Contributors: Yeshewatesfa Hundecha (SMHI), Peter Berg (SMHI), René Capell (SMHI), Christiana Photiadou (SMHI), Lisanne Nauta (WUR), Fulco Ludwig (WUR), Eveline van der Linden (WUR), Wietse Franssen (WUR), Jan Biermann (WUR), Sinclair Chinyoka (WUR)
Acronyms
Scope of the document
This document presents the setup and performance of an ensemble of hydrological models used for the Climate Impact Indicators (CII) calculations for the data set entitled "Hydrology related climate impact indicators from 1970 to 2100 derived from bias adjusted European climate projections".
The ensemble of hydrological models come at two different spatial resolutions: catchment and 5km grid. Due to the different drainage networks used for each resolution, the ensemble is treated separately for the two resolutions, and a multi-model ensemble is available for both.
This document describes the two hydrological models E-HYPE and VIC-WUR, their different setups and their performance evaluation.
The E-HYPE model
The E-HYPE model (Donnelly et al., 2016) is based on the continuous process-based hydrological model HYPE (HYdrological Predictions for the Environment; Lindström et al. 2010), which simulates components of the catchment water cycle and water quality. The model is semi-distributed, in which a river basin may be subdivided into multiple subcatchments, which are further subdivided into hydrological response units (HRUs) based on soil type and land use classes. It has conceptual routines for the major hydrological surface and subsurface processes (e.g., snow/ice accumulation and melting, evapotranspiration, surface and macropore flow, soil moisture, runoff generation, groundwater fluctuation, routing through rivers and lakes), land management , (irrigation, abstractions), and nutrient turnover (diffuse and point source releases, solid-matter and dissolved sub-surface pools, plant uptake, riverine transport) that are controlled by a number of parameters that are often linked to physiography to account for spatial variability and estimated through calibration. More information on HYPE model details can be found at https://hypeweb.smhi.se/model-water/
.
The model parameters can generally be categorized into HRU and general parameters. The HRU parameters are either soil or land use type dependent. Unique parameter values are estimated for each soil or land use type and are applied throughout the model domain. The general parameters, such as the routing parameters, are subcatchment scale parameters and may be assigned constant values throughout the model domain or estimated separately for different parameter regions. They can also be estimated using a regionalization approach as functions of catchment.
The model was setup for the pan-European domain and is referred to as E-HYPE (Donnelly et al. 2016). The version used here covers an area of 8.8 million km2 and is subdivided into 35,408 subcatchments with an average size of 248 km2. A range of physiographic data sources were used in setting up the model (Donnelly et al. 2016) and detailed information can be found at https://hypeweb.smhi.se/explore-water/. All E-HYPE models are driven by daily mean temperature and precipitation.
E-HYPEcatch: A multi-model ensemble
E-HYPEcatch is the original E-HYPE model with irregular polygons that delineate the sub-catchment. The shape files for the E-HYPEcatch models can be found here: http://doi.org/10.5281/zenodo.581451.
In order to account for the uncertainty in the estimated model parameters, a multi-model ensemble was generated. This was performed by first identifying the most sensitive parameters for the model calibrated using the procedure described in Hundecha et al. (2020). The dominant soil and land use parameters corresponding to the eight catchment groups, along with the general parameters were sampled using the SAFE sensitivity and uncertainty analysis toolbox (Pianosi et al. 2015) to generate 10000 randomly distributed samples with the latin hypercube strategy. The model performance was assessed in terms of NSE.
Parameter samples whose model performance was above a subjectively defined threshold (top 1% of the 10000 samples) were accepted based on a GLUE-type parameter identification approach (Beven and Binley, 1992). In the next step, the top 1% samples were combined across the eight groups to provide narrower identifiable parameter ranges and 15000 sets were sampled within these ranges with the same strategy as before. These new global sets were then evaluated against all calibration stations and parameter sets were chosen as in the previous step, providing a number of acceptable global parameter sets. Finally, the best 10 sets were selected as the multi-model ensemble, herein referred to as M01-M10.
E-HYPEgrid: Setting up a grid-based pan-European HYPE model
In order to set up E-HYPE at a 5km grid resolution over Europe with the same drainage network as the EFAS modeling system, the following data was obtained from EFAS: flow direction, mean elevation, fraction of water bodies, lake-mask, and stream channel length within each grid cell as well as the total upstream area draining to each grid cell.
Each grid cell is treated as a subcatchment, as in the standard E-HYPEcatch model. Hydrological response units are defined within each grid cell as a combination of land use and soil type, using the same classification as E-HYPEcatch.
