Urban flooding has become a major challenge in rapidly urbanizing regions due to increasing impervious surfaces, drainage limitations, and extreme rainfall events. Existing approaches often address hydrological modeling, spatial analysis, or Green Infrastructure (GI) planning independently, limiting their ability to provide an integrated decision-support framework. To address this gap, this study proposes GIStorm-Net, a hybrid framework that integrates hydrological feature engineering, GIS-based spatial validation, Hybrid XGBoost–Graph Neural Network (GNN) flood-risk prediction, and Hybrid Grey Wolf Optimizer–Particle Swarm Optimization (GWO–PSO) for GI planning. The framework also provides a foundation for future integration with satellite-enabled networks, enabling the incorporation of remote-sensing observations and real-time environmental sensor data for continuous urban flood-risk monitoring. Experimental results demonstrated that the proposed framework achieved an accuracy of 0.9798, precision of 0.9504, recall of 0.9466, F1-score of 0.9485, and ROC-AUC of 0.9869 for urban flood-risk prediction. The proposed framework is readily extensible to IoT environmental monitoring systems, enabling future integration of real-time remote sensing and sensor observations for adaptive urban flood-risk prediction and Green Infrastructure planning. Scenario-based optimization further estimated potential improvements of 39.6
All-India Rainfall (AIR) is influenced by several large-scale climate drivers, yet most statistical and machine-learning studies assess prediction skill only in terms of anomaly accuracy, without examining how well models reconstruct long-term components of the rainfall signal. This study introduces a phased residual-modelling framework that separately extracts and reinserts the mean, climatology, and trend before predicting the residual rainfall component using Indian Ocean Dipole (IOD) sea-surface temperatures at a three-month lead. The design, inspired by dynamical-model diagnostics, enables explicit evaluation of how SST forcing contributes to different temporal components of rainfall variability. Across four phases, results show that ANN and regression models reproduce interannual variability reasonably well (correlation ≈ 0.7), but climatological peaks and long-term trends are not reliably reconstructed unless explicitly supplied. Detrending stabilizes predictions by eliminating spurious low-frequency drift but does not increase interannual predictability, indicating limited SST control over AIR’s low-frequency variability. The analysis highlights the deterministic and non-deterministic components of the SST–AIR linkage and demonstrates how phased decomposition improves interpretability, offering a systematic approach to assess predictability limits in climate data-driven models.
Frequent and severe droughts disrupt ecosystem carbon and water processes, thereby altering ecosystem water use efficiency (WUE). However, the temporal variations in WUE and the contributions of its associated drivers during the drought, resistance, and recovery periods remain unclear. WUE is an integrated indicator of ecosystem carbon–water coupling and is important for evaluating ecosystem responses to drought. In this study, WUE anomalies during the drought, resistance, and recovery periods across different vegetation types in China from 1982 to 2018 at the 1-, 3-, and 12-month drought timescales were analysed. We further quantified the spatial patterns of the relative contributions of climatic and vegetation factors to WUE variations. During the drought period, the spatial proportion of negative WUE anomalies decreased with increasing drought timescale, whereas that of positive anomalies increased. The normalized difference vegetation index (NDVI) accounted for the greatest proportion of high-contribution areas across China, with a relative contribution exceeding 30
The Bouregreg Basin is one of the basins enriched with surface and groundwater resources that serve important purposes in the areas downstream of the Sidi Mohammed bin Abdullah Dam; however, due to declining rainfall, these resources are now at risk of depletion, placing the basin under a severe threat of drought. This study aimed to identify rainfall trends and analyse their effects on water resources and streamflow. Data from three stations representing the Bouregreg Basin, appropriately distributed across it, were used: the Tsalat, Lalla Chafia and Aguibat Eziar stations. These data include rainfall and streamflow for the period from 1961 to 2022. The methodology employed was based on the application of the Mann–Kendall test, the Sen slope, and the Standardised Precipitation Index (SPI) for the rainfall series, as well as the Streamflow Drought Index (SDI) for the hydrological series, Statistical methods identified a significant breakpoint between 1978 and 1980 for all stations. Since then, the average annual rainfall experienced a sharp decline. This decline was accompanied by a sharp drop in streamflow, with total reductions in flow ranging from 6.25 m3/s at the headwaters to 20.74 m3/s at the mouth. With regard to the frequency of droughts, over 62-year period, 22 years of drought were recorded, with severe to extreme drought peaks in 1994–1995 and 2006–2007. As for the correlation coefficient between the precipitation index (SPI) and the streamflow index (SDI), it was found that the response of streamflow (SDI) to precipitation (SPI) was very strong in autumn, moderate in winter and spring, and negative in summer.
Near-surface atmospheric responses to large surface modifications remain poorly quantified in high-altitude arid regions, particularly for lithium brine evaporation ponds, whose atmospheric effects have received little attention compared with hydrological and geochemical impacts. This study investigates near-surface atmospheric changes associated with extensive evaporation ponds in the Argentine Puna using a Before–After–Control–Impact (BACI) framework applied to reanalysis-based meteorological data. Austral winter conditions during 2010–2014 and 2019–2023 were compared between an impact domain containing the evaporation-pond systems and a nearby control domain. Statistical uncertainty was evaluated using a hierarchical moving-block bootstrap accounting for multi-year structure and short-term serial dependence. Differential changes were detected in air temperature (BACI = +0.55 °C), relative humidity (−2.7
