Rapid urbanization combined with the intensification of extreme rainfall under climate change has substantially increased flood risk in coastal megacities, particularly in data-scarce regions. Conventional flood risk assessments predominantly rely on hydrometeorological and physical indicators, often neglecting real-time societal response and infrastructure stress. This study develops an integrated framework for flood risk prediction by coupling social media intelligence with urban morphology to capture both physical vulnerability and dynamic public response during extreme rainfall events. The framework is applied to Karachi, Pakistan, using the 2022 monsoon floods as a representative climate extreme. A dataset of 8,732 flood-related georeferenced social media posts was analyzed using natural language processing, sentiment analysis, topic modeling, and machine learning techniques. Public Concern Index and Public Sentiment Index metrics were employed to quantify temporal variations in public response, revealing that flood-related discussions peaked during extreme rainfall days, with Public Concern Index values reaching 17.28

Drought risk assessment and the prediction of transition probability are essential for drought early warning and water resources management. However, studies on spatial transitional properties of drought risk considering the influence of neighboring regions remain limited. Using Hunan Province, China, as a case study, this study addresses this gap by applying Standardized Precipitation Index (SPI) and three drought risk indicators (i.e., vulnerability, exposure, and resilience), combining Markov chain and Moran’s I index. The key findings were: (1) The annual SPI series from 1960 to 2021 exhibited a fluctuating trend, initially increasing, then decreasing, and finally increasing again. Drought-prone areas were mainly clustered in Shaoyang, Hengyang, and the northern parts of Xiangxi and Yongzhou. (2) Approximately 24.6

Study region This study focuses on 1034 near-natural catchments grouped into eight hydroclimatic-physiographic types using climatic, topographic, soil, and hydrological attributes. Study focus Drought-to-flood abrupt alternation (DTF) may not directly translate from meteorological anomalies into streamflow hazards because catchments can filter, delay, or amplify atmospheric signals. Meteorological DTF (M-DTF) and hydrological DTF (H-DTF) events were identified using standardized weighted precipitation and runoff indices. A one-to-one event-matching framework was used to quantify propagation rate, delay, and intensity evolution from M-DTF to H-DTF. We evaluated the statistical associations of these metrics with meteorological forcing and catchment attributes. New hydrological insights for the region M-DTF events were more frequent, abrupt, and severe than H-DTF events, whereas H-DTF events persisted 2.9 times longer on average. Across the analyzed catchments, the mean catchment-level PR within the adopted 90-day matching window was 12%, consistent with substantial catchment filtering. Warm-humid and runoff-responsive catchments had higher propagation rates and shorter delays, whereas high-elevation and snow-dominated catchments had weaker and delayed propagation. Abruptness and severity were attenuated in 70% and 77% of catchments, respectively, although arid, low-runoff catchments sometimes showed greater hydrological severity than their matched meteorological events. These findings provide an empirical basis for regional DTF risk assessment and the future development and evaluation of streamflow-oriented early-warning approaches.

Differentiable modeling techniques have enabled end-to-end calibration of hydrological models and facilitated their integration with data-driven approaches. However, the potential of differentiable parameter learning across diverse climatic conditions remains underexplored, and a systematic evaluation of differentiable module replacement strategies remain limited. To address this, we develop a differentiable parameter learning model (D-HBV) based on the HBV-96 framework, in which Temporal Convolutional Network (TCN) and Convolutional Long Short-Term Memory (ConvLSTM) modules are incorporated to replace the runoff generation and routing components, respectively. Three hybrid models (D-T-M, D-H-C, and D-T-C) are constructed and applied to river basins spanning ten global climate types. Compared with the traditional HBV model, the D-HBV model improves NSE and KGE by 37.3