In order to exploit HYPE's capability of modeling lake dynamics with detailed natural and regulation lake processes, large lakes are allowed to extend over several grid cells. To accommodate for this, lake areas were added as additional computation grid point during the E-HYPEgrid setup and calibration phase using European and global lake and reservoir databases. Large lakes in E-HYPEgrid cover one or several EFAS grid cells completely, and surrounding grid cells partially, while smaller lakes and elongated river dams cover several grid cells partially. Since model results are calculated only for the outlet point of the lake, this point is representing the whole lake and any partial land computation cell will be replaced by the lake cell, following the reasoning that the water flows through the lake unit. To deliver E-HYPEgrid results on the original EFAS grid, the lake values have replaced the land values when the lake is only partial, following the reasoning that the water flows through the lake unit.
Climate simulations with E-HYPE
All versions of the E-HYPE models use the same approach to climate simulations. The driving regional climate model data of daily mean temperature and precipitation are first bias adjusted. Because the EFAS-Meteo reference data for the bias adjustment covers the period 1990-2018, which overlaps the initiation of the RCP-scenarios in 2006, there will be slight differences in the bias adjustment parameters depending on the scenario used. E-HYPEcatch and E-HYPEgrid are therefore simulating the reference period separately for each RCP, with slight differences due to the bias adjustment. The differences are, however, very small and the published data in the CDS catalogue therefore only provides a single historical data set which is that of historical extended with RCP4.5.
For E-HYPEcatch models, the bias adjusted 5km data were first mapped to the catchments by averaging using geometric weights based on the overlap of the 5 km side grid box with the catchment delineation. E-HYPEgrid uses one-to-one mapping between the grids, except for lakes that cover more than one grid point, for which the same mapping as for E-HYPEcatch is performed to map the average result to the outflow grid cell of the lake.
HYPE requires a spin-up time to equilibrate model states. Most hydrological states equilibrate quickly, i.e. after a few weeks of modelled data, but some states may take longer because of seasonal weather pattern (e.g. snow cover, dry season) or slow turn-over times (e.g. large lakes). Conventionally, spin-up is performed using a combination of a pre-computed static spin-up state with a dynamic spin-up period. The static spin-up state is computed once for a model domain over a multi-year period using some appropriate forcing data, and provides the starting point for a shorter, month to year-long dynamic spin-up with the forcing data used in the model simulation. This approach saves significant amounts of computation time compared to full dynamic spin-up. For climate simulations here, the models were initialised using a static state only. This approach will leave an imprint on model results for some time. However, a quality assessment showed that the imprint diminishes largely after a few weeks, and was deemed negligible particularly for the calculation of impact indicators, which are based on changes between long time windows.
The nitrogen and phosphorus modules within E-HYPEcatch use matter-bound soil pools for nutrient turnover modeling. These pools are much larger than the modeled turnover fluxes, by several orders of magnitude, and essentially unknown, and thus are an uncertain factor in the model concept. In addition, nutrient amounts in pools can drift unbounded in long model runs, which may or may not correspond to realistic changes. For transient climate change impact model runs with E- HYPEcatch, we therefore analysed soil pool dynamics using synthetic forcing data and developed a two-part strategy to minimize effects of soil pool instability: (1) Model runs were initialized with a 20-year spin-up period to stabilize soil pools after initial blind guesses, and (2) matter-bound nutrient pools were repeatedly re-set to the initialized state after 10 years during the transient climate runs. Re-setting at this frequency showed no impact on modeled in-stream nutrient dynamics. Climate impact assessments on riverine nutrients thus explicitly exclude impacts of potential soil pool changes because E-HYPEcatch's unknown ability to reproduce such effects.
Because of differences between E-HYPE and VIC-WUR definitions of soil moisture and evapotranspiration, we specifically define them for E-HYPE models as:
- Soil moisture content is computed as fraction of water content at field capacity in the root zone. Water content at field capacity in HYPE is a model parameter tied to soil class. The root zone, i.e. the plant-available water content of HYPE-modelled sub-surface water, is defined as the upper two sub-surface layers of a HYPE soil class. Soil classes consist of up to 3 computational layers.
- Evapotranspiration: Potential evapotranspiration (PET) from land and lake surface is estimated based on a modified Jensen-Haise/McGuinness model with parameters tied to land use classes. Actual evapotranspiration (AET) is computed as a fraction of PET, depending on soil moisture content. AET is zero for temperatures below snow threshold, and snow sublimation is not considered in the model.
