Testing and improving a distributed rainfall-runoff model using slope surface flow as a new observation metric

Reliable hydrologic modeling is essential for sustainable water resource management, accurate projections of future water availability, and the timely forecasting of extreme weather events. The severe flooding that occurred in July 2021, however, highlighted significant limitations in current predictive models, particularly their inadequacy in capturing unprecedented events. These limitations are largely attributed to factors such as sparse observational data, simulation uncertainties, and challenges in model calibration. Another critical factor contributing to flood events is the accumulation of soil moisture following prolonged rainfall, which often leads to soil saturation. Despite its importance, many existing flood forecasting systems inadequately represent soil moisture processes and the complex interactions between surface and subsurface hydrological components. In response to these shortcomings, this study explores the integration of machine learning techniques with hydrologic models to enhance the accuracy of flood predictions. Specifically, we explore the impact of incorporating soil water content (SWC), derived from The integrated surface-subsurface ParFlow/CLM model, into an artificial intelligence (AI) framework for flood forecasting. We developed a Gated Recurrent Unit (GRU) Convolutional model, which initially used eight atmospheric variables (including longwave and shortwave radiation, total precipitation, air temperature, air pressure, specific humidity, wind velocity components) to predict river discharge. In the subsequent stage, SWC was added as an additional input to evaluate its effect on the model’s performance. Our comparative analyses revealed that including SWC significantly improved the predictive accuracy of the AI model, highlighting its essential role in discharge estimation. Furthermore, integrating the ParFlow model’s outputs with machine learning techniques resulted in a substantial enhancement in flood prediction, outperforming both standalone AI and hydrological models. The AI model with SWC achieved an RMSE reduction of up to 51% and an increase in R2 from 0.634 to 0.916 at Kordel-Kyll. The findings demonstrate the potential of informing machine learning with integrated hydrologic model output to improve flood prediction systems, offering valuable insights for more effective water resource management. This graphical abstract presents an integrated modeling framework that combines a fully coupled subsurface-surface hydrological model, ParFlow/CLM, with a machine learning algorithm for river discharge prediction. The system was applied to a real flood event in Germany, demonstrating its practical relevance. Atmospheric forcings, including precipitation (APCP), temperature (Temp), specific humidity (SPFH), pressure (Press), wind components (UGRD, VGRD), and radiation (DLWR, DSWR), are collected along with high-resolution soil water content (SWC) data and observed river discharge. The detailed SWC data capture near-saturated soil conditions, which contributed to flooding despite only moderate, but sustained, rainfall. ParFlow/CLM simulates key land surface processes such as snow accumulation, overland flow, and evapotranspiration. Outputs from ParFlow/CLM and meteorological inputs are used as features in a hybrid machine learning model, specifically, a Gated Recurrent Unit (GRU)-based Convolutional Neural Network (CNN), which captures temporal dynamics for improved discharge forecasting. A coupling interface links the physical model with the machine learning component. Simulations correspond to observed flood periods and are evaluated using both deterministic and probabilistic metrics. Incorporating hydrological models with machine learning, especially with soil moisture as an input, improves both flood timing and peak prediction. This approach supports early warning systems, where soil moisture forecasts can be used alongside meteorological predictions to enhance model performance, particularly in areas with flatter terrain, where soil moisture plays a more significant role than in steep regions. GRU model trained on eight key atmospheric variables for discharge prediction. Soil Water Content (SWC) from ParFlow/CLM added as a new input feature. SWC inclusion significantly boosts flood prediction accuracy. Machine learning boosts hydrological modeling in capturing extreme flood events.