Study region China. Study focus This study investigated the propagation from soil moisture drought (SD) to vegetation drought (VD) across China during 1981–2020 using a three-dimensional (3-D) spatio-temporal clustering framework. SD and VD events were identified based on the Standardized Soil Moisture Index (SSMI) and Standardized Vegetation Health Index (SVHI), respectively. A spatio-temporal matching algorithm was developed to pair SD and VD events, enabling a quantitative analysis of propagation characteristics (i.e., propagation time, rate, direction, distance, probability, and thresholds). The key contributors were identified using an integrated XGBoost-SHAP approach. New hydrological insights for the region This study identified 144 SD and 59 VD events, among which 24 were successfully matched as propagation events. The average propagation time from SD to VD was 3.76 months, exhibiting a distinct west-to-east gradient. The thresholds for triggering VD increased with drought severity, with SD duration thresholds ranging from 1 to 6 months across most regions. Regional contributors to drought propagation time varied across China: energy-related factors were primary in humid regions, precipitation (PRE) in arid zones, and temperature (TMP) and root-zone soil moisture (SMrz) in semi-arid areas. This study provides a comprehensive framework for understanding multi-dimensional drought propagation and offers scientific support for drought early warning and ecological risk management in China.

Exploring the triggers of drought propagation is essential for understanding drought dynamics. Current research primarily provides probability thresholds for drought propagation based on Copula and Bayesian approaches. However, water resource managers are more interested in determining whether droughts can actually be triggered, rather than solely receiving probabilistic reminders. In this study, we propose a framework for identifying the discriminative model, key factors, and precipitation blocking thresholds that trigger meteorological-to-agricultural drought in the Xijiang River Basin (XRB). The results highlight the influence of non-effective precipitation days (NEPD), meteorological drought duration, the meteorological drought area and its spatial complexity (A_GAM) on triggering propagation. Daily precipitation exceeding 3 mm begins to mitigate drought propagation. Through analyzing 45 actual drought events using 36 models comprising 4 factor combinations and 9 machine learning methods, we found that the GANs-enhanced K-nearest neighbors (KNN) algorithm is the optimal discriminative model. Sensitivity analysis based on the model reveals that a reduction in NEPD (daily precipitation ≤ 3 mm) can decrease the propagation ratio by 11.1

The catchments located at high altitudes are significant sources of freshwater in the downstream societies but are becoming susceptible to the interactive effects of land use and climatic change. This study considers the effect of these control factors on the water yield and runoff dynamics at the Kunhar River Basin in northern Pakistan. This paper simulates past and future hydrological dynamics using the process-based Soil and Water Assessment Tool (SWAT), downscaled CMIP5 climate forecasts, and CA-Markov-based land use projections. The model exhibited high accuracy, with NSE = 0.89; R 2 = 0.75; PBIAS = -2.04% during calibration and NSE = 0.83; R 2 = 0.74; PBIAS = -13.51% during validation. Observations show that cropland convention to forest as well as urbanisation has transformed the runoff dynamics by 15%-30% in terms of total volume. The future projections under the RCP 4.5 and RCP 8.5 scenarios therefore indicate that average annual water yield will decrease by 35%-48%, with seasonal redistribution characterised by increased winter runoff and reduced summer runoff. The results exhibit significant differences in the pattern of streamflow, and such differences impact the well-being of ecosystems, water supply security, and risk levels of floods. In terms of mountainous areas, this study develops a practical framework of climate resilience initiatives and sustainable watershed management.

Floods pose a major socio-economic and environmental challenge in Pakistan, where climate-induced extremes have increasingly heightened the vulnerability of both urban and rural systems. Traditional post-disaster surveys often fail to capture the dynamic emotional and behavioral responses of affected communities in real time. This study employs an interdisciplinary social-media based analytical framework to examine public sentiment, thematic narratives, and risk responses during the 2022 Pakistan floods. A dataset of 156,842 Twitter posts was collected, of which 148,536 tweets remained after preprocessing. Sentiment analysis, Latent Dirichlet Allocation (LDA) topic modeling, and spatial visualization were applied to analyze public discourse. The results indicate that 58% of the tweets expressed negative sentiment, primarily related to loss, governmental criticism, and infrastructure failure, while 15% reflected positive sentiment associated with solidarity and humanitarian relief, and 27% were neutral, focusing mainly on information dissemination. Topic modeling identified eight dominant themes, with relief efforts (32%), property damage (27%), and recovery campaigns (21%) collectively accounting for 80% of the discourse. Spatial analysis revealed that discussion intensity was concentrated in the most flood-affected provinces, including Sindh, Punjab, and Khyber Pakhtunkhwa. By integrating textual and spatial analytics, the study provides insights into community-level perceptions and risk responses during climate-induced flood events. The findings demonstrate the potential of social media data as a complementary tool for disaster assessment, early situational awareness, and flood-risk analysis in climate-prone regions.

