Basin‐scale stream‐flow forecasting using the information of large‐scale atmospheric circulation phenomena

It is well recognized that the time series of hydrologic variables, such as rainfall and streamflow are significantly influenced by various large‐scale atmospheric circulation patterns. The influence of El Niño‐southern oscillation (ENSO) on hydrologic variables, through hydroclimatic teleconnection, is recognized throughout the world. Indian summer monsoon rainfall (ISMR) has been proved to be significantly influenced by ENSO. Recently, it was established that the relationship between ISMR and ENSO is modulated by the influence of atmospheric circulation patterns over the Indian Ocean region. The influences of Indian Ocean dipole (IOD) mode and equatorial Indian Ocean oscillation (EQUINOO) on ISMR have been established in recent research. Thus, for the Indian subcontinent, hydrologic time series are significantly influenced by ENSO along with EQUINOO. Though the influence of these large‐scale atmospheric circulations on large‐scale rainfall patterns was investigated, their influence on basin‐scale stream‐flow is yet to be investigated. In this paper, information of ENSO from the tropical Pacific Ocean and EQUINOO from the tropical Indian Ocean is used in terms of their corresponding indices for stream‐flow forecasting of the Mahanadi River in the state of Orissa, India. To model the complex non‐linear relationship between basin‐scale stream‐flow and such large‐scale atmospheric circulation information, artificial neural network (ANN) methodology has been opted for the present study. Efficient optimization of ANN architecture is obtained by using an evolutionary optimizer based on a genetic algorithm. This study proves that use of such large‐scale atmospheric circulation information potentially improves the performance of monthly basin‐scale stream‐flow prediction which, in turn, helps in better management of water resources. Copyright © 2007 John Wiley & Sons, Ltd.     

Real‐time recurrent learning neural network for stream‐flow forecasting

Various types of neural networks have been proposed in previous papers for applications in hydrological events. However, most of these applied neural networks are classified as static neural networks, which are based on batch processes that update action only after the whole training data set has been presented. The time variate characteristics in hydrological processes have not been modelled well. In this paper, we present an alternative approach using an artificial neural network, termed real‐time recurrent learning (RTRL) for stream‐flow forecasting. To define the properties of the RTRL algorithm, we first compare the predictive ability of RTRL with least‐square estimated autoregressive integrated moving average models on several synthetic time‐series. Our results demonstrate that the RTRL network has a learning capacity with high efficiency and is an adequate model for time‐series prediction. We also investigated the RTRL network by using the rainfall–runoff data of the Da‐Chia River in Taiwan. The results show that RTRL can be applied with high accuracy to the study of real‐time stream‐flow forecasting networks. Copyright © 2002 John Wiley & Sons, Ltd.     

A comparison of snowmelt‐derived streamflow from temperature‐index and modified‐temperature‐index snow models

Reliable estimation of the volume and timing of snowmelt runoff is vital for water supply and flood forecasting in snow‐dominated regions. Snowmelt is often simulated using temperature‐index (TI) models due to their applicability in data‐sparse environments. Previous research has shown that a modified‐TI model, which uses a radiation‐derived proxy temperature instead of air temperature as its surrogate for available energy, can produce more accurate snow‐covered area (SCA) maps than a traditional TI model. However, it is unclear whether the improved SCA maps are associated with improved snow water equivalent (SWE) estimation across the watershed or improved snowmelt‐derived streamflow simulation. This paper evaluates whether a modified‐TI model produces better streamflow estimates than a TI model when they are used within a fully distributed hydrologic model. It further evaluates the performance of the two models when they are calibrated using either point SWE measurements or SCA maps. The Senator Beck Basin in Colorado is used as the study site because its surface is largely bedrock, which reduces the role of infiltration and emphasizes the role of the SWE pattern on streamflow generation. Streamflow is simulated using both models for 6 years. The modified‐TI model produces more accurate streamflow estimates (including flow volume and peak flow rate) than the TI model, likely because the modified‐TI model better reproduces the SWE pattern across the watershed. Both models also produce better performance when calibrated with SCA maps instead of point SWE data, likely because the SCA maps better constrain the space‐time pattern of SWE.     

