Abstract Discrete wavelet (DW) analysis is commonly employed to decompose raw observed data into stochastic and deterministic components, aiming to improve model accuracy. Nonetheless, the DW technique is recognized for its significant limitations, often leading to inaccurate predictions of daily rainfall. In this study, a dynamic adaptive wavelet algorithm is introduced to decompose the raw observed data into components, aiming to address the drawbacks associated with DW analysis. The ultimate goal is to enable the estimation of daily rainfall with enhanced accuracy and precision in missing value imputation. The decomposed stochastic and deterministic components derived from the maximum overlap discrete wavelet (MODW) and DW techniques are incorporated as inputs in compressed sensing (CS) framework. This procedure facilitates the creation of novel methodologies MODW-based compressed sensing (MODW-CS) and DW-based compressed sensing (DW-CS) for daily rainfall prediction. These methodologies are deployed across scenarios utilizing 50%, 70%, and 90% of the available data. Twenty-two years of validated daily rainfall data were collected from five stations situated within the Euphrates-Tigris Basin. The novel methodologies were assessed against the CS model, with root-mean squared error (RMSE), mean absolute error (MAE), and coefficient of efficiency (CE) being taken into account. Findings revealed that both the MODW-CS and DW-CS models outperformed the CS model, with enhanced prediction accuracy observed through the utilization of decomposed signals. However, the MODW-CS model exhibited superior performance compared with the DW-CS model for all scenarios. This suggests that the MODW algorithm possesses a dynamic adaptive structure enabling it to capture the natural structure of observed data more effectively than DW analysis. The findings of this study have significant implications across various domains, including hydrology, agriculture, flood control, drought management, and weather forecasting. Accurate predictions of missing data points in time series are essential for better understanding and forecasting rainfall event behavior.
Abstract Effective management of dam spillway water releases is crucial for mitigating flood hazards in snow-fed basins. This study examines the April 2024 flood event that severely impacted Uralsk city and its surrounding areas in Kazakhstan. Using hydrological and hydraulic modeling, this study evaluates the role of upstream dam spillway releases in Russia as a primary driver of the flood event. Hydrological modeling was conducted using the HBV-light model, incorporating calibrated and validated streamflow data from ground stations. Snow water equivalent (SWE) and snow cover area (SCA) simulations were validated using ERA5-Land products. The flood extent and water depths were determined using the HEC-RAS two-dimensional (2D) hydraulic model, validated against satellite and drone imagery. Scenario-based analyses compare flood conditions with and without dam reservoir water contributions. The results indicate that an abrupt spillway release with a peak discharge of 2,170 m 3 / s increased the simulated flood inundation area by approximately 10% and locally increased flood depths by up to 1 m. These findings underscore the importance of proactive dam management in snow-fed basins to minimize flood risk. This study provides insights into the critical relationship between water resource management and flood dynamics, offering valuable guidance for sustainable development and flood risk mitigation in snow-covered regions.
Abstract Nonstationarity in extreme rainfall, driven by an intensifying hydrological cycle under a changing climate, challenges the reliability of conventional stationary intensity–duration–frequency (IDF) curves used in hydrologic design. Analyzing annual and seasonal timeframes, we employed innovative trend analysis to detect trends in maximum rainfall over durations of 1 to 48 h. A set of 10 covariate combinations, resulting in 9,225 models per location and timeframe, was used. Nonstationarity is initially introduced in the location parameter and subsequently in the location and scale parameters of the generalized extreme value distribution. Annual trends indicate an increase in 24 h rainfall in northern and central regions, contrasted by decreases in coastal southern and northern areas. The southwest (SW) monsoon trends indicate increases in the north and southern peninsular regions, with decreases in the north and west. In contrast, the northeast (NE) monsoon trends show increases in the north, east, and south as well as decreases in the southern peninsular areas. Nonstationarity predominantly affects the generalized extreme value location parameter in the western, southern coastal, and southern peninsular regions during the southwest monsoon and in the central and southern peninsular regions during the northeast monsoon. The eastern, southern peninsular, and central areas, traditionally dominated by the NE monsoon, exhibit more extreme conditions during the SW monsoon. In contrast, the south coastal, central, and northern regions, typically driven by the SW monsoon, show increased extremes of the NE monsoon. Annual nonstationary rainfall intensities increase up to 58.4% (Dehradun) and 72.9% (Varanasi); SW monsoon rises reach 37.5%–50.9%, while NE monsoon increases range 20.3%–42.2%. The global temperature anomaly, El Niño–southern oscillation phase changes, and regional temperature change are identified as significant covariates, with temporal nonstationarity limited to northern regions for specific durations. These findings underscore the need for region-specific IDF curves to address the evolving patterns of extreme rainfall under climate change.