Drought is thought of as one of the gravest climate-related hazards of the agro-dependent regions facing water stress like Pakistan where socio-economic stability and food security are under threat because of the increasing hydro-climatic variability. In this paper, a more advanced drought risk forecasting model has been analyzed, which integrates the multi-scale Standardized Precipitation Indices, climatic zoning based on the Köppen-Geiger classification, as well as the deep learning model to predict and identify different drought dynamics in various climatic regions. The spatiotemporal pattern analysis of droughts is considered to be the SPI of 1, 3, 6, 9, and 12-month. Besides that, performance of hybrid model (RNN, LSTM, BiLSTM and CNN-LSTM) has been compared with the old algorithm (SVM and empirical model using Penman-Monteith. The results indicate that the intensity and the length of droughts have been increasing at a high pace within the semi-arid and coastal desert regions of the country. Long-term trends in droughts were best modeled by SPI-9 and SPI-12. The models of deep learning are significantly superior to the baseline methods. The CNN-LSTM would be the best to use in the long-term prediction, whereas BiLSTM appears to be more efficient in the short-term predictions. These outcomes indicate that deep neural networks can learn non-linear climatic dynamics besides providing action on-lead-time drought risk management data. The proposed prediction system also forms the basis of the anticipatory governance in providing information on crop planning, deferral of irrigation, regulation of abstraction of groundwater, and drought-contingency plan. To improve the progress of the climate-risk-intelligence developments, to be incorporated in the policy-making, the strengthening of the national early-warning possibilities and resilience planning can be applied. In the light of an interdisciplinary effort of applying climate diagnostics in predictive analytics, the research offers a scaling route to the utilization of information in both tracking drought and climate-adjusting water management in not only Pakistan but also other disaster-prone regions in South Asia.

Accurate medium-long-term streamflow forecasting is crucial for flood mitigation and water-resource management across the Yangtze River Basin. Single deep-learning approaches remain challenged by non-stationarity, intricate long-range dependencies, and extreme-event sparsity. We propose a framework that integrates empirical mode decomposition (EMD) to decompose daily discharge, temporal convolutional networks (TCN) to extract multi-scale features, and gated recurrent units (GRU) to generate multi-step forecasts, with final outputs obtained via linear recombination, namely EMD-TCN-GRU. When the forecast horizon is set at 3 days, the model—trained on 2013–2022 Wuhan observations—records an R² of 0.9951 and a MAPE of 2.87

The global water crisis is driven by human overconsumption, depleting resources faster than they can regenerate. This threatens ecosystems and exacerbates drought-related water shortages, which may escalate into international conflicts if not properly managed. This study assesses the effectiveness of drought indices in evaluating and managing droughts to determine their role in developing mitigation plans against water scarcity. It also highlights strategies to reduce drought risks and prevent future droughts through proactive approaches. Drought indices measure severity and frequency in various regions to identify patterns of water availability in the most drought-prone areas. In addition to the discussed drought indices, traditional and innovative risk management strategies for drought mitigation will be analysed, particularly focusing on the most vulnerable areas. Communities and mankind are becoming more exposed to meteorological disasters ranging from droughts to flooding attributed to minimal or excess rain. Analysis of multi-dimensional aspects of droughts can lead to better understanding of how human activities, such as mining and deforestation, increase this natural occurrence. By applying this knowledge, we can avoid practices that overheighten the effects of dry weather conditions. The growing human population means new approaches are necessary for food sustainability and water security that balance with environmental protection and natural resource conservation. The study concludes that large-scale engineering projects for drought relief may harm ecosystems and agricultural land, advocating for sustainable, eco-friendly solutions. It emphasises the need for improved water management and conservation efforts, prioritising natural water replenishment over artificial interventions. The findings advocate for sustainable, nature-based water management to reduce ecological damage, enhance long-term resilience, and emphasise global cooperation to prevent conflicts, linking drought management with ecosystem sustainability.

