Climate change impact on river flows and catchment hydrology: a comparison of two spatially distributed models

Abstract

Hydrological models have been widely used to assess changes in stream discharges by climate change; however, concern might arise on the accurateness of the model predictions under the changed conditions particularly during low flow periods. In this study, two spatially distributed hydrological models MIKE SHE and WetSpa, each representing a different model complexity in terms of process description, data needs, parameter space, degree of calibration, were compared in their estimation of the climate change impact on the flow regimes in a medium-sized catchment in Belgium. The fully integrated, physically based MIKE SHE model, comprising a three-dimensional groundwater flow and river model, was applied to better understand the groundwater flow and groundwater-river interactions under changed climate conditions. Both models were able to capture the flow dynamics very well with high efficiencies and simulated the flow extremes very accurately. The groundwater heads in MIKE SHE and their seasonal variation had a high model performance. The two models simulated similar changes to future flows because of climatic changes. Peak flows were expected to increase or decrease depending, taking the large uncertainty in future climate trends into account. The model structural uncertainties on these high flow predictions were rather limited. Low flows were expected to reduce because of drier conditions by future climate change, indicating elevated low flow risks for Belgium. However, the projected low flow changes differed significantly over the models and even exceeded the uncertainty by the expected climate trends. Smaller impact was predicted by considering the groundwater physics and river interaction in the MIKE SHE model. These changes to the surface water regimes were verified by the changes in groundwater heads. Future groundwater conditions also point towards drier conditions with a small decrease in heads for future summer and autumn periods. Projected variations in winter heads depend on the climate scenario.

Significant uncertainty in global scale hydrological modeling from precipitation data errors

“…Only if one single forcing dataset is employed, as for example in a flow forecasting system (Candogan Yossef et al, 2011, it may be feasible to estimate the parameters from discharge and/or other data. In addition the current study used one single hydrological model, focusing on parameter and forcing uncertainty and ignoring hydrological model uncertainty, whereas several studies demonstrated that different models may respond differently to (climate) changes in forcing data (Vansteenkiste et al, 2012;Ficklin and Barnhart, 2014). The current study should be repeated with other GHMs to test whether the statement that forcing uncertainty is larger than parameter uncertainty is in general valid.…”

Understanding the Effects of Parameter Uncertainty on Temporal Dynamics of Groundwater-Surface Water Interaction

“…For each scenario, three CRCM outputs were averaged as a mean. A similar approach was applied in various hydrological studies [3,8]. Then those outputs were distributed as daily values in the particular month by using the delta change method.…”

“…During the last decade, many researchers have used different hydrologic models to quantify these exchange processes. For instance, Scibek and Allen [3] model; Van Roosmalen et al [4] used the DK model (The National Water Resource model for Denmark); Goderniaux et al [5] used the HydroGeoSphere model; Stoll et al [6] used the MIKE Système Hydrologique Européen (MIKE-SHE) model; Jackson et al [7] used the coupled Zoom Object-Oriented Distributed Recharge Model (ZOODRM) and Zoom Object-Oriented Quasi-3-Dimensional Model (ZOOMQ3D); Vansteenkiste et al [8] used the MIKE-SHE, and Water and Energy Transfer between Soil, Plants and Atmosphere (WetSpa) models; El Hassan et al [9] used the Gridded Surface Subsurface Hydrologic Analysis (GSSHA) model; Wu et al [10] used Groundwater and Surface-water FLOW (GSFLOW) model; Faramarzi et al [11] used the Soil and Water Assessment Tool (SWAT) model. Most of the parameters (e.g., soil properties, surface roughness) in hydrologic models used for GW-SW interaction simulations require intensive field measurements [12], and they are always associated with uncertainty.…”

Nitrogen and phosphorus transformations and balance in a pond-ditch circulation system for rural polluted water treatment

“…The coupling of unsaturated zone and saturated zone, which is not involved in the conceptual models, efficiently integrates the surface water and groundwater on a physical basis, allowing it to meet the requirements of simulating regions of more complicated hydrological processes (Christiaens and Feyen, 2001;Doummar et al, 2012;Sørensen, 1996) and allowing for the solution of more comprehensive hydrological issues of water resources management. With respect to the hydrological modeling of landscape changes, the changed parameters of MIKE SHE, unlike in previous conceptual models, have clear physical meanings and detailed input and output clearly articulate the spatial heterogeneity and time continuity of different hydrological processes, making hydrological forecasts more realistic (Kalantari et al, 2014;Sultana and Coulibaly, 2011;Vansteenkiste et al, 2013;Wijesekara et al, 2014).…”

“…The best model combination methods are a trimmed mean (constructed from the central four or six predictions each day) and a weighted mean ensemble (with weights calculated from calibration performance) that places relatively large weights on the better performing models. Furthermore, two distributed models of MIKE SHE and WetSpa were compared for their estimation of climate change impact on flow regimes in a medium-sized catchment in Belgium and it was found that the model structural uncertainties for high flow predictions were rather limited, but the projected low flow changes differed significantly over the models and even exceeded the uncertainty by the expected climate trends (Vansteenkiste et al, 2013).…”