Abstract The choice of minimum interevent time (MIT)—a criterion used to discern independent events from a continuous rainfall record—can affect the resulting partial duration series (PDS) and depth-duration-frequency values (DDFs). However, the mechanisms by which MIT choice affects regional DDFs—and how they are modulated by the choice of regional frequency analysis (RFA) framework—remain unclear. Here, we investigated these effects for short-duration ( ≤ 2 h ) precipitation, over a range of MITs (1 to 8 h) and average recurrence intervals (ARIs between 0.75 and 100 years), by applying two standard RFA frameworks—the station-year and the index-flood methods—to four study areas spanning diverse geoclimates and rain gauge densities. We found a robust preservation of tail-event identity for all regions: the magnitudes and ranks of the greatest exceedances that govern tail-events probabilities remain invariant to MIT. Shorter MITs increase the proportion of large exceedances in the PDS, and when combined with invariant tail behavior, produce a systematic sensitivity: MIT effects are most pronounced at low ARIs ( ≤ 2 years ), with smaller MIT values yielding larger DDFs. The index-flood method propagates station-level sensitivity to the regional scale through its index variable, whereas the station-year method attenuates it via spatial pooling. MIT choice introduces only modest variability in regional DDFs (for ARIs ≤ 25 years )—often within ± 1 % for most sites across all study regions—relative to dominant uncertainties such as sampling variability. Nonetheless, MIT is important for identifying independent events; in practice, a reasonably selected MIT combined with an appropriate RFA technique results in negligible effects on regional DDFs.
Abstract Accurate monthly rainfall prediction is essential for water-resource planning, agricultural management, and flood-hazard mitigation, particularly in rapidly urbanizing tropical cities. This study presents a hybrid wavelet-based ARIMA model (HWAM) that couples discrete wavelet decomposition with individually optimized autoregressive integrated moving average (ARIMA) models to forecast monthly rainfall in Bangalore, India. The Daubechies wavelet of order 6 at decomposition level 4 is applied to a 123-year record (1901–2023) to separate the signal into one approximation subseries and four detail subseries, each of which is independently modeled by an appropriate ARIMA process. Forecasts are reconstructed by summing the component predictions. The HWAM is trained on 80% of the data (1901–1998) and validated on the remaining 20% (1999–2023). On the validation set, the HWAM achieves a mean absolute error (MAE) of 0.221 mm, a root-mean-square error (RMSE) of 0.662 mm, a mean absolute percentage error (MAPE) of 1.121%, and a coefficient of determination ( R 2 ) of 0.998. To demonstrate the comparative performance of the HWAM, its results are benchmarked against four state-of-the-art machine-learning and deep learning models: Transformer, bidirectional long short-term memory (BiLSTM), convolutional neural network–bidirectional LSTM (CNN-BiLSTM), and extreme gradient boosting (XGBoost). The HWAM achieves the lowest MAPE (1.121%) and highest R 2 (0.998) among all five models, outperforming the standalone ARIMA baseline by approximately 85–95% and surpassing all machine-learning benchmarks across every error metric under the same univariate forecasting setting. The novelty of this work lies in the systematic selection and optimization of wavelet parameters for long-record monthly rainfall prediction in a monsoon-dominated urban watershed, comprehensive benchmarking against contemporary deep-learning models, and the generation of reliable 3-year-ahead monthly forecasts that can directly support urban infrastructure planning and climate-adaptation decisions in Bangalore.
