Droughts are increasingly threatening ecological balance, agricultural productivity, and socio-economic resilience—especially in semi-arid regions like the Inner Mongolia segment of China’s Yellow River Basin. This study presents a hybrid drought forecasting framework integrating machine learning (ML) and deep learning (DL) models with high-resolution historical and downscaled future climate data. TerraClimate observations (1985–2014) and bias-corrected CMIP6 projections (2030–2050) under SSP2-4.5 and SSP5-8.5 scenarios were utilized to develop and evaluate the models. Among the tested ML algorithms, Random Forest (RF) demonstrated the best trade-off between accuracy and interpretability and was selected for feature importance analysis. The top-ranked predictors—precipitation, solar radiation, and maximum temperature—were used to train a Long Short-Term Memory (LSTM) network. The LSTM outperformed all ML models, achieving high predictive skill (R² = 0.766, CC = 0.880, RMSE = 0.885). Scenario-based projections revealed increasing drought severity and variability under SSP5-8.5, with mean PDSI values dropping below −3 after 2040 and deepening toward −4 by 2049. The high-emission scenario also exhibited broader uncertainty bands and amplified interannual anomalies. These findings highlight the value of hybrid AI–climate modeling approaches in capturing complex drought dynamics and supporting anticipatory water resource planning in vulnerable dryland environments.

Desertification presents major environmental challenges in Central Asia, driven by climatic and anthropogenic factors. The present study quantifies desertification risk through an integrated approach using Bayesian networks and the ESAS model, offering a holistic perspective on desertification dynamics. Four key variables—vegetation cover, precipitation, land-use intensity, and soil quality—were incorporated into a Bayesian model to evaluate their influence on desertification. A probabilistic model was developed to gauge desertification intensity, with simulations conducted at 200 geospatial points. Hazard maps for 2030–2050 were produced under climate scenarios SSP245 and SSP585, incorporating projected land-use changes. All procedures for desertification risk assessment, land-use mapping, and climate downscaling were performed using the Google Earth Engine platform. The findings suggest a 4% increase in desertification risk under SSP245 and an 11% increase under SSP585 by 2050, with the greatest threats observed in western regions such as Kazakhstan, Uzbekistan, and Turkmenistan. Sensitivity analysis indicated that vegetation quality exerts the strongest influence on desertification, reflected by a Vegetation Quality Index (VQI) ranging from 1.582 (low in Turkmenistan) to 1.692 (very low in Kazakhstan). A comparison of the Bayesian and ESAS models revealed robust alignment, evidenced by an R² value of 0.82, a Pearson correlation coefficient of 0.76, and an RMSE of 0.18. These results highlight the utility of Bayesian networks as an effective tool for desertification assessment and scenario analysis, underscoring the urgency of targeted land management and proactive climate adaptation. Although reclaimed land presents opportunities for afforestation and sustainable agriculture, carefully considering potential trade-offs with biodiversity and ecosystem services remains essential.

This study highlights the importance of incorporating temporal landscape dynamics in landslide susceptibility assessments (LSAs). Two models are compared: one integrates multi-temporal data, while the other relies solely on present conditions. Artificial Digital Elevation Models (ADEMs) for 1960, 1980, 2000, and 2020 were generated using a landscape manipulation tool to simulate phases of land degradation and rehabilitation, thereby enabling the assessment of susceptibility over time. Landslide occurrences were simulated to increase over time—primarily as a result of anthropogenic changes, such as deforestation and land use alterations—with partial stabilization following conservation efforts. Both models achieved identical AUC values of 0.97, but the blind model misclassified stable areas and missed historically unstable regions. While conventional performance metrics such as ROC curves provide insights into model accuracy, they fail to detect misclassifications arising from temporal landscape changes, leading to overestimation in some areas and underestimation in others, especially in evolving environments. This study demonstrates that neglecting temporal landscape evolution leads to flawed susceptibility maps, potentially misguiding hazard mitigation efforts. To improve LSA accuracy, the study advocates for integrating multi-temporal thematic maps and adopting performance metrics that assess temporal robustness. It emphasizes the need for a shift from a static-to-hazard paradigm to a temporally evolved susceptibility-to-hazard framework for more accurate hazard and risk predictions.