The VIC-WUR model
The Variable Infiltration Capacity (VIC) model (Liang et al., 1994) is a semi-distributed macroscale hydrological model and forms the fundamentals of the VIC-WUR model. The VIC-WUR (version 2.0.0) model is an extension of the VIC-5 model and includes additional modules to account for anthropogenic influences (Hamman et al., 2018, Droppers et al., 2020). A conceptual overview of the model is presented in Figure 1. The model was been setup in natural flow conditions and water balance mode with three thermal nodes. Lakes and frozen soils are not simulated. A single elevation band is used, because heterogeneity in topography is preserved by the high spatial resolution of the model. Sub-daily timesteps of 6 hours are used to solve the energy and water balance for each cell individually; i.e. there is no water or energy exchange between cells. The snow module also uses a 6 hourly timestep and calculates the snow variables in the surface and vegetation layers. Soil parameters and vegetation characteristics (Lawrence and Chase, 2007) are used from respectively 0.5 and 0.05 degrees gridded data and remapped to the model domain using ECMF nearest neighbour algorithm. The parameter file includes 3 vertical soil layers and 16 vegetation classes. There is surface runoff whenever precipitation occurs under saturated top soil conditions, while precipitation under unsaturated conditions infiltrates into the soil layers. The upper two soil layers interact with the canopy layer and provide water for transpiration. The third and deepest soil layer is assumed to be outside the root zone and accounts for subsurface runoff (baseflow). VIC-WUR uses an internal routing module (Lohmann et al, 1996, 1998) to simulate discharge based on the unit hydrographs of both stream inflow and outflow. The unit hydrographs are computed from the EFAS flow direction, channel length, mean elevation (slope) and the SCS dimensionless unit hydrograph using a flow velocity of 1.5 m/s and diffusion value of 2000 m/s2.
Inputs to the VIC-WUR model are 6h temporal resolution of temperature, precipitation, solar radiation, thermal radiation, wind speed, surface pressure and humidity. As data at this high temporal resolution and not available from the observational data set EFAS-Meteo, nor from the climate models, the available data were disaggregated using different methods depending on the use case.
Figure 1: Schematic presentation of the water and energy fluxes in the VIC-WUR model. Source: https://vicwur.readthedocs.io/
Climate simulations with VIC-WUR
In total 8 historical simulations were conducted for the period 1970-2005 with 1970 discarded as a spin up period for soil moisture in the third layer and snowpack in mountainous areas. It is important to highlight that a single historical simulation was conducted with the bias adjustment based on a combination of the historical and RCP4.5 data for the period 1990-2018 (see description for the HYPE model for more details). The historical simulations generated the model restart files that were used for all scenarios for the period 2006-2100.
To transfer the daily input data to the required 6h timesteps, the data was disaggregated. The disaggregation of bias adjusted precipitation equally distributed daily precipitation over 6 hourly intervals. However, bias adjusted temperature was treated differently to factor in the diurnal variation of air temperature. First, daily bias correction factors were calculated from adjusted and non-adjusted temperature data. Second, 6 hourly temperature data was calculated from daily minimum and maximum near surface temperatures using the metGeneratoR package (Franssen 2020) which is based on the principles of Bohn et al. (2013). Finally, daily correction factors were added to the generated sub daily temperature data in order to preserve the bias correction. Other required forcing data were disaggregated to 6 hourly intervals using daily GCM data and the metGeneratoR software. More information on bias correction can be found in C3S_424_SMHI_3.1b_Biasadjustment.Because of differences between E-HYPE and VIC-WUR definitions of soil moisture and evapotranspiration, we specifically define them for VIC-WUR as:
- Soil moisture output (OUT_ROOTMOIST) from the VIC-WUR model is defined as the soil moisture content in the upper two soil layers which contain roots. The root zone layer is under influence of transpiration by root uptake of the vegetation layer. The model uses 3 interconnected soil layers in total and the layers may vary in depth.
- The evapotranspiration calculated in VIC-WUR includes evaporation from bare soil and vegetation, transpiration from vegetation and snow sublimation from surface and vegetation. The evapotranspiration is calculated for each soil layer using the Penman- Monteith approach which considers the varying canopy resistance due to different vegetation types, soil types, soil moisture and atmospheric conditions. Evapotranspiration takes place at potential rate until the water availability becomes limited. The evapotranspiration might become negative under specific conditions.
Evaluation of the model ensemble over Europe
The E-HYPE models (catchment and grid versions separately) and the VIC-WUR model were evaluated using thion of the performance of the E-HYPE and VIC-WUR models setup at a 5km grid will be presented. The performance of the grid version of E-HYPE will also be compared with the corresponding performance of the catchment resolution E-HYPE and its multi-model extension.
