Streamflow Simulation Research Articles

DECIPHeR-GW v1: a coupled hydrological model with improved representation of surface–groundwater interactions

Abstract. Groundwater is a crucial part of the hydrologic cycle and the largest accessible freshwater source for humans and ecosystems. However, most hydrological models lack explicit representation of surface–groundwater interactions, leading to poor prediction performance in groundwater-dominated catchments. This study presents DECIPHeR-GW v1 (Dynamic fluxEs and ConnectIvity for Predictions of HydRology and GroundWater), a new surface–groundwater hydrological model that couples a model based on hydrological response units (HRUs) and a two-dimensional gridded groundwater model. Using a two-way coupling method, the groundwater model component receives recharge from HRUs, simulates surface–groundwater interactions, and returns groundwater levels and groundwater discharge to HRUs, where river routing is then performed. Depending on the storage capacity of the surface water model component and the position of the modelled groundwater level, three scenarios are developed to derive recharge and capture surface–groundwater interactions dynamically. Our coupled model was set up at 1 km spatial resolution for the groundwater model, and the average size of the surface water HRUs was 0.31 km2. The coupled model was calibrated and evaluated against daily flow time series from 669 catchments and groundwater level data from 1804 wells across England and Wales. The model provides streamflow simulation with a median Kling–Gupta efficiency (KGE) of 0.83 across varying hydro-climates, such as wetter catchments with a maximum mean annual rainfall of 3577 mm yr−1 in the west and drier catchments with a minimum of 562 mm yr−1 in the east of Great Britain, as well as diverse hydrogeological conditions including chalk, sandstone, and limestone. Higher KGE values are found in particular for the drier chalk catchments in southeast England, where the average KGE for streamflow increased from 0.49 in the benchmark DECIPHeR model to 0.7. Furthermore, our model reproduces temporal patterns of the groundwater level time series, with more than half of the wells achieving a Spearman correlation coefficient of 0.6 or higher when comparing simulations to observations. Simulating 51 years of daily data for the largest catchment, the Thames at the Kingston River basin (9948 km2), takes approximately 17 h on a standard CPU, facilitating multiple simulations for model calibration and sensitive analysis. Overall, this new DECIPHeR-GW model demonstrates enhanced accuracy and computational efficiency in reproducing streamflow and groundwater levels, making it a valuable tool for addressing water resources and management issues over large domains.

Assessing the adequacy of traditional hydrological models for climate change impact studies: a case for long short-term memory (LSTM) neural networks

Abstract. Climate change impact studies are essential for understanding the effects of changing climate conditions on water resources. This paper assesses the effectiveness of long short-term memory (LSTM) neural networks compared to traditional hydrological models for these studies. Traditional hydrological models, which rely on simplified process parameterization with a limited number of parameters, are examined for their capability to accurately predict future hydrological streamflow under scenarios of significant warming. In contrast, LSTM models, known for their capacity to learn from extensive sequences of data and capture temporal dependencies, present a promising alternative. This study is performed on 148 catchments, comparing four traditional hydrological models, each calibrated specifically on each catchment, against two LSTM models. The first LSTM model is trained regionally across the 148 catchments, while the second incorporates data from an additional 1000 catchments at the continental scale, many located in climate zones representative of the future climate within the study domain. The climate sensitivity of all six hydrological models is assessed using four simple climate scenarios (+3, +6 °C, −20 %, and +20 % mean annual precipitation) and an ensemble of 22 CMIP6 GCMs under the SSP5-8.5 scenario. Results indicate that LSTM-based models demonstrate a different climate sensitivity compared to traditional hydrological models. Moreover, analyses of precipitation elasticity to streamflow and multiple streamflow simulations on analogue catchments suggest that the continental LSTM model performs better and is therefore better suited for climate change impact studies – a conclusion that is also supported by theoretical arguments.

Assessing climate change impact on watershed hydrological processes and stream temperature by considering CO2 emissions.

Assessing climate change impact on watershed hydrological processes and stream temperature by considering CO2 emissions.

