithm under various soil moisture products at different time aggregation (AGGR) scales. In addition, different satellite soil moisture products were merged to evaluate whether the combined soil moisture products could improve the performance of the SM2RAIN model. Finally, the rainfall estimates with different soil moisture data were further evaluated and compared with two benchmark rainfall products (IMERG and ERA5). Results indicate that: (1) Overall, SM2RAIN-SMAP has the highest rainfall estimation accuracy, but with the time aggregation scale up to 30 days, the mean R of the four rainfall estimates could reach above 0.8 and the mean value of Kling-Gupta Efficiency (KGE) could reach above 0.8. (2) Combined satellite soil moisture products can significantly improve the rainfall estimates. The SM2RAIN model performed the best when SMAP and ASCAT soil moisture products were combined. (3) Using the SMAP product or combined soil moisture products yielded more accurate rainfall estimates than the two benchmark rainfall products (IMERG and ERA5).

Graph regularized nonnegative matrix factorization (GNMF) has been widely used in data representation due to its excellent dimensionality reduction. When it comes to clustering polluted data, GNMF inevitably learns inaccurate representations, leading to models that are unusually sensitive to outliers in the data. For example, in a face dataset, obscured by items such as a mask or glasses, there is a high probability that the graph regularization term incorrectly describes the association relationship for that sample, resulting in an incorrect elicitation in the matrix factorization process. In this article, a novel self-initiated unsupervised subspace learning method named robust nonnegative matrix factorization with self-initiated multigraph contrastive fusion (RNMF-SMGF) is proposed. RNMF-SMGF is capable of creating samples with different angles and learning different graph structures based on these different angles in a self-initiated method without changing the original data. In the process of subspace learning guided by graph regularization, these different graph structures are fused into a more accurate graph structure, along with entropy regularization, ${L_{2,1/2}}$ -norm constraints to facilitate the robust learning of the proposed model and the formation of different clusters in the low-dimensional space. To demonstrate the effectiveness of the proposed model in robust clustering, we have conducted extensive experiments on several benchmark datasets and demonstrated the effectiveness of the proposed method. The source code is available at: https://github.com/LstinWh/RNMF-SMGF/.

Chlorophyll-a plays an essential biochemical role in the eutrophication process, and is widely considered an important water quality indicator for assessing human activity’s effects on aquatic ecosystems. Herein, 20 years of moderate resolution imaging spectroradiometer (MODIS) data were applied to investigate the spatiotemporal patterns and trends of chlorophyll-a concentration (Chla) in the eutrophic Lake Taihu, based on a new empirical model. The validated results suggested that our developed model presented appreciable performance in estimating Chla, with a root mean square error (MAPE) of 12.95 μg/L and mean absolute percentage error (RMSE) of 29.98%. Long-term MODIS observations suggested that the Chla of Lake Taihu experienced an overall increasing trend and significant spatiotemporal heterogeneity during 2002–2021. A driving factor analysis indicated that precipitation and air temperature had a significant impact on the monthly dynamics of Chla, while chemical fertilizer consumption, municipal wastewater, industrial sewage, precipitation, and air temperature were important driving factors and together explained more than 81% of the long-term dynamics of Chla. This study provides a 20 year recorded dataset of Chla for inland waters, offering new insights for future precise eutrophication control and efficient water resource management.

Dense soil moisture observation networks serve as the primary means to validate large-scale satellite soil moisture products. However, maintaining intensive observations demands substantial labor and financial resources. It is therefore crucial to address the issue of how to sparsify ground observations while still achieving comparable validation results. Accordingly, this study takes an agro-pastoral interconnected zone of the Shandian River basin as an example. Representative soil moisture sites and the minimal number of required sites (NRS) were investigated at different sampling depths (3 cm and 50 cm) and at multiple scales (3 km, 9 km, 36 km and 100 km). The average soil moisture estimated from the representative sites was then applied to validate SMAP L2, L3, and L4 multi-scale soil moisture products. The findings indicate that the spatial heterogeneity of soil moisture within the Shandian River basin remained relatively consistent across varying spatial scales. However, it exhibited a notable increase in heterogeneity when moving vertically into deeper soil layers. Representative sites can more accurately determine the average soil moisture in a region, showing that SMAP notably overestimated root zone soil moisture and underestimated surface soil moisture. This research can serve as a theoretical guide for watershed-scale soil moisture estimation as well as a solid scientific foundation for improving the architecture of the watershed soil moisture network.