As warmer temperatures enhance atmospheric moisture, hydrological droughts tend to intensify in most regions of the globe. Consequently, younger generations are expected to face a more severe risk of hydrological drought during their lifetimes, emphasizing the critical issue of intergenerational inequity due to climate change. To quantify exposure to hydrological drought across generations, we constructed a cascade model chain for drought simulation using hybrid terrestrial models, based on 5 GCM outputs under SSP5-85, five hydrological models and a deep learning model. We then projected future univariate and bivariate hydrological drought evolution in 4091 river basins, and quantified lifetime exposure to drought for the age groups born in 2020 and 1960. Drought severity and duration are projected to increase substantially in the Eastern America, Southern Brazil and Western Europe, over 79 % of basins. Extreme droughts far beyond historical records are expected to become more frequent and impact Western Europe in particular. Of note, the exposure of the different age groups to hydrological drought shows a notable disequilibrium. Exposure of people born in 2020 to hydrological drought hazards is projected to increase by 12 % over the late 21st century compared to those born in 1960, indicating that the acceleration of climate change is expected to increase the lifetime risk of future generations. The exposure factor of the newborns is 1.4 times higher than that of 80 years of age under warming condition. Our findings underscore that future drought conditions under extreme warming pose a significant threat to the living conditions of younger generations.

Accurate calculation of downstream water level during non-spilling periods is crucial for the safe operation of hydropower stations. Inaccurate water level calculations can lead to frequent modifications of power generation plans, seriously affecting the safe, stable, and efficient operation of hydropower stations. Currently, the significant errors in calculating downstream water level during non-spilling periods urgently need to be addressed. A multiple linear regression calculation method based on decision tree classification is proposed. This method first employs the classification and regression tree (CART) algorithm, using long-series measured data from hydropower stations to establish decision tree classification rules and classify data of different operation working conditions. An independent multiple linear regression method is established for the numerical fitting of every data classification, which avoids the problem that it is hard for a single model to describe the overall data characteristics. When the method is applied to calculate downstream water level during non-spilling periods at the Three Gorges Hydropower Station, the results show that the calculated downstream water level are closer to the actual downstream water level process compared to the function fitting method, which improves the calculation accuracy of downstream water level. The research findings can accurately calculate downstream water level fluctuations, reduce the impact of downstream water level on the operating head, and provide technical support for improving the refined scheduling level of hydropower stations.

The extensive and gradual onset of drought prompts critical examination of the alterations and engagement among substantial demographics during the drought's advancement and the consequent effects of such shifts on drought detection. This research examines Fb data from 2016 to 2024 to investigate the role of online engagement in drought management. This study evaluates public discourse on drought-related matters through the analysis of five fundamental terms. The research employs topic modeling and sentiment analysis to assess regional awareness and utilizes machine learning techniques (Random Forest, Naive Bayes) in conjunction with Bag of Words to forecast drought progression. The research highlights the potential of Fb data in facilitating real-time drought management, offering significant hydrological insights. The study elucidates regional disparities in drought awareness through the examination of key terminology and sentiment, revealing that some regions exhibit a more rapid reaction to water scarcity, as indicated by Fb engagement. Furthermore, the incorporation of machine learning algorithms such as Random Forest and Naive Bayes facilitates a predictive paradigm for detecting prospective drought hotspots through online discourse analysis. The study confirmed that participation in online communities successfully (p ≤ 0.04) alleviates the impact of drought and, on Facebook, significantly enhanced drought awareness across various regions of Pakistan (p ≤ 0.5), as confirmed through statistical analysis with a paired t-test and regression analysis over labeled sentiment and topic-classified data. Fb engagement may function as a proactive indicator, assisting policymakers and hydrologists in optimizing water resource allocation in drought-prone areas, thus enhancing drought mitigation strategies.