Rapid population growth, economic development, land-use modifications, and climate change are the major driving forces of growing hydrological disasters like floods and water stress. Reliable flood modelling is challenging due to the spatiotemporal changes in precipitation intensity, duration and frequency, heterogeneity in temperature rise and land-use changes. Reliable high-resolution precipitation data and distributed hydrological model can solve the problem. This study aims to develop a distributed hydrological model using Machine Learning (ML) algorithms to simulate streamflow extremes from satellite-based high-resolution climate data. Four widely used bias correction methods were compared to select the best method for downscaling coupled model intercomparison project (CMIP6) global climate model (GCMs) simulations. A novel ML-based distributed hydrological model was developed for modelling runoff from the corrected satellite rainfall data. Finally, the model was used to project future changes in runoff and streamflow extremes from the downscaled GCM projected climate. The Johor River Basin (JRB) in Malaysia was considered as the case study area. The distributed hydrological model developed using ML showed Nash–Sutcliffe efficiency (NSE) values of 0.96 and 0.78 and Root Mean Square Error (RMSE) of 4.01 and 5.64 during calibration and validation. The simulated flow analysis using the model showed that the river discharge would increase in the near future (2020–2059) and the far future (2060–2099) for different Shared Socioeconomic Pathways (SSPs). The largest change in river discharge would be for SSP-585. The extreme rainfall indices, such as Total Rainfall above 95th Percentile (R95TOT), Total Rainfall above 99th Percentile (R99TOT), One day Max Rainfall (R × 1day), Five-day Max Rainfall (R × 5day), and Rainfall Intensity (RI), were projected to increase from 5% for SSP-119 to 37% for SSP-585 in the future compared to the base period. The results showed that climate change and socio-economic development would cause an increase in the frequency of streamflow extremes, causing larger flood events.

This study suggests a radical approach to hydrologic predictions in ungauged basins, addressing the long standing challenge of issuing predictions when in-situ river discharge does not exist. A simple but powerful rationale for measuring and modeling river discharge is proposed, using coupled advances in hydrologic modeling and satellite remote sensing. Our approach presents a Surrogate River discharge driven Model (SRM) that infers Surrogate River discharge (SR) from remotely sensed microwave signals with the ability to mimic river discharge in varying topographies and vegetation cover, which is then used to calibrate a hydrological model enabling physical realism in the resulting river discharge profile by adding an estimated mean of river discharge via the Budyko framework. The strength of SRM comes from the fact that it only uses remotely sensed data in prediction. The approach is demonstrated for 130 catchments in the Murray Darling Basin (MDB) in Australia, a region of high economic and environmental importance. The newly proposed SR (SRL, representing L-band microwave) boosts the Nash-Sutcliffe Efficiency (NSE) of modeled flow, showing a mean NSE of 0.54, with 70% of catchments exceeding NSE 0.4. We conclude that SRM effectively predicts high-flow and low-flow events related to flood and drought. Overall, this new approach will significantly improve catchment simulation capacity, enhancing water security and flood forecasting capability not only in the MDB but also worldwide.

The purpose of this article is to determine the response of river discharge to precipitation variability in the Gucha-Migori River Basin, where recurrent flood events persist despite the limited application of non-structural mitigation measures. Understanding the rainfall–runoff relationship is essential for improving flood risk management and decision-making. The study employed the Hydrologic Engineering Centre–Hydrologic Modelling System (HEC-HMS) to simulate river discharge using precipitation data for the period 1969–2015. The model’s performance was evaluated through R² and Nash–Sutcliffe efficiency during calibration and validation. A correlation analysis was also done to determine how seasonal rainfall relates to river discharge. The results show that the HEC-HMS model achieved moderate performance, with R² and Nash values of 0.52 and 0.36 during calibration, and 0.42 and 0.31 during validation, respectively. The findings further indicate a statistically significant positive relationship between seasonal precipitation and river discharge at the 0.05 significance level, confirming that precipitation variability strongly influences discharge patterns in the basin. The study concludes that precipitation variability is a key driver of river discharge and that HEC-HMS is a suitable tool for simulating rainfall–runoff processes. It is recommended that hydrological modelling be integrated into flood risk assessment and early warning systems to enhance preparedness and support sustainable water resource management in the basin. The findings provide a basis for improving flood response planning and reducing vulnerability to flood hazards.

