Explore the vast range of titles available.

Determining rainfall runoff responses of catchments to unprecedented climate conditions is an issue which has largely eluded the hydrologic community for many years. Conceptual rainfall runoff models are used globally to predict runoff for regional water resources management and planning. However, obtaining parameter values suitable for future climate conditions requires approaches that consider conditions beyond historical periods. This paper takes advantage of data from 207 Australian catchments to determine model parameters that most closely produce expected rainfall runoff coefficients (ratio of runoff to rainfall) for a wide range of environmental conditions. This was done for two popular rainfall runoff models, GR4J and Sacramento. In a two‐step process, parameters were first selected that could adequately reproduce observed runoff coefficients across the 207 catchments. Acceptable parameter sets were stored in a library from which, in the second step, parameters were selected for each individual catchment according to various goodness‐of‐fit metrics. Performance of this calibration approach was compared with a classical optimization employed for each catchment (DELO—Differential Evolution Local Optimization). The study found performance trade‐offs using the parameter library based calibration compared to DELO for metrics such as Nash‐Sutcliffe Efficiency and percentage bias. The library‐based calibration exhibited behavior that more closely aligned with expectations under perturbed climate conditions, compared to DELO parameters. Results also showed tolerable estimates of rainfall runoff coefficient using DELO parameters at many sites when rainfall is reduced by no more than 25%. However, there is a high risk of under‐ or over‐estimating runoff coefficients at larger reductions.

Accurate short-term rainfall–runoff prediction is essential for flood mitigation and safety of hydraulic structures and infrastructures. This study investigates the capability of four machine learning methods (MLM), optimal pruning extreme learning machine (OPELM), multivariate adaptive regression spline (MARS), M5 model tree (M5Tree, and hybridized MARS and Kmeans algorithm (MARS-Kmeans), in hourly rainfall–runoff modeling (considering 1-, 6- and 12-h horizons). Their results are compared with a conceptual method, Event-Based Approach for Small and Ungauged Basins (EBA4SUB) and multi-linear regression (MLR). Hourly rainfall and runoff data gathered from Ilme River watershed, Germany, were divided into two equal parts, and MLM were validated considering each part by swapping training and testing datasets. MLM were compared with EBA4SUB using four events and with respect to three statistics, root-mean-square errors (RMSE), mean absolute error (MAE) and Nash–Sutcliffe efficiency (NSE). Comparison results revealed that the newly developed hybridized MARS-Kmeans method performed superior to the OPELM, MARS, M5Tree and MLR methods in prediction of 1-, 6- and 12-h ahead runoff. Comparison with conceptual method showed that all the machine learning models outperformed the EBA4SUB and OPELM provided slightly better performance than the other three alternatives in event-based rainfall–runoff modeling.Graphic abstract

Conceptual rainfall‐runoff models are widely used for hydrological applications, yet many fail to capture multi‐annual storage dynamics that are critical during prolonged droughts or shifting climate conditions. This study investigates whether specific structural components enable models to represent such long‐term behavior. We hypothesize that three components are necessary: (a) a store responsible for long‐term behavior, (b) disconnection of that store from direct streamflow generation, and (c) a water loss mechanism from that store (e.g., evapotranspiration or groundwater export) that allows gradual depletion. We systematically tested this hypothesis using 46 daily models from the Modular Assessment of Rainfall‐Runoff Models Toolbox. Each model was tested in three Australian headwater catchments that have previously been identified as subject to multi‐annual declines in storage during the “Millennium” Drought (1997–2010). In addition, we included a synthetic scenario designed to evaluate delayed streamflow recovery after an idealized multi‐annual drought. Models with all three structural components were significantly more successful: 5 of these 9 models passed both tests, compared to much lower success rates (1 model) among the 37 without the full structure. However, structure alone was not sufficient, as some models with the hypothesized structure failed due to restrictive internal formulations or calibration outcomes that prevented activation of the intended slow‐storage processes. These findings provide guidance for identifying or adapting conceptual models in applications where hydrological memory and long‐term drought response are important.