Air pollution in cities is still a major problem for developing nations, worsened by growing industrialization and urbanization. The research used sentiment analysis techniques for social media data to assess the opinions of the general public concerning air quality (AQ). The research used sentiment analysis techniques on social media data to assess the perspectives of the general populace concerning air quality. This research examines unstructured information gathered from Facebook, aiming to assess emotional reactions linked to air quality. The study correlates with environmental factors such as industrial emissions, traffic congestion, and seasonal agricultural burning. It leverages social media as data source for the novel assessment of air quality in real-time, particularly in areas with no or poor monitoring systems. Regression analysis showed a significant positive association between PM2.5 concentrations and negative sentiment (β = 0.42, p < 0.01), underlining the health risks linked to poor air quality. The findings with pollution metrics could formulate the urban policy, enhance public cognizance, and foster civic participation in environmental governance for developing regions.

Study region: This study aims at the Kunhar River Basin, Pakistan, that has been facing repeated flood occurrences on a recurring basis. As the flood susceptibility of this area is high, its topographic complexity demands correct predictive modeling for strategic flood planning. Study focus: We developed a system of flood susceptibility mapping based on Geographic Information Systems (GIS), Principal Component Analysis (PCA), and Support Vector Machine (SVM) classification. Four kernel functions were applied, and the highest-performing was the Radial Basis Function (SVM-RBF). The model was validated and trained using historical flood inventories, morphometric parameters, and hydrologic variables, and feature dimensionality was reduced via PCA for increased efficiency. New hydrological insights: The SVM-RBF model recorded an AUC of 0.8341, 88.02% success, 84.97% predictability, 0.89 Kappa value, and F1-score of 0.86, all of which indicated high predictability. Error analysis yielded a PBIAS of +2.14%, indicating negligible overestimation bias but within limits acceptable in hydrological modeling. The results support the superiority of the SVM-RBF approach compared to conventional bivariate methods in modeling flood susceptibility over the complex terrain of mountains. The results can be applied in guiding evidence-based flood mitigation, land-use planning, and adaptive management in the Kunhar River Basin.

Climate change alters river runoff regimes, affecting the safe operation of hydropower stations. This study proposed an optimization scheduling and risk analysis framework for cascade hydropower under climate change using the Qingjiang cascade hydropower stations as a case study. The framework has three stages. Firstly, a hydrological model coupling GCMs with SWAT under CMIP5 scenarios is established to predict future runoff. Secondly, cascade hydropower optimization scheduling under climate change is performed using the POA (Progressive Optimization Algorithm). Thirdly, a risk assessment index system is established, including risks of insufficient power generation, insufficient output, and water abandonment. The POMR (Probability Optimization Method for the Risk) is applied to calculates power scheduling risks. Results show that the simulated annual average runoff at Changyang Station increases by 6.0, 8.7, and 13.2% under the RCP2.6, RCP4.5, and RCP8.5 scenarios, respectively. Annual power generation for the Qingjiang cascade is projected to rise by 6.2-16.5%, with increases of 5.2-12.9% during flood seasons and 7.5-19.9% in non-flood seasons. Comprehensive risk rates decline to 0.1767, 0.1706, and 0.1630 across the scenarios. This research provides scientific and technical support for managing water resources and operating the Qingjiang cascade under climate change.

Study region: Pakistan. Study focus: Enhanced social media data accessibility facilitates the effective detection of disaster-associated information, exemplified by urban heatwaves. Social media elucidates human behavior and geographical trends in Sindh Province, Pakistan, amidst intensifying extreme heat events. The utilization of this data facilitates a comprehensive examination of heatwaves, thereby enhancing the real-time management and mitigation strategies within urban environments. Innovative insights: This study introduces an innovative framework for the real-time extraction and analysis of heatwave-related data from Facebook in areas impacted by disasters. This framework employs topic modeling and transfer learning for examining heatwave disaster dynamics. It extracts temperature and humidity information from both textual content and visual representation using state-of-the-art deep learning techniques and fuses these insights via decision-making processes. A fine-grained location corpus specific to urban heatwaves was developed using a named entity recognition (NER) model. The BERT-BiLSTM-CRF model was then utilized to extract heatwave data accurately. In the case of southeastern Pakistan, the extracted heatwave points aligned with 86 % of officially documented incidents, predominantly occurring near urban areas. This framework improves situational awareness and supports real-time spatiotemporal analyses of the urban heatwaves, thus providing a new approach for effective disaster management and response.