Timely and accurate detection of burned areas is crucial for assessing fire damage and contributing to ecosystem recovery efforts. In this study, we propose a framework for detecting fire-affected vegetation anomalies on the basis of a ResNet deep learning (DL) algorithm by merging spectral and textural features (ResNet-IST) and the vegetation abnormal spectral texture index (VASTI). To train the ResNet-IST, a vegetation anomaly dataset was constructed on high-resolution 30 m fire-affected remote sensing images selected from the Global Fire Atlas (GFA) to extract the spectral and textural features. We tested the model to detect fire-affected vegetation in ten study areas across four continents. The experimental results demonstrated that the ResNet-IST outperformed the VASTI by approximately 3% in terms of anomaly detection accuracy and achieved a 5–15% improvement in the detection of the normalized burn ratio (NBR). Furthermore, the accuracy of the VASTI was significantly greater than that of NBR for burn detection, indicating that the merging of spectral and textural features provides complementary advantages, leading to stronger classification performance than the use of SFs alone. Our results suggest that deep learning outperforms traditional mathematical models in burned vegetation anomaly detection tasks. Nevertheless, the scope and applicability of this study are somewhat limited, which also provides directions for future research.

To preserve ecological integrity and promote sustainable progress in the Three-River Headwaters Region (TRHR), it is vital to understand the vegetation alteration patterns and the sensitivity of these patterns to climatic and anthropogenic influences. In this study, we retrieved the fractional vegetation cover (FVC) through the dimidiate pixel model, driven by MODIS reflectance data from 2001 to 2022, and analyzed its spatiotemporal variations and responses to climate variation and human activities via partial correlation and residual analyses. The results indicated that the FVC retrieval accuracy reached 84.2%. From 2001 to 2022, the growing season FVC displayed a fluctuating yet overall increasing trend, with an average growth rate of 0.23% per year (p < 0.01). The vegetation significantly improved in 50.72% of the TRHR, with the Yellow River source area exhibiting the most notable improvement. However, 67.42% of the TRHR experienced a transition from improvement to degradation in vegetation, indicating a pessimistic outlook for future changes. Partial correlation analysis revealed that temperature had a pronounced influence on the southwestern Yellow River Basin and the southern Yangtze River Basin, whereas precipitation had a substantial effect on the southwestern and northeastern sections of the Yellow River Basin. Additionally, residual analysis revealed that climate change served as the predominant factor behind the changes in the FVC, whereas the anthropogenic intervention contributed substantially to vegetation improvements in the northeastern and western portions of the Yellow River Basin. Our study provides scientific support for the construction of ecological security barriers and the harmonious development of humans and nature in the TRHR.

Accurate terrestrial evapotranspiration (ET) estimation is crucial for understanding land–atmosphere interactions, evaluating ecosystem functions, and supporting water resource management, particularly across climatically diverse regions. To address the limitations of traditional ET models, we propose a simple yet robust Sigmoid-RH model that characterizes the nonlinear relationship between relative humidity and ET. Unlike conventional approaches such as the Penman–Monteith or Priestley–Taylor models, the Sigmoid-RH model requires fewer inputs and is better suited for large-scale applications where data availability is limited. In this study, we applied the Sigmoid-RH model to estimate ET over mainland China from 2001 to 2018 by using satellite remote sensing and meteorological reanalysis data. Key driving inputs included air temperature (Ta), net radiation (Rn), relative humidity (RH), and the normalized difference vegetation index (NDVI), all of which are readily available from public datasets. Validation at 20 flux tower sites showed strong performance, with R-square (R²) ranging from 0.26 to 0.93, Root Mean Squard Error (RMSE) from 0.5 to 1.3 mm/day, and Kling-Gupta efficiency (KGE) from 0.16 to 0.91. The model performed best in mixed forests (KGE = 0.90) and weakest in shrublands (KGE = 0.27). Spatially, ET shows a clear increasing trend from northwest to southeast, closely aligned with climatic zones, with national mean annual ET of 560 mm/yr, ranging from less than 200 mm/yr in arid zones to over 1100 mm/yr in the humid south. Seasonally, ET peaked in summer due to monsoonal rainfall and vegetation growth, and was lowest in winter. Temporally, ET declined from 2001 to 2009 but increased from 2009 to 2018, influenced by changes in precipitation and NDVI. These findings confirm the applicability of the Sigmoid-RH model and highlight the importance of hydrothermal conditions and vegetation dynamics in regulating ET. By improving the accuracy and scalability of ET estimation, this model can provide practical implications for drought early warning systems, forest ecosystem management, and agricultural irrigation planning under changing climate conditions.

Accurately estimating latent heat flux (LE) across water-limited grassland ecosystems is critically hampered by strong land-surface heterogeneity and pronounced intra-annual variability. Here, we proposed a ConvTransformer framework by integrating MODIS remote sensing products, China Meteorological Forcing Dataset (CMFD) data, and eddy covariance observations from six grassland sites to estimate daily LE across the Inner Mongolia grasslands. The model was evaluated using a leave-one-site-out cross-validation strategy and compared with three widely used machine learning models, including random forest (RF), gradient boosting regression trees (GBRT), and support vector regression (SVR). Across the six validation sites, the ConvTransformer achieved an average R² of 0.69, an RMSE of 12.56 W m⁻², a Bias of 0.67 W m⁻², and an average KGE of 0.82. Although RF produced slightly higher R² values at several individual sites, the ConvTransformer exhibited the highest overall KGE and the most stable performance, indicating superior cross-site generalization. Based on the trained model, a 1 km daily LE dataset for the Inner Mongolia grasslands during 2003–2018 was generated. The estimated LE revealed a distinct decreasing gradient from southeast to northwest and marked seasonality, with summer dominating the annual latent heat exchange. These results suggest that the ConvTransformer constitutes an effective framework for regional LE estimation, while also offering a valuable alternative for ecohydrological studies and regional water-resource assessment in water-limited grassland ecosystems.

Understanding ecosystem carbon processes relies heavily on the reliable assessment of gross primary productivity (GPP) yet remains challenging in the Inner Mongolia grasslands due to data scarcity and high uncertainty among existing products. We developed a ConvTransformer-based framework that exploits complementary information from satellite observations and meteorological datasets to enhance the representation of complex spatiotemporal dependencies in grassland ecosystems. Grounded in leave-one-site-out cross-validation across six eddy covariance sites, the model achieved average performance metrics of R² = 0.59, RMSE = 1.40 g C m⁻² d⁻¹, Bias = −0.31 g C m⁻² d⁻¹, and KGE = 0.46, outperforming traditional machine learning models (RF, GBRT, and SVR) as well as the light use efficiency model (EC-LUE) in both accuracy and robustness. Using this framework, we generated a daily GPP dataset at spatial granularity of 1 km for the Inner Mongolia grasslands from 2003 to 2018. The results reveal a clear spatial gradient, with GPP decreasing from southeast to northwest. Comparisons with established products, including FLUXCOM, BESS V2, and PML V2, show strong spatial consistency and reduced discrepancies, supporting the reliability of the estimates. Overall, the proposed framework provides an effective approach for characterizing regional carbon dynamics and supports long-term ecological monitoring in semi-arid regions.