A Framework to Determine Present and Future Effects of Rain‐on‐Snow on Spring Hydrology and Nutrient Loading in the Lake Erie Basin

ABSTRACTRain‐on‐snow (ROS) events occur when temperatures allow for liquid precipitation to fall onto an existing snowpack. Although ROS is a fundamental facet of winter and spring hydrology in the Great Lakes Basin with the potential to result in severe flooding and influence water quality issues, its role in this region is understudied compared with alpine regions. Many watershed models do not comprehensively characterise the ROS process and thus may misrepresent hydrological and water quality outputs. Here, we elucidate the importance of ROS on spring water balance and nutrient loading in the Big Creek watershed, part of the Lake Erie Basin (LEB), through an ensemble of statistical, hydrological, and climate modelling tools. We found that spring flow events with enhanced ROS melt are conducive to excessive loading export from both agricultural and natural land uses. The incorporation of a novel ROS routine into the Soil and Water Assessment Tool (SWAT) model demonstrated that the modified version improved performance in 76% of 504 random streamflow simulations. The ROS characterisation can more accurately recreate the magnitude of extreme flow events in spring, which is a commonly reported shortcoming of the SWAT model. The ROS submodel simulated earlier shifts in snowmelt, water yield and evapotranspiration by 1 month compared with the original model. In examining a climate scenario associated with modest greenhouse gas emission changes, we found that monthly average streamflow over the 21st century is projected to remain relatively stable, but the occurrence of extreme flow conditions will increase. ROS event frequency is projected to increase in February and March and decrease in April in urban and natural land uses, but agricultural areas will only experience a slight decline, suggesting that the landscape attributes play an important role in localised shifts in ROS event frequency. We contend that ongoing watershed modelling work must include the ROS process to improve representation of critical facets of hydrology and water quality that could be extrapolated to other more complex watersheds within the LEB and elsewhere.

Improve streamflow simulations by combining machine learning pre-processing and post-processing

Improve streamflow simulations by combining machine learning pre-processing and post-processing

Understanding Hydrological Responses to Land Use and Land Cover Change in the Belize River Watershed

Increasing forest destruction from land use and land cover change (LULCC) has altered catchment hydrological processes worldwide. This trend is also endemic to the Belize River Watershed (BRW), a significant source of land and water resources for Belize. This study aims to understand LULCC impacts on BRW hydrological responses from 2000 to 2020 by applying the widely used Soil and Water Assessment Tool (SWAT). This study identified historical trends in LULCC in the BRW and explored an alternative 2020 land cover scenario to elucidate the role of protected forests for hydrological response regulation. A SWAT model for the BRW was developed at the monthly timescale and calibrated on in situ streamflow using SWAT Calibrations and Uncertainty Programs (SWAT-CUP). The results showed that the BRW SWAT model performed satisfactorily for streamflow simulation at the Benque Viejo (BV) gauge station but performed variably at the Double Run (DR) gauge station. Overall, the findings revealed watershed-level increases in monthly average sediment yield (34.40%), surface runoff (24.95%), streamflow (16.86%), water yield (16.02%), baseflow (11.58%), and percolation (3.40%), and decreases in monthly average evapotranspiration (ET) (3.52%). In conclusion, the BRW SWAT model is promising for uncovering the hydrological impacts of LULCCs with opportunities for further model improvement.

Quantifying the Added Values of a Merged Precipitation Product in Streamflow Prediction over the Central Himalayas

Gridded precipitation datasets (GPDs) have complemented gauge-based measurements in global hydrology by providing spatiotemporally continuous rainfall estimates for streamflow prediction. However, these datasets suffer from uncertainties in space and time, particularly in complex terrains like the Himalayas. Merging multi-GPDs offers a potential solution to reduce such uncertainties, but the actual contribution of the merged product to hydrological modeling remains underexplored in data-scarce and topographically complex regions. Here, we applied a gauge-independent merging technique called Signal-to-Noise Ratio optimization (SNR-opt) to merge three precipitation products: ERA5, SM2RAIN, and IMERG-late. The resulting Merged Gridded Precipitation Dataset (MGPD) was evaluated using the hydrological model (HYMOD) across three major river basins in the Central Himalayas (Koshi, Narayani, and Karnali). The results show that MGPD significantly outperforms the individual GPDs in streamflow simulation. This is evidenced by higher Nash–Sutcliffe Efficiency (NSE) values, 0.87 (Narayani) and 0.86 (Karnali), compared to ERA5 (0.83, 0.82), SM2RAIN (0.83, 0.85), and IMERG-Late (0.82, 0.78). In Koshi, the merged product (NSE = 0.80) showed slightly lower performance than SM2RAIN (NSE = 0.82) and ERA5 (NSE = 0.81), likely due to the poor performance of IMERG-Late (NSE = 0.69) in this basin. These findings underscore the value of merging precipitation datasets to enhance the accuracy and reliability of hydrological modeling, especially in ungauged or data-scarce mountainous regions, offering important implications for water resource management and forecasting.