Abstract Rainfall received across the India during monsoon season is very crucial for the country’s economy. Predicting monthly rainfall plays a crucial role in efficient water resources management and in reducing the risk of hydrological disasters. Looking to the drawback of general circulation models (GCMs), it is better to assess hydroclimatic teleconnections (HCTs) of monthly Indian summer monsoon rainfall (ISMR) for its prediction. The climate system is dynamic and nonlinear. These nonlinearities are not well represented by conventional linear statistical techniques. Machine learning (ML) provides in-depth understanding of complex nonlinear data structures. To the best of the authors’ knowledge, no earlier study has performed an assessment of HCTs and prediction of monthly ISMR by formulating robust ML models by employing eighteen circulation indices, which is performed in the present study. Assessment of input significance and input independence, systematic data division in such a way that statistical parameters of subsets and all data are similar, followed by use of Bayesian optimization for development of ML model, are necessary for formulating robust ML models and it is not performed by any earlier study according to the authors’ best knowledge. Therefore, it is performed in the present study. Thus, the present study employed four ML techniques, namely, adaptive boosting, extreme gradient boosting regression, random forest (RF) regression, and support vector machine for assessment of aforesaid HCTs. The aforementioned analysis is performed for two periods viz. 1951–2000 and 1951–2014 to investigate the alteration of aforesaid HCTs along with time. It is observed that HCTs of monthly ISMR vary with time. The present study also showed that all the formulated ML models are robust and do not show any overfitting/underfitting. It is also observed that the performance of the RF technique is better as compared to the other three ML techniques.
Abstract The performance of weather radar nowcast for use in urban hydrology is often evaluated based on observed weather radar rainfall. This may not, however, accurately reflect the performance of urban hydrological rainfall–runoff predictions that rely on weather radar nowcasts. This study highlights the significant role of both spatial scale and evaluation methodology in assessing the performance of radar-based rainfall nowcasts for urban hydrological applications. We investigate whether the weather radar nowcast performance at mesoscales corresponds to that at smaller urban catchment scales and if the weather radar observation-based evaluation of nowcast performance aligns with the hydrological runoff prediction performance. We apply the COTREC nowcasting technique in both a small Danish catchment and a larger spatial domain, which corresponds to a mesoscale, and provide evidence that nowcasting performance at the mesoscale is not indicative of its performance at the catchment scale. When mesoscale representativeness is assumed for event-accumulated rainfall depths, nowcast performance at the catchment scale tends to be overestimated during low-intensity rainfall events and underestimated during high-intensity events. Conversely, time series evaluations show less pronounced differences across spatial scales. By comparing weather radar observation-based nowcast evaluations with hydrologically based evaluations, the performance of hydrological predictions is further demonstrated to differ from that of weather radar rainfall predictions. For lead times ≤ 10 min , the weather radar observation-based evaluation overestimates the hydrological prediction skill. However, the hydrological prediction skill is underestimated during high-intensity rainfall events if weather radar observation-based evaluations are assumed representative when evaluating time series.
Abstract Based on multisource urban flood data from Yangpu District during 2025, this study develops a hierarchical sequential decision-making framework that decouples risk detection and severity assessment through a dual-output long short-term memory (LSTM) architecture. Compared to conventional multiclass LSTMs, the framework achieves superior overall accuracy (99.77% versus 99.51%) and F1-score (0.8818 versus 0.4503), with a significantly lower false alarm rate (0.02% versus 0.06%). The framework demonstrates strong warning detection capability (62.92% exact level matching) and perfect performance on high-level warnings (Levels II and III). Compared to 79.78%, the underwarning rate of 37.08% indicates the framework’s effectiveness in alleviating the underwarning issue.