Using high resolution tracer data to constrain water storage,flux and age estimates in a spatially distributed rainfall‐runoff model

Models simulating stream flow and conservative tracers can provide a representation of flow paths, storage distributions and mixing processes that is advantageous for many predictive purposes. Compared with models that only simulate stream flow, tracer data can be used to investigate the internal consistency of model behaviour and to gain insight into model performance. Here, we examine the strengths and weaknesses of a data‐driven, spatially distributed tracer‐aided rainfall‐runoff model. The model structure allowed us to assess the influence of landscape characteristics on the routing and mixing of water and tracers. The model was applied to a site in the Scottish Highlands with a unique tracer data set; ~4 years of daily isotope ratios in stream water and precipitation were available, as well as 2 years of weekly soil and ground water isotopes. The model structure was based on an empirically based, lumped tracer‐aided model previously developed for the catchment. The best model runs were selected from Monte Carlo simulations based on dual calibration criteria using objective functions for both stream isotopes and discharge at the outlet. Model performance for these criteria was reasonable (Nash–Sutcliffe efficiencies for discharge and isotope ratios were ~0.4–0.6). The model could generally reproduce the variable isotope signals in the soils of the steeper hill slopes where storage was low, and damped isotope responses in valley bottom cells with high storage. The model also allowed us to estimate the age distributions of internal stores, water fluxes and stream flow. Average stream water age was ~1.6 years, integrating older groundwater in the valley bottom and dynamic younger soil waters. By tracking water ages and simulating isotopes, the model captured the changes in connectivity driven by distributed storage dynamics. This has substantially improved the representation of spatio‐temporal process dynamics and gives a more robust framework for projecting environmental change impacts. Copyright © 2016 The Authors Hydrological Processes Published by John Wiley & Sons Ltd.     

A two‐step‐ahead recurrent neural network for stream‐flow forecasting

In many engineering problems, such as flood warning systems, accurate multistep‐ahead prediction is critically important. The main purpose of this study was to derive an algorithm for two‐step‐ahead forecasting based on a real‐time recurrent learning (RTRL) neural network that has been demonstrated as best suited for real‐time application in various problems. To evaluate the properties of the developed two‐step‐ahead RTRL algorithm, we first compared its predictive ability with least‐square estimated autoregressive moving average with exogenous inputs (ARMAX) models on several synthetic time‐series. Our results demonstrate that the developed two‐step‐ahead RTRL network has efficient ability to learn and has comparable accuracy for time‐series prediction as the refitted ARMAX models. We then investigated the two‐step‐ahead RTRL network by using the rainfall–runoff data of the Da‐Chia River in Taiwan. The results show that the developed algorithm can be successfully applied with high accuracy for two‐step‐ahead real‐time stream‐flow forecasting. Copyright © 2003 John Wiley & Sons, Ltd.     

Removal of water‐surface reflection effects with a temporal minimum filter for UAV‐based shallow‐water photogrammetry

The recent development of structure‐from‐motion (SfM) and multi‐view stereo (MVS) photogrammetry techniques has enabled semi‐automatic high‐resolution bathymetry using aerial images taken by consumer‐grade digital cameras mounted on unmanned aerial vehicles (UAVs). However, the applicability of these techniques is sometimes limited by sun and sky reflections at the water surface, which render the point‐cloud density and accuracy insufficient. In this research, we present a new imaging technique to suppress the effect of these water‐surface reflections. In this technique, we order a drone to take a short video instead of a still picture at each waypoint. We then apply a temporal minimum filter to the video. This filter extracts the smallest RGB values in all the video frames for each pixel, and composes an image with greatly reduced reflection effects. To assess the performance of this technique, we applied it at three small shallow‐water sites. Specifically, we evaluated the effect of the technique on the point cloud density and the accuracy and precision of the photogrammetry. The results showed that the proposed technique achieved a far denser point cloud than the case in which a randomly chosen frame was used for each waypoint, and also showed better overall accuracy and precision in estimating water‐bottom elevation. The effectiveness of this new technique should depend on the surface wave state and sky radiance distribution, and this dependence, as well as the applicability to large areas, should be investigated in future research. Copyright © 2018 John Wiley & Sons, Ltd.     

Flood hydrograph reconstruction from the peak flow value in ephemeral streams using a simplified robust single‐parameter model

Flood hydrographs from ephemeral streams in arid areas provide valuable information for assessing run‐off and groundwater recharge. However, such data are often scarce or incomplete, especially in hyper‐arid regions. The hypothesis of this study was that it is possible to reconstruct a hydrograph of a specific point along an ephemeral stream with the knowledge of only the peak flow rate of a flood event at that point and that this can be done at almost every point along the stream. The feasibility of this approach lies in the shape of the recession stage of the flood hydrograph, which is known to be a repeating phenomenon. The recession stage comes immediately after the peak flow rate, when it begins its decline, and lasts until the flood is extinguished. A general shape of the flood recession stage can be provided. Because the recession stage represents ~80% of the duration of a flood event, it can provide a general idea of the flood hydrograph's shape. A simple model based on geometric progression is suggested to describe the repeating recession stage of a flood. The advantage of the proposed model is that it requires only one parameter: the recession characteristic at a fixed point along the ephemeral stream, termed recession coefficient q. By knowing the recession coefficient of a fixed point and the peak flow rate of a flood event at that point, one can plot the flood hydrograph. A good agreement is shown between the observed and computed values of the recession stage. Copyright © 2016 John Wiley & Sons, Ltd.     