Integrating Ground and Satellite-based Precipitation Data for Streamflow Simulation and Soil Erosion Hotspot Mapping in the Data-Scarce Ruvu River Basin, Tanzania

Integrating Ground and Satellite-based Precipitation Data for Streamflow Simulation and Soil Erosion Hotspot Mapping in the Data-Scarce Ruvu River Basin, Tanzania

Comparative Analysis of SWAT Model Performance under Varying Land Use and Soil Data Resolutions in Eastern India

This study investigated the impact of land use/land cover (LULC) and soil data resolution on hydrological modeling using the Soil and Water Assessment Tool (SWAT) in the Bargarh command area of the Mahanadi River Basin, India. Six scenarios combining different resolutions of LULC (1:250000, 1:50000, and 10m spatial resolution) and soil data (1:5000000 and 1:50000) were used to simulate streamflow. The SWAT model was calibrated (2000-2012) and validated (2013-2020) using the SUFI-2 algorithm. Model performance was evaluated using the Nash-Sutcliffe efficiency (NSE), coefficient of determination (R2), percent bias (P-bias), R-factor, and P-factor. The results showed that the simulated streamflow matched well with the observed data for all scenarios, with the D1L1S1, D1L3S1, and D1L3S2 setups exhibiting the best agreement. The NSE values ranged from 0.76 to 0.89, R2 from 0.86 to 0.91, P-bias from -3.8% to 31.7%, R-factor from 0.56 to 1.20, and P-factor from 0.73 to 0.91 during calibration and validation. This study highlights the importance of the combination and classification of datasets in hydrological modeling, indicating that fitted parameter values do not adjust uniformly for all uncertainty sources. These findings provide guidance for selecting appropriate precision levels of input data for future applications to improve the accuracy and reliability of hydrological predictions. This study also highlights the practical relevance of hydrological modeling for water resource planning and management. The findings provide guidance on selecting appropriate input data resolutions to improve the accuracy and reliability of streamflow predictions using the SWAT model.

Exploring Flood Predictability in Taiwan through Coupled Atmospheric–Hydrological and High-Performance Hydrodynamic Models

Abstract Effective flood simulation capabilities can tremendously support early warning and disaster prevention. To examine the applicability of a fully physics-based and high-performance flood simulation and forecasting modeling framework for a flood-prone region in Taiwan, we conduct a numerical experiment that couples the Weather Research and Forecasting (WRF) Model, WRF-Hydrological modeling system (WRF-Hydro), and the Two-Dimensional Runoff Inundation Toolkit for Operational Needs (TRITON) to perform integrated rainfall, streamflow, and flood simulations. We first use the coupled WRF and WRF-Hydro (WWH) to predict rainfall and streamflow and then drive TRITON with the predicted streamflow hydrographs to simulate flood depth and inundation area. With the refined spatial resolution and parameterization, this framework can better predict rainfall with reasonable spatial patterns. Although WWH could overestimate the amount of rainfall in some areas, the uncertain rainfall–streamflow predictions produce reasonable flood maps able to pinpoint regions at risk of flooding. In terms of model efficiency, the graphics processing unit–based computation can yield a speed-up factor as high as ∼13 compared to the central processing unit–based computation, promoting the efficacy of the coupled modeling framework in practical real-time flood forecasting. Significance Statement The purpose of this study is to develop and assess a modeling framework that integrates a coupled atmospheric–hydrological model with a high-performance hydrodynamic model for rainfall, streamflow, and flood simulations in a holistic manner. By integrating these state-of-the-art models, we aim to advance our understanding of regional flood predictability subject to the chaotic nature of hydrometeorological forecasts. Based on our case study in Taiwan, we demonstrate that the proposed modeling framework can identify inundation hotspots despite the found uncertainties in rainfall–streamflow predictions. We also demonstrate that using the high-performance hydrodynamic model can significantly enhance simulation efficiency, suggesting the need for a modeling framework like the one proposed in this study for real-time flood forecasting.