Deep learning (DL) models such as Long‐Short‐Term‐Memory (LSTM) networks have achieved exceptional predictive accuracy in rainfall–runoff modeling. Yet these models learn from statistical correlations rather than hydrologic insights, raising the question of whether their internal functional reasoning is physically reliable. Despite previous studies highlighting unexpected outcomes from LSTMs under long‐term climate shifts, functional realism—defined as the extent to which a model's internal functioning aligns with defensible mechanisms of streamflow generation—remains largely underexplored. We introduce a hydrology‐specific Explainable AI (XAI) framework that opens the black‐box of LSTM. It extracts nonlinear, lag‐dependent, and time‐varying Impulse Response Functions (IRFs) which quantify the functional relationships that LSTM uses to reflect the isolated influence of precipitation (P), temperature (T), and potential evapotranspiration (PET) on simulated streamflow. IRFs reveal how LSTMs internalize streamflow generation during events, offering a catchment hydrology perspective for evaluating model realism. Applying this framework to 672 North American catchments with strong LSTM predictive skill, we find that high accuracy often masks hydrologically implausible reasoning: in over 70% of rain‐dominated basins, short‐term temperature rises unexpectedly raise simulated streamflow and enhance celerity rate even without rainfall; in snow‐dominated regions, PET is misattributed as a driver of snowmelt‐related flow and enhances the catchment's celerity rate. We conclude that correlation‐driven learning can compromise the robustness of LSTM‐based forecasts under weather extremes and short‐term and long‐term climatic shifts. Our framework bridges deep learning with hydrologic understanding and offers a scalable diagnostic for assessing the functional realism of DL models across diverse catchment types.

Hydrological processes such as rainfall–runoff generation are inherently complex and often difficult to teach, particularly to students without prior background or interest. To address this challenge, we developed the gamified rainfall–runoff model SplashTune, using the educational programming language Scratch to support intuitive and exploratory learning of watershed hydrology (). The model visualizes rainfall, infiltration, surface and subsurface flow, and runoff to river through particle‐based animation, emphasizing visual clarity and interactivity. Players manipulate land surface conditions, including land cover and soil moisture, to match simulated hydrographs with predefined targets, mimicking parameter tuning in hydrological modeling. A scoring system based on Nash–Sutcliffe efficiency provides immediate feedback and encourages repeated trial‐and‐error exploration. A workshop with high school students in Japan demonstrated the model's educational effectiveness. Pre‐ and post‐questionnaires revealed notable gains in understanding of both basic and complex hydrological concepts. Game‐based interaction was particularly effective in enhancing quantitative understanding, such as peak timing and soil moisture effects, compared to traditional lecture‐style instruction. Participants also reported that the gamified model made learning more enjoyable and memorable, especially due to its visual clarity, interactivity, and score‐based feedback. This approach fosters the development and communication of perceptual models of hydrological processes, enabling learners to refine their understanding through interactive simulation. Beyond classroom use, such tools offer potential to promote shared hydrological literacy and engagement across educational, policy, and public domains. Our findings highlight the value of combining scientific modeling with game‐based learning to support both conceptual understanding and inclusive participation.

The soil moisture is a quantity of paramount importance in the study of hydrologic phenomena and soil‐atmosphere interaction. Because of its high spatial and temporal variability, the soil moisture monitoring scheme was investigated here both for soil moisture retrieval by remote sensing and in view of the use of soil moisture data in rainfall‐runoff modeling. To this end, by using a portable Time Domain Reflectometer, a sequence of 35 measurement days were carried out within a single year in seven fields located inside the Vallaccia catchment, central Italy, with area of 60 km2. Every sampling day, soil moisture measurements were collected at each field over a regular grid with an extension of 2000 m2. The optimization of the monitoring scheme, with the aim of an accurate mean soil moisture estimation at the field and catchment scale, was addressed by the statistical and the temporal stability. At the field scale, the number of required samples (NRS) to estimate the field‐mean soil moisture within an accuracy of 2%, necessary for the validation of remotely sensed soil moisture, ranged between 4 and 15 for almost dry conditions (the worst case); at the catchment scale, this number increased to nearly 40 and it refers to almost wet conditions. On the other hand, to estimate the mean soil moisture temporal pattern, useful for rainfall‐runoff modeling, the NRS was found to be lower. In fact, at the catchment scale only 10 measurements collected in the most “representative” field, previously determined through the temporal stability analysis, can reproduce the catchment‐mean soil moisture with a determination coefficient, R2, higher than 0.96 and a root‐mean‐square error, RMSE, equal to 2.38%. For the “nonrepresentative” fields the accuracy in terms of RMSE decreased, but similar R2 coefficients were found. This insight can be exploited for the sampling in a generic field when it is sufficient to know an index of soil moisture temporal pattern to be incorporated in conceptual rainfall‐runoff models. The obtained results can address the soil moisture monitoring network design from which a reliable soil moisture temporal pattern at the catchment scale can be derived.