Method for performance evaluation
To assess the model performance at each of the discharge gauging stations used for model evaluation, we use three model efficiency measures: the Nash-Sutcliffe Efficiency (NSE), Kling-Gupta Efficiency (KGE), and percentage model bias (Pbias). In addition, the model performance in terms of reproducing the extremes of daily flow is assessed by evaluating how well the model captures the 5th and 95th percentile of the long-term daily flow. All models are driven by the EFAS-Meteo forcing data over the period 1990-2019, and are evaluated at 352 selected validation stations across Europe.
All variants of the E-HYPE model were calibrated using a set of stations, which are independent from the validation stations. The number of stations used for calibration of the catchment resolution model (E-HYPEcatch) and members of its multi-model extension (M01-M10) and the grid version of the model (E-HYPEgrid) are different. There are 52 calibration stations that are common to all variants of the model and the models' performance at these stations is also evaluated.
In addition to bias adjusted temperature and precipitation from the EFAS-Meteo dataset, the VIC- WUR model was forced by non-bias adjusted solar radiation, thermal radiation, wind speed, surface pressure and humidity forcing from the ERA5 dataset. ERA5 data was aggregated from 1-hourly to 6-hourly intervals, while EFAS-Meteo data was disaggregated from daily values. Vapor pressure forcing was calculated by substituting dewpoint temperature through the Clausius–Clapeyron equation.
Performance of E-HYPE
Calibration of all variants of the E-HYPE model was performed for the period 1991 – 2002 with an objective of maximizing the overall Nash-Sutcliffe Efficiency (NSE) at the stations used for calibration (Hundecha et al., 2016). The models were then evaluated at the validation set of stations over the same time period. Since the available discharge data after 2002 at many of both the calibration and validation sets of stations is not enough to allow a reasonable temporal model validation, temporal model validation over a non-overlapping time period was not performed. Instead, the models were evaluated over the entire forcing data period (1990-2019).
As shown in Figure 2, model calibration resulted in a median NSE of 0.6 and 0.53, respectively, for the E-HYPEcatch and E-HYPEgrid models at the 52 calibration stations over the calibration period (1991-2002). For the 10 members of the multi-model ensemble extension of the E-HYPEcatch model, the median NSE ranges between 0.51 and 0.56. As shown in Figure 2, the distribution of the KGE for each model is similar with that of the corresponding NSE, with a slightly higher median value and less spread across the stations. All variants of the E-HYPE generally underestimate the long-term flow at many of the calibration stations over the calibration period (Figure 3). However, the median Pbias at the 52 calibration stations is within 10% for both E-HYPEcatch and E-HYPEgrid. The multi-model extension of the E-HYPEcatch generally underestimates the long-term flow more than both the E-HYPEcatch and E-HYPEgrid at the calibration stations over the calibration period (see Figure 3).
Evaluation of the models at the independent validation stations over the same period over which calibration was performed (1991-2002) shows that in terms of NSE, E-HYPEcatch and its multi- model extension perform worse than in the calibration stations with median values across all the 352 stations of 0.53 for E-HYPEcatch and between 0.33 and 0.49 for members of its multi-model extension (Figure 2). The spread of NSE across the stations is higher in the validation stations, meaning that there are stations with higher NSE values than the maximum NSE in the calibration stations, but also stations with smaller values than the minimum in the calibration set of stations. For E-HYPEgrid, the median NSE is comparable with the corresponding value in the calibration stations (0.52) but the spread across the stations is higher. In terms of KGE, E-HYPE grid generally performs better in the validation stations than in the calibration stations, with a median value of 0.61, but with more spread across the stations (Figure 3). The median KGE for E-HYPEcatch in the validation stations is similar with that of the calibration stations (0.61), but with more spread across stations. The median KGE for each member of the multi-model extension of E-HYPEcatch is consistently lower in the validation stations than in the calibration stations, with more spread across the stations. Interestingly, the distributions of NSE and KGE in the validation stations over the model calibration period are similar for E-HYPEgrid and E-HYPEcatch (see Figures 1 and 2). In terms of Pbias, both E-HYPE grid and E-HYPEcatch slightly overestimate the long-term flow at many of the validation stations, with median values of 3.3% and 7.6%, respectively, over the calibration period (Figure 3). Members of the multi-model extension of E-HYPEcatch perform well in the validation stations in terms of Pbias, with a median value of around 0% and less spread across stations compared with the corresponding spread in the calibration stations.
Figure 2: Distributions of NSE across the calibration and validation stations for the different variants of E- HYPE (For each group of stations, left boxes are for the calibration period and right boxes are for the entire model period).