Impact of baseflow separation on improving streamflow and its extremes with a hybrid model coupling hydrological and machine learning models

Accurate streamflow simulation is crucial for understanding hydrological processes and predicting extreme events. However, traditional hydrological models and machine learning (ML) models both have limitations, particularly in simulating extreme values. To address these challenges, this study builds a hybrid model that integrates the Distributed Hydrology Soil Vegetation Model (DHSVM) with the Long Short-Term Memory (LSTM) network. A baseflow separation method is incorporated into models to investigate its impact on streamflow. The effectiveness of the model is assessed in the Xiangjiang River Basin, compared against standalone DHSVM and LSTM models. The results demonstrate that the hybrid model outperforms DHSVM and LSTM, raising streamflow NSE to 0.9 and reducing PBIAS to -1.3%. Baseflow separation enhances the hybrid model’s baseflow and low flow, cutting PBIAS from 8.5% and 53.5% to 0.8% and 19.7%, respectively. These findings underscore that baseflow separation can improve streamflow simulations, including the accuracy of extreme values.

Incorporating multi-timescale data into a single long short-term memory network to enhance reservoir-regulated streamflow simulation

Incorporating multi-timescale data into a single long short-term memory network to enhance reservoir-regulated streamflow simulation

A Novel Stream Network Upscaling Scheme for Accurate Local Streamflow Simulations in Gridded Global Hydrological Models

Abstract Large‐scale hydrological models are progressing toward sub‐kilometer resolutions to achieve “locally relevant hydrological simulations.” However, grid‐based domain representations introduce significant errors in streamflow within small catchments, a challenge that remains unresolved by state‐of‐the‐art modeling schemes, such as 8‐directional gridded routing (D8). Here, we introduce the Subgrid Catchment Conservation (SCC) scheme to enhance streamflow estimation at any location within the domain using a coarse resolution (i.e., 1 km or more), continental scale, grid‐based hydrological model (HM). Gridded HMs not preserving the DEM‐reference (or subgrid) catchment area is usually referred as the catchment size problem. SCC allows multiple outflow from grid cells, a key feature that preserves the subgrid catchment area of all points of interest across scales. SCC is a general concept; however, for demonstration purposes, it has been implemented in the mesoscale hydrological model, mHM. We employ a global setup with 62 large‐scale domains encompassing 5,256 streamflow gauging stations, and a regional setup encompassing 187 stations in the Rhine river basin. We found that the widely used D8 scheme's efficacy diminishes drastically for catchments under 30 times the grid size. SCC demonstrates remarkable consistency in streamflow in the regional experiment with nine out of 10 stations exceeding the mean flow benchmark (KGE −0.41) across 1–100 km model resolutions. In addition, SCC's ability to resolve multiple points of interests in a grid leads to greater modeling flexibility. By addressing the catchment size problem, SCC marks a significant advancement for global‐scale simulations producing locally relevant streamflow.

Evaluation of SWAT-C Model Applicability for TOC Simulation and Analysis of TOC Load Characteristics in the Geumho River Basin