Continuous simulation for design flood estimation is increasingly becoming a viable alternative to traditional event‐based methods. The advantage of continuous simulation approaches is that the catchment moisture state prior to the flood‐producing rainfall event is implicitly incorporated within the modeling framework, provided the model has been calibrated and validated to produce reasonable simulations. This contrasts with event‐based models in which both information about the expected sequence of rainfall and evaporation preceding the flood‐producing rainfall event, as well as catchment storage and infiltration properties, are commonly pooled together into a single set of “loss” parameters which require adjustment through the process of calibration. To identify the importance of accounting for antecedent moisture in flood modeling, this paper uses a continuous rainfall‐runoff model calibrated to 45 catchments in the Murray‐Darling Basin in Australia. Flood peaks derived using the historical daily rainfall record are compared with those derived using resampled daily rainfall, for which the sequencing of wet and dry days preceding the heavy rainfall event is removed. The analysis shows that there is a consistent underestimation of the design flood events when antecedent moisture is not properly simulated, which can be as much as 30% when only 1 or 2 days of antecedent rainfall are considered, compared to 5% when this is extended to 60 days of prior rainfall. These results show that, in general, it is necessary to consider both short‐term memory in rainfall associated with synoptic scale dependence, as well as longer‐term memory at seasonal or longer time scale variability in order to obtain accurate design flood estimates. Key Points Short‐ and long‐term memory in precipitation is important for flood modelling There are advantages to using a continuous rainfall‐runoff modelling framework Influence of catchment and climate‐specific factors on antecedent precipitation

Flash floods frequently occur in steep mountainous catchments, where topographic gradients play a crucial role in concentrating and accelerating runoff processes. Two‐dimensional (2D) shallow water equations (SWE) are widely used to simulate such events; however, these equations are classically derived assuming small bottom slopes, typically less than 10%. This assumption, seldom questioned, may introduce inaccuracies when SWE models are applied to mountain watersheds, where slopes commonly exceed the theoretical range of validity. This study addresses this inconsistency by revisiting the theoretical foundations of SWE‐based modeling and extending the recently proposed 2D Steep‐Slope Shallow Water Equations (SSSWE) to overland flow. The formulation explicitly accounts for steep‐slope geometric effects, offering a more physically consistent description of flow dynamics in mountainous terrain. A kinematic‐wave‐based reference framework in one‐ and two‐dimensional domains is presented to quantify errors arising from the use of conventional SWE under the small‐angle approximation on steep slopes. Results from idealized test cases (including a V‐shaped catchment) show that the standard SWE predicts lower equilibrium water depths and response times and higher flow velocities than the improved steep‐slope model. These discrepancies become substantial for slopes steeper than 20°. To incorporate steep‐slope effects without modifying existing SWE‐based numerical solvers, an analytically derived, slope‐dependent adjustment to the Manning roughness coefficient is proposed; in 2D, the correction is anisotropic. Overall, this study provides the first systematic analytical assessment of inaccuracies in SWE‐based flash‐flood modeling for steep watersheds, addressing a long‐standing theoretical gap and improving the physical realism of hydrodynamic models in mountainous environments.

In forested, seasonally dry watersheds, winter rains commonly replenish water storage deficits in the vadose zone before recharging underlying hillslope groundwater systems that sustain streamflow. However, the relative inaccessibility of the subsurface limits our understanding of how groundwater recharge is moderated by vadose zone storage deficits generated by plant‐water uptake. Here, we compare groundwater recharge inferred from the storage‐discharge relationship with independent, distributed estimates of deficits across 12 undisturbed California watersheds. We find accrued dry season deficits primarily driven by evapotranspiration insufficiently explain inter‐annual variability in the amount of precipitation required to generate groundwater recharge due to continued deficit accumulation between wet season storms. Tracking the deficit at the storm event‐scale, however, reveals a characteristic response in groundwater to increasing rainfall not captured in the seasonal analysis that may improve estimates of the rainfall required to generate recharge and streamflow on a per‐storm basis. Our findings demonstrate the potential for existing public data sets to better capture water partitioning within the subsurface and thus improve the prediction of rainfall‐runoff behavior and summer water availability in rainfall‐dominated, seasonally dry basins using a combined deficit‐recharge approach.

This paper discusses the notion of similarity often used in the regionalization studies of hydrological models. We compare two different visions of similarity: the apparent similarity defined on the basis of observable catchment properties, and behavioral similarity judged through the use of hydrological models. These two visions are generally assumed to be merged in regionalization studies: Catchments having apparently similar physical characteristics are assumed to have a similar hydrological behavior. In this paper, we wished to test the validity of this assumption. To this aim, we defined behavioral (hydrological) similarity on the basis of model parameter transferability. Then pools of hydrologically similar catchments are compared with pools of apparently physically similar catchments, as identified on the basis of physiographic catchment descriptors. The overlap between the two pools of similar catchments is analyzed, making it possible to judge the efficiency of the physical similarity measure and to identify hydrologically similar catchments in an ungauged context. The results show that the overlap between the two pools is significant for only 60% of the catchments. For the other catchments, two major reasons were identified as contributing to the lack of overlap: (1) these catchments often have a quite specific hydrological behavior and (2) the role of the underground properties of the catchment on its hydrological behavior was not found to be accurately described by the available physical descriptors, meaning that more relevant catchment descriptors should be sought to better describe the geological and lithological context in hydrological terms.