Floods are one of the main natural disaster threats to the safety of people’s lives and property. Flood hazards intensify as the global risk of flooding increases. The control of flood disasters on the basin scale has always been an urgent problem to be solved that is firmly associated with the sustainable development of water resources. As important nonengineering measures for flood simulation and flood control, the hydrological and hydraulic models have been widely applied in recent decades. In our study, on the basis of sufficient remote-sensing and hydrological data, a hydrological (Xin’anjiang (XAJ)) and a two-dimensional hydraulic (2D) model were constructed to simulate flood events and provide support for basin flood management. In the Chengcun basin, the two models were applied, and the model parameters were calibrated by the parameter estimation (PEST) automatic calibration algorithm in combination with the measured data of 10 typical flood events from 1990 to 1996. Results show that the two models performed well in the Chengcun basin. The average Nash–Sutcliffe efficiency (NSE), percentage error of peak discharge (PE), and percentage error of flood volume (RE) were 0.79, 16.55%, and 18.27%, respectively, for the XAJ model, and those values were 0.76, 12.83%, and 11.03% for 2D model. These results indicate that the models had high accuracy, and hydrological and hydraulic models both had good application performance in the Chengcun basin. The study can a provide decision-making basis and theoretical support for flood simulation, and the formulation of flood control and disaster mitigation measures in the basin.

Behind every robust result is a robust method: Perspectives from a case study and publication process in hydrological modelling

ABSTRACT: The shift from rural to urban living has resulted in natural revegetation of abandoned rural areas located in mountainous regions. These land-use changes, combined with recent temperature and precipitation trends, affect water resources and sediment yields. Consequently, surface runoff, water infiltration, and sediment production/transport are impacted. Studies of this impact in mountainous areas are limited and partial, restricted to small basins. Therefore, this study examines the changes in both water and sediment fluxes in the northern draining region of the Ebro Basin. For the study, the SWAT+ hydrological model was employed. SWAT+ is a semi-distributed, deterministic, continuous basin model that operates on a daily time step. The model requires several inputs: a Digital Elevation Model (DEM), reservoir locations, land use map and index, soil type map and properties, as well as climatic data. The Ebro basin was divided into eight sub-basins and a hydrological model was developed for each. Calibration and validation processes were conducted in two steps: firstly, hydrological calibration and validation was performed and afterwards the process was repeated for the sediments. Firstly, hydrological calibration and validation were conducted at available and relevant gauging stations within the sub-basins. In total, over 30 stations were used as control points. The precision of the calibration and validation was evaluated using the Nash-Sutcliffe Efficiency (NSE). Secondly, sediment calibration was conducted using available reservoir bathymetry data or sediment yield estimates from scientific literature. The sediment calibration aimed to reproduce values of the same order of magnitude as those derived from bathymetry or literature data. The resulting hydrologic and sedimentologic model was re-run and NSE, maintaining or improving the NSE values obtained from the hydrological calibration. NSE values, thus, ranged from 0.53 to 0.95 depending on the gauging station. The mean NSE for the entire basin under study was 0.75, indicating that a well-established hydrologic and sedimentologic model was achieved. Future work will involve, firstly, applying climate change scenarios from CMIP6 and comparing the current results to observe the response to climate change. Secondly, generating downscaled land-use change maps to accurately represent the evolution of land-use change in each sub-basin. This approach will allow for the analysis of the impacts of climate change alone and in combination with land-use change. ACKNOWLEDGMENTS: This work is funded by the European Research Council (ERC) through the Horizon Europe 2021 Starting Grant program under REA grant agreement number 101039181 - SEDAHEAD.

Hydro-climatology study of the Ogooué River basin using hydrological modeling and satellite altimetry

Precise streamflow prediction is fundamental for effective water resources management, flood risk mitigation, and sustainable agricultural planning, particularly in regions dependent on rainfed agriculture. This study evaluates the prediction capability of four hydrological models of parameter-efficient distribution (PED), Hydrologiska Byråns Vattenbalansavdelning (HBV), Hydrological Engineering Center-Hydrological Modeling System (HEC-HMS), and Soil and Water Assessment Tool (SWAT) in the Koga Watershed, Ethiopia. The models were calibrated from 1997 to 2006 and validated from 2007 to 2011 using observed daily streamflow. During calibration, the PED model showed the best performance with coefficient of determination (R2) (0.79), Nash-Sutcliffe efficiency (NSE) (0.782), root mean square error (RMSE) (0.42), and percentage of bias (PBIAS) (7.56%), while SWAT simulated the highest flows, and HEC-HMS slightly overestimated flows. During validation, PED had an excellent performance (R2 = 0.70, NSE = 0.72, RMSE = 0.65, and PBIAS = 16%), whereas HBV had minimum flows, and SWAT forecasted minimal flows. Inclusively, the PED model is found to be the most suitable model for flow prediction in the watershed established due to its consistency for sustainable water resource management. The findings provide valuable insights for selecting suitable hydrological models to improve water resource planning and execution.