This study evaluated the applicability of the SWAT-C model for Total Organic Carbon (TOC) simulation on a daily basis in the Geumho River basin and analyzed the spatiotemporal characteristics of TOC loads. Model calibration and validation results showed “very good” performance for streamflow simulation during both calibration (2018-2020) and validation (2021-2022) periods based on R2 and NSE indicators. TOC simulation demonstrated “satisfactory” or better performance for the same periods. However, a tendency to underestimate TOC loads was observed during peak flow events. TOC loads showed a high correlation with precipitation (<i>r</i> = 0.99, <i>p</i> < 0.05) and exhibited distinct seasonal patterns, with the highest loads in summer and the lowest in winter. Analysis of TOC composition revealed that Dissolved Organic Carbon (DOC) accounted for an average of 82% of TOC loads while Particulate Organic Carbon (POC) contributed 18% in the Geumho River basin, with higher POC ratios observed during the fall season compared to other seasons. This study is significant as it represents the first application of daily TOC simulation using the SWAT-C model in the Geumho River basin in Korea, validating the model's applicability and comprehensively analyzing the temporal characteristics of TOC loads. The results provide scientific basis for implementing TOC Total Maximum Daily Load (TMDL) regulations in the Geumho River basin and contribute to developing effective water quality management strategies.

Evaluasi Data Hujan Berbasis Satelit untuk Menentukan Debit Aliran Masuk Waduk Selorejo Menggunakan Model HBV-96

Effective reservoir management can be supported by applying rainfall-runoff hydrological models. However, one of the main challenges of such models lies in the availability of reliable rainfall data. Satellite-based rainfall data offer a viable alternative to address this issue. This study aims to evaluate the reliability of satellite-based rainfall data for hydrological applications, specifically for simulating reservoir inflow using the HBV-96 model in the Selorejo Reservoir. The rainfall data used in this study include satellite-based datasets from TRMM, GPM, and RCM, tested in both raw and corrected forms. The HBV-96 model parameters were calibrated using observed rainfall data from 1998 to 2008, achieving a correlation coefficient and Nash-Sutcliffe Efficiency (NSE) of 0.86 and 0.72, respectively, for simulated streamflow. The model's performance was subsequently verified using observed rainfall data from 2009 to 2016, yielding consistent results with a correlation coefficient and NSE of 0.832 and 0.71, respectively. These calibrated parameters were then applied to the satellite rainfall datasets. The findings reveal that, in general, corrected TRMM satellite rainfall data using regression equations were not suitable for hydrological modelling. However, TRMM data corrected using duration curves significantly reduced deviations by up to 50% compared to raw data and provided better-simulated streamflow results, aligning more closely with observed streamflow. Conversely, RCM rainfall data, whether raw or corrected, performed poorly in the HBV model, with negative NSE values. Meanwhile, the bias-corrected GPM satellite rainfall data demonstrated the best performance in the HBV model, with a maximum deviation of only 5.81%.

Monthly hydropower generation data for Western Canada to support Western-US interconnect power system studies

Hydroelectric power generation in Western Canada significantly contributes to power grid operations of the North American Western Interconnection through substantial generation, some of which is exported to the United States (U.S.). However, the lack of publicly available hydropower generation datasets poses challenges for future market projections and resource adequacy evaluations. We present a simulation-based monthly power system model-ready hydropower generation dataset for 110 facilities in British Columbia and Alberta from 1981 to 2019. These monthly hydropower generation estimates are developed from integrated hydrologic model simulations of runoff and reservoir-operated streamflow, followed by scaling that considers diversion inflow constraints based on hydropower water license information. To address the lack of comparable hydropower generation records, we conduct step-by-step evaluations for simulated runoff, regulated streamflow, and hydropower generation using available observations or estimates. The presented hydropower dataset aims to enhance the representation of hydropower resources in Western Canada, supporting power grid system studies for the Western Interconnection of the U.S. and Canada.

Assessing Coincidence of Satellite Acquisitions and Flood Events to Predict Suitability for Flood Map Synthesis

Flooding is a global problem that impacts people, communities, and governments every year. A better understanding of flooding in an area can enable an improved emergency response before a flood hits. Flood maps are a crucial tool to translate what, for most, is an abstract streamflow into a more understandable and actionable representation of who and what is at risk. Satellite-based flood maps are a useful tool that has potential global applications. We developed methods to determine areas that are suitable for generating satellite-based synthetic flood maps. For our processes, we used Forecasting Inundation Extents using REOF analysis (FIER), a data-driven method of synthesizing flood maps by correlating extracted spatial and temporal patterns from satellite imagery with historical hydrological variables. To overcome the limitation of only using places where gauges are installed, we used large-scale hydrological models, namely the National Water Model (NWM) and the GEOGLOWS Streamflow Model, to provide simulated retrospective streamflow data to train our model. We evaluated locations where both optical and radar imagery would be suitable for creating these models. The procedures we developed and the results that we obtained are potentially transferable to many satellite data sources and methods of model generation.