Physically‐based modelling of double‐peak discharge responses at Slapton Wood catchment

Heavy winter rainfall produces double‐peak hydrographs at the Slapton Wood catchment, Devon, UK. The first peak is saturation‐excess overland flow in the hillslope hollows and the second (i.e. the delayed peak) is subsurface stormflow. The physically‐based spatially‐distributed model SHETRAN is used to try to improve the understanding of the processes that cause the double peaks. A three‐stage (multi‐scale) approach to calibration is used: (1) water balance validation for vertical one‐dimensional flow at arable, grassland and woodland plots; (2) two‐dimensional flow for cross‐sections cutting across the stream valley; and (3) three‐dimensional flow in the full catchment. The main data are for rainfall, stream discharge, evaporation, soil water potential and phreatic surface level. At each scale there was successful comparison with measured responses, using as far as possible parameter values from measurements. There was some calibration but all calibrated values at one scale were used at a larger scale. A large proportion of the subsurface runoff enters the stream from three dry valleys (hillslope hollows), and previous studies have suggested convergence of the water in the three large hollows as being the major mechanism for the production of the delayed peaks. The SHETRAN modelling suggests that the hillslopes that drain directly into the stream are also involved in producing the delayed discharges. The model shows how in the summer most of the catchment is hydraulically disconnected from the stream. In the autumn the catchment eventually ‘wets up’ and shallow subsurface flows are produced, with water deflected laterally along the soil‐bedrock interface producing the delayed peak in the stream hydrograph. Copyright © 2007 John Wiley & Sons, Ltd.     

Enforced self‐organizing map neural networks for river flood forecasting

Self‐organizing maps (SOMs) have been successfully accepted widely in science and engineering problems; not only are their results unbiased, but they can also be visualized. In this study, we propose an enforced SOM (ESOM) coupled with a linear regression output layer for flood forecasting. The ESOM re‐executes a few extra training patterns, e.g. the peak flow, as recycling input data increases the mapping space of peak flow in the topological structure of SOM, and the weighted sum of the extended output layer of the network improves the accuracy of forecasting peak flow. We have investigated an ESOM neural network by using the flood data of the Da‐Chia River, Taiwan, and evaluated its performance based on the results obtained from a commonly used back‐propagation neural network. The results demonstrate that the ESOM neural network has great efficiency for clustering, especially for the peak flow, and super capability of modelling the flood forecast. The topology maps created from the ESOM are interesting and informative. Copyright © 2007 John Wiley & Sons, Ltd.     

Prediction of episodic acidification in North‐eastern USA: an empirical/mechanistic approach

Observations from the US Environmental Protection Agency's Episodic Response Project (ERP) in the North‐eastern United States are used to develop an empirical/mechanistic scheme for prediction of the minimum values of acid neutralizing capacity (ANC) during episodes. An acidification episode is defined as a hydrological event during which ANC decreases. The pre‐episode ANC is used to index the antecedent condition, and the stream flow increase reflects how much the relative contributions of sources of waters change during the episode. As much as 92% of the total variation in the minimum ANC in individual catchments can be explained (with levels of explanation >70% for nine of the 13 streams) by a multiple linear regression model that includes pre‐episode ANC and change in discharge as independent variables. The predictive scheme is demonstrated to be regionally robust, with the regional variance explained ranging from 77 to 83%. The scheme is not successful for each ERP stream, and reasons are suggested for the individual failures. The potential for applying the predictive scheme to other watersheds is demonstrated by testing the model with data from the Panola Mountain Research Watershed in the South‐eastern United States, where the variance explained by the model was 74%. The model can also be utilized to assess ‘chemically new’ and ‘chemically old’ water sources during acidification episodes. Copyright © 1999 John Wiley & Sons, Ltd.     