A representative of meteorological data-constrained basin, Ayeyarwady, in Myanmar, Southeast Asia, is set for flow simulation and forecasting at 15 locations using a range of hydrological modeling approaches: conceptual lumped (GR4J), hybrid-lumped [Identification of unit Hydrographs And Component flows from Rainfall Evapotranspiration and Streamflow Catchment Wetness Index (IHACRES CWI)], semidistributed [Hydrological Engineering Center-Hydrological Modeling System (HEC-HMS)], and relatively distributed [Soil and Water Assessment Tool (SWAT)]. Using daily rainfall data from 51 surface rainfall stations (over an area of approximately 400,000 km2) and coarse monthly evaporation inputs from global sources, the models are calibrated (validated) against observed flows for 2001–2009 (2010–2014) using the performance indicators Nash-Sutcliffe efficiency (NSE), percentage bias (PBIAS), RMSE-observations standard deviation ratio (RSR), and volumetric efficiency . The developed models were then integrated with rainfall forecasts from the Weather Research and Forecasting Model for 2015–2018 and assessed for biases against observed flows. The NSE values were favorable for GR4J (median NSE=0.9), followed by IHACRES (NSE=0.86), SWAT (NSE=0.81) and HEC-HMS (NSE=0.77) during calibration and GR4J (NSE=0.87) and the latter three (NSE=0.83) during validation. Lumped models were found to have comparable, albeit better in simulating low, median, and high quantiles of flows during both calibration and validation periods, compared to other models of varying complexity set for the study basin. The hydrometeorological coupling also revealed that GR4J yielded the least while HEC-HMS yielded the highest biases (up to 30-fold at some stations) in daily flow forecasting. The analysis suggested that while process-based and relatively complex models may exhibit better performance in data-rich basins, simple conceptual models like GR4J are useful for daily flow simulation and forecasting in data-constrained basins of the region.

Flooding has increasingly posed significant challenges in the Western Cape, South Africa, with the September 2023 floods in Franschhoek underscoring the vulnerability of the region to extreme rainfall events. During this event, the area received over 220 mm of rainfall within 48 hours, resulting in extensive flooding that inundated approximately 500 hectares, displaced over 1,000 residents, and caused substantial damage to infrastructure. This study developed an integrated Flood Risk Information System (FRIS) designed for flood-prone regions in the Western Cape, utilizing Earth Observation (EO) technologies, hydrological modelling, and Geographic Information Systems (GIS).The system integrated historical flood data, municipal hydrological observations, and real-time environmental variables, including rainfall, river discharge, and soil moisture, to enhance flood risk prediction, monitoring, and response. Hydrological modelling was conducted using the HEC-HMS (Hydrologic Engineering Center’s Hydrologic Modeling System) and SWAT (Soil and Water Assessment Tool) models. Machine learning algorithms, including Random Forest (RF) and Gradient Boosting Machine (GBM), were implemented to predict flood probabilities. Model outputs were validated against observed data from the local municipality, which included flood extent maps and river discharge measurements.The system demonstrated high accuracy in predicting flood extents, with the HEC-HMS model achieving a Nash-Sutcliffe Efficiency (NSE) of 0.88 and a Root Mean Square Error (RMSE) of 12% compared to observed discharge data. The machine learning models yielded flood prediction accuracies of 87% (RF) and 91% (GBM) when compared to observed flood extents. Google Earth Engine (GEE) was used to process large EO datasets, allowing for real-time flood mapping and risk analysis.The FRIS proved instrumental in being able to model the September 2023 floods by providing accurate predictions and mapping, enabling disaster management agencies to target evacuation efforts and allocate resources effectively. However, further improvements are planned, including incorporating finer-resolution rainfall and topographic data, expanding the system’s spatial coverage, and integrating socio-economic indicators to assess community vulnerability better.This study highlighted the potential of combining EO, GEE, GIS, and advanced hydrological models in improving flood risk management. The FRIS provides a powerful framework for mitigating flood impacts and protecting vulnerable communities, aligning with broader efforts to enhance climate adaptation and resilience in the Western Cape.