Exploring the Controlling Factors of Watershed Streamflow Variability Using Hydrological and Machine Learning Models

Abstract Studying streamflow processes and controlling factors is crucial for sustainable water resource management. This study demonstrated the potential of integrating hydrological models with machine learning by constructing two machine learning methods, Extreme Gradient Boosting (XGBoost) and Random Forest (RF), based on the input and output data from the Soil and Water Assessment Tool (SWAT) and comparing their streamflow simulation performances. The Shapley Additive exPlanations (SHAP) method identified the controlling factors and their interactions in streamflow variation, whereas scenario simulations quantified the relative contributions of climate and land use changes. The results showed that when integrated with the SWAT model, XGBoost demonstrated better streamflow simulation performance than RF. Among the key factors influencing streamflow variation, area was the most important, with precipitation having a stronger impact than temperature, positively affecting streamflow when exceeding 550 mm. Different land use types exerted nonlinear impacts on streamflow, with notable differences and threshold effects. Specifically, grassland, cropland, and forest positively contributed to streamflow when their proportions were below 50%, above 20%, and between 30% and 50%, respectively. Nonlinear interaction effects on streamflow between land use types resulted in positive or negative contributions at specific proportion thresholds. Furthermore, precipitation was not dominant in the interaction with land use. Streamflow changes were primarily driven by drastic land use changes, which contributed 55.71%, while climate change accounted for 44.27%. This integration of hydrological models with machine learning revealed the complex impacts of climate and land use changes on streamflow, offering scientific insights for watershed water resource management.

TSWS: An observation-based streamflow dataset of Tianshan Mountains watersheds (1901–2019)

Due to scarcity of data and complex hydrological conditions in the Tianshan region, long-term and complete streamflow data are lacking. This study produced a multi-basin streamflow dataset, named Tianshan watershed streamflow (TSWS) dataset, by comparing the results of Hydrologiska Byråns Vattenavdelning and Long Short-Term Memory models, and analyzed spatiotemporal variation of streamflow.TSWS dataset provides daily streamflow data for 56 watersheds and monthly streamflow data for 89 watersheds in the Tianshan Mountains in 1901–2019. The streamflow simulations of 40 watersheds (daily scale) and 70 watersheds (monthly scale) passed the S-tests (Nash-Sutcliffe efficiency ≥0.5, percent bias ≤25%, and ratio of the root-mean-square error to the standard deviation of measured data ≤0.7). The dataset showed an overall increasing trend of streamflow, especially from 1990 to 2019; spatially, it showed higher streamflow in the west and south, and lower streamflow in the east and north. The dataset provides the first comprehensive simulation and its long time series will provide an important reference for climatic and hydrological studies.

SWAT simulation of streamflow in Sironko catchment on Mt. Elgon, Eastern Uganda

We determined the 2025-2040 impact of climate and land use land cover change on streamflow in the Sironko catchment. LULCC was predicted using Cellular Automata Markov. The climate was statistically downscaled from the 29 GCMs of the IPCC Fifth Coupled Model Intercomparison Project using the delta method of the Agricultural Model Inter-comparison and Improvement Project under five climate regimes and Representative Concentration Pathways 4.5 and 8.5. Streamflow was simulated using the Soil and Water Assessment Tool. One sample t-test confirmed significant differences in the 1980–2010 and 2025–2040 streamflow. A second-degree regression model computed the contributions of climate and LULCC on streamflow. Significant differences (p < 0.05) were detected in the 1980–2010 and 2025–2040 streamflow. Climate change will increase streamflow by 7.6–49.20% and 6.50–59.20% under RCP 4.5 and RCP 8.5 while LULCC will increase streamflow by 4.10–39.20% and 1.30–45.30% under RCP 4.5 and RCP 8.5 calling for a multifaceted approach to mitigate the identified impacts.