This study employed Deep Abstract Networks (DANets), independently and in combination with the Whale Optimization Algorithm (WOA), to generate high-resolution susceptibility maps for drought and wildfire hazards in the Oroqen Autonomous Banner in Inner Mongolia. Presence samples included 309 wildfire points from MODIS active fire data and 200 drought points derived from a custom Standardized Drought Condition Index. DANets-WOA models showed clear performance improvements over their solitary counterparts. For drought susceptibility, RMSE was reduced from 0.28 to 0.21, MAE from 0.17 to 0.11, and AUC improved from 85.7% to 88.9%. Wildfire susceptibility mapping also improved, with RMSE decreasing from 0.39 to 0.36, MAE from 0.32 to 0.28, and AUC increasing from 78.9% to 85.1%. Loss function plots indicated improved convergence and reduced overfitting following optimization. A pairwise z-statistic analysis revealed significant differences (p < 0.05) in susceptibility classifications between the two modeling approaches. Notably, the overlap of drought and wildfire susceptibilities within the forest–steppe transitional zone reflects a climatically and ecologically tense corridor, where moisture stress, vegetation gradients, and human land-use converge to amplify multi-hazard risk beyond the sum of individual threats. The integration of DANets with the WOA demonstrates a robust and scalable framework for dual hazard modeling.

This study aims to analyze the driving factors and threshold responses of the NDVI across different regional scales in Hunan Province, revealing the main influences on vegetation cover and the corresponding threshold effects and providing essential data for precise future afforestation planning. We use NDVI data and its associated driving factors, employing correlation analysis methods to investigate the spatial differentiation and threshold effects of vegetation driving factors at different regional scales. First, various analytical techniques, including Sen’s trend analysis, the Mann–Kendall significance test, and the Hurst index, are applied to assess changes in vegetation cover between 2000 and 2020 and to predict future trends. Second, to explore the differences in vegetation’s driving mechanisms at different regional scales, the optimal parameters-based geographic detector model is employed, which integrates continuous variable discretization methods and selects the optimal parameter set by maximizing explanatory power. This approach is particularly suitable for analyzing nonlinear relationships. Lastly, threshold regression analysis is conducted on the key driving factors identified through the optimal parameters-based geographic detector model. The results show that vegetation cover in most areas of Hunan significantly increased from 2000 to 2020; however, our predictions suggest slight degradation in the future. The optimal parameters-based geographic detector model identified topography and geomorphology as the primary factors affecting the spatial and temporal distribution of the NDVI, with notable regional differences in other factors. The influence of natural factors has weakened over time, while anthropogenic activities increasingly affect vegetation. Moreover, dual-factor influences exhibit stronger explanatory power than single-factor influences. The threshold response analysis reveals that slope is a key factor influencing the NDVI, with a positive threshold relationship observed at both the provincial and subregional scales, although the threshold points vary by subregion. The temperature and NDVI are negatively correlated, with varying threshold points across regions. The abovementioned research findings suggest that future afforestation efforts in Hunan should take into account the distinct characteristics of each subregion. Afforestation strategies should be tailored based on the specific threshold relationships observed in each area to enhance their effectiveness.

Analyzing the dominant drivers of the Food-Water-Land-Ecosystem (FWLE) nexus in the future is important for improving sustainable development in dryland ecosystems. However, the future trajectories of food–water–land–ecosystem interactions in typical drought-prone regions remain poorly understood. To address this gap, this study coupled the Gray Multi-Objective Programming with Patch-generating Land Use Simulation (GMOP-PLUS) model and applied spatial analysis methods (including longitudinal and zonal statistical analysis, trade-off synergy analysis, and redundancy analysis) to examine the spatiotemporal differentiation patterns of the FWLE nexus in Xinjiang under different development scenarios. Over the past two decades, water yield in Xinjiang’s agricultural landscapes has declined by 57.4%, primarily due to land-use and land-cover changes. Under the 2030 sustainable development scenario, a custom optimization developed via the GMOP model that balances economic and ecological objectives, crop production and habitat quality are projected to increase by 47.9% and 55.1%, respectively. Moreover, redundancy analysis results indicate that the driving contribution of precipitation on the FWLE nexus is expected to reach 76.9% by 2030. These findings provide a clear delineation of priority spatial units for improvement within Xinjiang agro-ecosystem and offer a strategic pathway for balancing ecological conservation and economic development.