Monsoon regions are characterized by a pronounced seasonality of rainfall. Model-based analysis of water resources in such an environment has to take account of the specific natural conditions and the associated water management. Especially, plant phenology, which is predominately water driven, and water management, which aims at reducing water shortage, are of primary importance. The aim of this study is to utilize the Soil and Water Assessment Tool (SWAT) in a monsoon-driven region in the Indian Western Ghats by using mainly generally available input data and to evaluate the model performance under these conditions. The test site analyzed in this study is the meso-scale catchment of the Mula and Mutha Rivers (2036 km2) upstream of the city of Pune, India. Most input data were derived from remote sensing products or from international archives. Forest growth in SWAT was modified to account for the seasonal limitation of water availability. Moreover, a dam management scheme was derived by combining general dam management rules with reservoir storage capacity and estimated monthly outflow rates from river discharge. With these model adaptations, SWAT produced reasonable results when compared to mean daily discharge measured in three of four subcatchments during the rainy season (Nash-Sutcliffe efficiencies 0.58, 0.63, and 0.68). The weakest performance was found at the gauge downstream of four dams, where the simple dam management scheme failed to match the combined management effects of the four dams on river discharge (Nash-Sutcliffe efficiency 0.10). Water yield was underestimated by the model, especially in the smallest (headwater) subcatchment (99 km2). Due to the absence of rain gauges in these headwater areas, the extrapolation errors of rainfall estimates based on measurements at lower elevations are expected to be large. Moreover, there is some indication that evapotranspiration might be underestimated. Nevertheless, it can be concluded that using generally available data in SWAT model studies of monsoon-driven catchments provides reasonable results, if key characteristics of monsoon regions are accounted for and processes are parameterized accordingly.

Given the evident impact of climate change, the frequency of severe flood events has increased worldwide. For various risk-reduction measures, covering all rivers in a country or regions including small-to-medium-sized rivers, flood risk assessment and real-time forecasting based on large-domain and high-resolution distributed rainfall–runoff models are fundamental. Due to limited observed records in such small-to-medium-sized rivers, the used distributed model must be robust and physically sound with the regionalized model parameters. Specifically, rather than optimizing parameters in many independent river basins, leading to a patched parameter distribution, regionalization should reflect the spatial distribution of hydrological signatures, such as soil and geology types. However, optimizing the parameters with existing methods incurs computational costs, posing difficulties in the parameter regionalization of large-domain and high-resolution distributed runoff models. To address this challenge, we propose a parameter regionalization method based on conditional probability. The key feature of this method is that the calibration phase calculation assumes spatially uniform parameter sets within the calibrating basins, significantly reducing computational costs. However, the resulting parameter sets are spatially distributed corresponding to the region’s pre-prepared soil or geological maps. It was achieved by introducing the Bayes’ theorem to estimate the conditional probability of the parameter set. The proposed method was applied to the distributed rainfall–runoff–inundation (RRI) model developed for Japan with a resolution of 150 m. The model performance in the validation phase, in which the performance was evaluated with 2723 flood events at 711 gauging stations, the median Nash–Sutcliffe efficiency (NSE) being 0.87, comparable or even improved to the performance in the calibration phase (NSE = 0.83) with 525 flood events at 75 dam reservoirs. Overall, the obtained nationwide high-resolution model is robust with good performance, even in ungauged basins. Furthermore, the proposed regionalization is a simple and useful way reflecting spatially distributed hydrologic signatures in the model parameters, and it can be utilized for any distributed rainfall–runoff model.

Modelling the impact of past and future climate scenarios on streamflow in a highly mountainous watershed: A case study in the West Seti River Basin, Nepal

Calibrating a hydrological model in ungauged small river basins of the northeastern Tibetan Plateau based on near-infrared images