Impact of dissipation and dispersion terms on simulations of open‐channel confluence flow using two‐dimensional depth‐averaged model

The flow patterns in confluence channel and the simulation of confluence flow are more complex than that in straight channel. Additional terms in the momentum equations, i.e. dissipation terms, denoting the impact of turbulence, and dispersion terms, denoting the vertical non‐uniformity of velocity, show great impacts on the accuracy of numerical simulations. The dissipation terms, i.e. the product of eddy viscosity coefficient and velocity gradient, are much larger than those of the flow in straight channel. In this study, the zero equation model and the depth‐averaged k‐ε model are used to analyse the impact of eddy viscosity. Meanwhile, the dispersion terms in the momentum equation, depending on the vertical non‐uniformity of velocity, are usually neglected in routine simulation. With the use of detailed experimental data for verification, this study presents the distribution of parameters of vertical non‐uniformity and the intimated connection between non‐uniformity parameters and accuracy of numerical simulations of confluence flow with depth‐averaged models. The results present that simulation accuracy of confluence flow is very sensitive to the turbulence modes, which cannot be handled by normal, simple turbulence model. On the contrary, the impact of dispersion terms is both flow‐condition‐dependent and place‐dependent, and such impact is negligible when secondary circulation is weak. The results indicate the key elements in modelling confluence flow and are helpful for selecting suitable numerical model and solving engineering problems encountered in confluence channel. Copyright © 2013 John Wiley & Sons, Ltd.     

A geomorphology‐based integrated stream–aquifer interaction model for semi‐gauged catchments

Many of the existing stream–aquifer interaction models available in the literature are very complex with limited applicability in semi‐gauged and ungauged catchments. In this study, to estimate the influent and effluent subsurface water fluxes under limited geo‐hydrometeorological data availability conditions, a simple stream–aquifer interaction model, namely, the variable parameter McCarthy–Muskingum (VPMM) hillslope‐storage Boussinesq (hsB) model, has been developed. This novel model couples the VPMM streamflow transport with the hsB groundwater flow transport modules in online mode. In this integrated model, the surface water–groundwater flux exchange process is modelled by the Darcian approach with the variable hydraulic heads between the river stage and groundwater table accounting for the rainfall forcing. Considering the exchange fluxes in the hyporheic zone and lateral overland flow contribution, this approach is field tested in a typical 48‐km stretch of the Brahmani River in eastern India to simulate the streamflow and its depth with the minimum Nash–Sutcliffe efficiency of 94% and 88%; the maximum root mean square error of 134 m³/s and 0.35 m; and the minimum index of agreement of 98% and 97%, respectively. This modelling approach could be very well utilized in data‐scarce world‐river basins to estimate the stream–aquifer exchange flux due to rainfall forcings.     

Parametric time‐domain identification of multiple‐input systems using decoupled output signals

Civil engineering structures are often subjected to multidirectional actions such as earthquake ground motion, which lead to complex structural responses. The contributions from the latter multidirectional actions to the response are highly coupled, leading to a MIMO system identification problem. Compared with single‐input, multiple‐output (SIMO) system identification, MIMO problems are more computationally complex and error prone. In this paper, a new system identification strategy is proposed for civil engineering structures with multiple inputs that induce strong coupling in the response. The proposed solution comprises converting the MIMO problem into separate SIMO problems, decoupling the outputs by extracting the contribution from the respective input signals to the outputs. To this end, a QR factorization‐based decoupling method is employed, and its performance is examined. Three factors, which affect the accuracy of the decoupling result, including memory length, input correlation, and system damping, are investigated. Additionally, a system identification method that combines the autoregressive model with exogenous input (ARX) and the Eigensystem Realization Algorithm (ERA) is proposed. The associated extended modal amplitude coherence and modal phase collinearity are used to delineate the structural and noise modes in the fitted ARX model. The efficacy of the ARX‐ERA method is then demonstrated through identification of the modal properties of a highway overcrossing bridge. Copyright © 2013 John Wiley & Sons, Ltd.     

Simulation of flow and solute coupled 2‐D problems with velocity‐dependent dispersion coefficient based on the network method

The governing equations of coupled density‐driven flow and solute transport problems in porous media, with velocity‐dependent dispersion coefficient, are strongly nonlinear and must be solved numerically. This contribution presents a network model, based on the network simulation method, capable of simulating the transient solution to this kind of problem efficiently and with a relatively low computational time. The mathematical model is formulated using the stream function and concentration variables. Simulation of the network model is carried out in the standard electric circuit simulation code, Pspice. The present model is first applied to simulate the original benchmark Henry problem, and the solution is compared with those obtained by other authors. A study of the grid size is also carried out. In addition, the modified version of Simpson and Clement of the Henry problem, as well as the groundwater flow in the closed desert basin of Pilot Valley, is studied using the proposed model. Copyright © 2011 John Wiley & Sons, Ltd.     