Abstract Water quality in a watershed is influenced by both climatic variability and human-induced changes. Best management practices (BMPs) are targeted strategies designed to improve water quality within watersheds and river basins. This study focuses on identifying critical areas for sediment and nutrient loading under projected climate and anthropogenic changes in the Rengali catchment of the Brahmani River basin in India. The effectiveness of three BMPs was assessed using the Soil and Water Assessment Tool (SWAT). Future land use and land cover changes were projected using three hybrid machine learning models: artificial neural network with cellular automata (ANN-CA), multilayer perceptron with cellular Automata and Markov chain (MLP-CA-MC), and logistic regression with cellular Automata and Markov chain (LR-CA-MC). Climate change projections were obtained from 10 general circulation models (GCMs) under the CMIP6 framework, with the most suitable models for the region selected using the compromise programming method. The three BMPs—filter strips, stone/soil bunds, and contouring—were applied in erosion-prone areas, and their effectiveness in reducing sediment and nutrient loads was evaluated at the subbasin scale. The study estimated the annual sediment yield, nitrogen loading, and phosphorus loading to be 1,449.02 t / ha / year , 925.52 kg / ha / year , and 1,141 kg / ha / year , respectively. Based on these findings, seven subbasins were identified as erosion hotspots due to their high levels of sediment and nutrient export. Implementation of best management practices resulted in notable reductions in sediment and nutrient loads. Stone or soil bunds reduced sediment by 48.70%, nitrogen by 41.01%, and phosphorus by 46.26%. Filter strips achieved reductions of 31.96% for sediment, 32% for nitrogen, and 34.02% for phosphorus. In comparison, contouring showed minimal impact relative to filter strips. These findings can inform the selection of appropriate BMPs for field-scale water resource management, contributing to the reduction of sediment and nutrient concentrations in surface waters.
Abstract Infiltration-based low-impact development (LID) practices are widely implemented to mitigate urban runoff impacts; however, their design is commonly guided by volumetric water-quality criteria that obscure critical trade-offs among hydrologic performance, geometry, soil properties, and cost. This study introduces the LID-design atlas (LIDA), a physics-based design database that systematically maps hydrologic performance and life cycle cost trade-offs across a broad space of infiltration-based LID configurations. LIDA is populated using thousands of numerical simulations generated with the Darcy–Richards analysis of infiltration in nature-based low-impact development (DRAIN-LID) model, which resolves one-dimensional variably saturated vertical flow through engineered media using a mixed-form Darcy–Richards formulation coupled with synthetic runoff inflow hydrographs. Three numerical design scenarios, spanning residential and roadside retrofit applications, are used to demonstrate how LIDA enables rapid screening of feasible designs based on peak-flow reduction, time-to-peak delay, detention time, ponding depth, and cost constraints. The results reveal strong threshold behavior in system response, with feasible design regions governed by interactions among footprint, media depth, and soil hydraulic conductivity rather than by storage volume alone. In space-constrained settings, increases in media depth are shown to be a more cost-efficient lever than footprint expansion, whereas lower-permeability media substantially restrict feasible design space. By explicitly resolving the temporal redistribution of runoff and exposing cost–performance trade-offs that are not captured by volume-based sizing methods, LIDA provides a transparent and computationally efficient framework for early-stage screening and comparison of infiltration-based LID designs.
Abstract Remote sensing precipitation products (RSPPs) are indispensable for hydrological modeling in regions with sparse and uneven rain gauge coverage, particularly in complex terrains. Their ability to deliver spatially continuous, high-resolution rainfall estimates enables improved runoff simulations, flood forecasting, and water resource assessments across heterogeneous hydroclimatic regimes. This study quantifies the manifestation of variability in using different RSPPs (viz., IMERG, MSWEP, GPCP, and PERSIANN) to simulate streamflow using rainfall-runoff model (HYSIM). The study encompasses diverse Indian watersheds—including Halia (Krishna), Lowara (Saurashtra), Mannot (Narmada), Muri (Subernarekha), and Saklespur (Cauvery)—to capture broad spatial, climatic, and topographic variability for a comprehensive assessment of RSPPs. Results demonstrate superior performance by the IMERG Final Run product over orographically influenced watersheds (e.g., Halia and Saklespur) to estimate high-intensity, short-duration rainfall regimes. In contrast, MSWEP exhibited higher skill in semiarid and large river watersheds (e.g., Krishna, Narmada, Saurashtra) with more homogeneous rainfall distributions, benefitting from its multisource integration of satellite retrievals, in situ gauge records, and reanalysis products. Both IMERG and MSWEP achieved satisfactory model performance ( N S E > 0.5 , R 2 > 0.6 ) in Halia, Lowara, Mannot, and Muri watersheds, while in the high-relief, high-rainfall Sakleshpur watershed, only IMERG maintained acceptable accuracy. The results underscore the necessity of precipitation product selection based on watershed-specific physiographic and climatic characteristics, and reaffirm that RSPPs can serve as a robust alternative to gauge-based rainfall data for operational hydrology, enabling enhanced predictive capability in flood forecasting, water resources planning, and watershed management.