Stream flow forecasting using neuro‐wavelet technique

This paper proposes the application of a neuro‐wavelet technique for modelling monthly stream flows. The neuro‐wavelet model is improved by combining two methods, discrete wavelet transform and multi‐layer perceptron, for one‐month‐ahead stream flow forecasting and results are compared with those of the single multi‐layer perceptron (MLP), multi‐linear regression (MLR) and auto‐regressive (AR) models. Monthly flow data from two stations, Gerdelli Station on Canakdere River and Isakoy Station on Goksudere River, in the Eastern Black Sea region of Turkey are used in the study. The comparison results revealed that the suggested model could increase the forecast accuracy and perform better than the MLP, MLR and AR models. Copyright © 2008 John Wiley & Sons, Ltd.     

Identifying and linking source areas of flow and P transport in dairy‐grazed headwater catchments,North Island,New Zealand

We examined the applicability of the critical‐source area (CSA) concept to the dairy‐grazed 192‐ha Upper Toenepi catchment and its 8·7‐ha Kiwitahi sub‐catchment, New Zealand. We evaluated if phosphorus (P) transport from land into stream is dominated by saturation‐excess (SE) and infiltration‐excess (IE) runoff during stormflow and by sub‐surface (<1·5 m depth) flows during baseflow. We measured stream flow and shallow groundwater levels, collected monthly stream, tile drain (TDA) and groundwater samples, and flow‐proportional stream samples from the Kiwitahi sub‐catchment, and determined their dissolved reactive phosphorus (DRP) and total phosphorus (TP) concentrations. In the Kiwitahi sub‐catchment, during storm events, IE contributions were significant. Contributions from SE appeared significant in the Upper Toenepi catchment. However, in both catchments, sub‐surface contributions dominated stormflow and baseflow periods. Absence of water table at the surface and the water table gradient towards the stream indicated that P transport during events was not limited to surface runoff. The dynamics of the groundwater table and the occurrence of SE areas were influenced by proximity to the stream and hillslope positions. Baseflow accounted for 42% of the annual flow in the Kiwitahi sub‐catchment, and contributed 37 and 52% to the DRP and TP loads, respectively. The P transport during baseflow appeared equally important as P losses from CSAs during stormflow. The close resemblance in P levels between groundwater and stream samples during baseflow demonstrates the importance of shallow groundwater for stream flow. In the Upper Toenepi catchment, contributions from effluent ponds (EFFs) dominated P loads. Management strategies should focus on controlling P release from EFFs, and on decreasing Olsen P concentrations in soil to minimize leaching of P via sub‐surface flow to streams. Research is needed to quantify the role of sub‐surface flow as well as to expand management strategies to minimize P transfers during stormflow and baseflow conditions. Copyright © 2010 John Wiley & Sons, Ltd.     

Multi‐decadal variability of drought risk,eastern Australia

A number of previous studies have identified changes in the climate occurring on decadal to multi‐decadal time‐scales. Recent studies also have revealed multi‐decadal variability in the modulation of the magnitude of El Niño–Southern Oscillation (ENSO) impacts on rainfall and stream flow in Australia and other areas. This study investigates multi‐decadal variability of drought risk by analysing the performance of a water storage reservoir in New South Wales, Australia, during different climate epochs defined using the Inter‐decadal Pacific Oscillation (IPO) index. The performance of the reservoir is also analysed under three adaptive management techniques and these are compared with the reservoir performance using the current ‘reactive’ management practices. The results indicate that IPO modulation of both the magnitude and frequency of ENSO events has the effect of reducing and elevating drought risk on multi‐decadal time‐scales. The results also confirm that adaptive reservoir management techniques, based on ENSO forecasts, can improve drought security and become significantly more important during dry climate epochs. These results have marked implications for improving drought security for water storage reservoirs. Copyright © 2004 John Wiley & Sons, Ltd.     

Field investigation and modelling of coupled stream discharge and shallow water‐table dynamics in a small Mediterranean catchment (Sardinia)

In the semi‐arid Mediterranean environment, the rainfall–runoff relationships are complex because of the markedly irregular patterns in rainfall, the seasonal mismatch between evaporation and rainfall, and the spatial heterogeneity in landscape properties. Watersheds often display considerable non‐linear threshold behavior, which still make runoff generation an open research question. Our objectives in this context were: to identify the primary processes of runoff generation in a small natural catchment; to test whether a physically based model, which takes into consideration only the primary processes, is able to predict spatially distributed water‐table and stream discharge dynamics; and to use the hydrological model to increase our understanding of runoff generation mechanisms. The observed seasonal dynamics of soil moisture, water‐table depth, and stream discharge indicated that Hortonian overland‐flow was negligible and the main mechanism of runoff generation was saturated subsurface‐flow. This gives rise to base‐flow, controls the formation of the saturated areas, and contributes to storm‐flow together with saturation overland‐flow. The distributed model, with a 1D scheme for the kinematic surface‐flow, a 2D sub‐horizontal scheme for the saturated subsurface‐flow, and ignoring the unsaturated flow, performed efficiently in years when runoff volume was high and medium, although there was a smoothing effect on the observed water‐table. In dry years, small errors greatly reduced the efficiency of the model. The hydrological model has allowed to relate the runoff generation mechanisms with the land‐use. The forested hillslopes, where the calibrated soil conductivity was high, were never saturated, except at the foot of the slopes, where exfiltration of saturated subsurface‐flow contributed to storm‐flow. Saturation overland‐flow was only found near the streams, except when there were storm‐flow peaks, when it also occurred on hillslopes used for pasture, where soil conductivity was low. The bedrock–soil percolation, simulated by a threshold mechanism, further increased the non‐linearity of the rainfall–runoff processes. Copyright © 2013 John Wiley & Sons, Ltd.     

Soil moisture states,lateral flow,and streamflow generation in a semi‐arid,snowmelt‐driven catchment

Hydraulic connectivity on hillslopes and the existence of preferred soil moisture states in a catchment have important controls on runoff generation. In this study we investigate the relationships between soil moisture patterns, lateral hillslope flow, and streamflow generation in a semi‐arid, snowmelt‐driven catchment. We identify five soil moisture conditions that occur during a year and present a conceptual model based on field studies and computer simulations of how streamflow is generated with respect to the soil moisture conditions. The five soil moisture conditions are (1) a summer dry period, (2) a transitional fall wetting period, (3) a winter wet, low‐flux period, (4) a spring wet, high‐flux period, and (5) a transitional late‐spring drying period. Transitions between the periods are driven by changes in the water balance between rain, snow, snowmelt and evapotranspiration. Low rates of water input to the soil during the winter allow dry soil regions to persist at the soil–bedrock interface, which act as barriers to lateral flow. Once the dry‐soil flow barriers are wetted, whole‐slope hydraulic connectivity is established, lateral flow can occur, and upland soils are in direct connection with the near‐stream soil moisture. This whole‐slope connectivity can alter near‐stream hydraulics and modify the delivery of water, pressure, and solutes to the stream. Copyright © 2005 John Wiley & Sons, Ltd.     

Modification of the SWAT model to simulate regional groundwater flow using a multicell aquifer

The soil and water assessment tool (SWAT) has been widely used and thoroughly tested in many places in the world. The application of the SWAT model has pointed out that 2 of the major weaknesses of SWAT are related to the nonspatial reference of the hydrologic response unit concept and to the simplified groundwater concept, which contribute to its low performance in baseflow simulation and its inability to simulate regional groundwater flow. This study modified the groundwater module of SWAT to overcome the above limitations. The modified groundwater module has 2 aquifers. The local aquifer, which is the shallow aquifer in the original SWAT, represents a local groundwater flow system. The regional aquifer, which replaces the deep aquifer of the original SWAT, represents intermediate and regional groundwater flow systems. Groundwater recharge is partitioned into local and regional aquifer recharges. The regional aquifer is represented by a multicell aquifer (MCA) model. The regional aquifer is discretized into cells using the Thiessen polygon method, where centres of the cells are locations of groundwater observation wells. Groundwater flow between cells is modelled using Darcy's law. Return flow from cell to stream is conceptualized using a non‐linear storage–discharge relationship. The SWAT model with the modified aquifer module, the so‐called SWAT‐MCA, was tested in 2 basins (Wipperau and Neetze) with porous aquifers in a lowland area in Lower Saxony, Germany. Results from the Wipperau basin show that the SWAT‐MCA model is able (a) to simulate baseflow in a lowland area (where baseflow is a dominant source of streamflow) better than the original model and (b) to simulate regional groundwater flow, shown by the simulated groundwater levels in cells, quite well.