Figure 3: Distributions of KGE across the calibration and validation stations for the different variants of E- HYPE (For each group of stations, left boxes are for the calibration period and right boxes are for the entire model period).
Figure 4: Distributions of Pbias across the calibration and validation stations for the different variants of E- HYPE (For each group of stations, left boxes are for the calibration period and right boxes are for the entire model period).
The model performance over the entire modelling period (1990-2019) is generally similar with the corresponding performance in each group of stations (calibration and validation stations) in the model calibration period for both E-HYPEgrid and E-HYPEcatch, as well as for many of the multi- model ensemble members, with some slight differences with respect to the different metrices (Figures 1-3). Generally, E-HYPE catch performs better in terms of NSE and KGE than E-HYPEgrid in the calibration stations but their performances in the validation stations are similar. The performance of members of the multi-model extension of E-HYPEcatch in both the calibration and validation stations is generally lower than the corresponding performance of E-HYPEcatch.
However, their performance in the calibration stations is better than the corresponding performance of E-HYPEgrid. In the validation stations, E-HYPEgrid performs better, with comparable performance with E-HYPEcatch.
E-HYPEgrid underestimates the long-term flow slightly more in the calibration stations than E- HYPEcatch. In the validation stations, however, E-HYPEcatch overestimates the long-term flow while E-HYPEgrid captures it well, with a median value of Pbias close to 0%. E-HYPEcatch predicts systematically higher discharge relative to E-HYPEgrid (Figure 4). Members of the multi-model extension of E-HYPEcatch underestimate the long-term flow in the calibration stations more strongly than both E-HYPEcatch and E-HYPEgrid. In the validation stations, however, they capture the long-term flow well, with a similar performance as E-HYPEgrid (Figure 4).
Figure 5 shows the spatial pattern of performance of E-HYPEcatch and E-HYPEgrid in the calibration stations in terms of NSE and KGE. It shows that there is no clear spatial gradient in the model performance. Where one of the models performs well, the other also performs well. E-HYPEcatch shows a better performance at several stations. There is no clear spatial gradient in the performance of both models in terms of Pbias too, as shown in Figure 6. Both models have a more or less similar spatial pattern in performance.
Figure 5: Spatial distribution of NSE and KGE in the calibration stations over the entire modeling period (1990-2019).
Figure 6: Spatial distribution of Pbias in the calibration stations over the entire modeling period (1990-2019).
The model performance in terms of NSE and KGE in the validation stations shows a spatial pattern that tends to be best in the north-west of the model domain and deteriorates towards the south- east for both models (Figure 7). In terms of Pbias, the two models do not show a similar pattern and the performance in both models is spatially variable with no clear pattern (Figure 8). Both models overestimate the long-term flow in south-western Russia and E-HYPEgrid underestimates the long- term flow in the Seine basin in France.
Figure 7: Spatial distribution of NSE and KGE in the validation stations over the entire modeling period (1990- 2019).
Figure 8: Spatial distribution of Pbias in the validation stations over the entire modeling period (1990-2019).
Evaluation of the two models in terms of their ability in reproducing the extremes of daily flow (the 5th and 95th percentiles of the long-term daily flow) in the validation stations shows that both models reproduce the extremes fairly well with comparable skill (Figure 9). Both models generally underestimate the high flows (the 95th percentile flow) in bigger catchments while they do well in smaller catchments. E-HYPEgrid generally underestimates the low flows (the 5th percentile flow) but E-HYPEcatch underestimates the low flows in bigger catchments and overestimates them in smaller catchments.
Figure 9: Scatter plot of modelled and observed 5th and 95th percentiles of long-term daily flow at the validation stations.
In conclusion, while both E-HYPEcatch and E-HYPEgrid are superior to the members of the multi- model extension of E-HYPEcatch, the two models have a competing performance. E-HYPEcatch, generally has achieved high NSE and KGE values at more stations than E-HYPEgrid but E-HYPEcatch displays a stronger wet bias at more stations than E-HYPEgrid.
Performance of VIC-WUR
The VIC-WUR model was run for the period 1990 -2018. The first year (1990) was used as model spin up time to create an initial state. The remaining period (1991-2018) is used for model validation. No observed discharge was available for 29/353 selected validation stations during this period. The calibration of VIC-WUR differs substantially from E-HYPE, in that it is physically based, and no direct calibration on the discharge is performed. The model calibration has been performed globally at a resolution of 0.5 degrees using a different dataset many years ago (Nijssen et al. 2001). This model calibration was reused for this setup.
Figure 9 shows the distribution of the three model performance indicators at the validation stations. Median values of 0.25, 0.44 are found for respectively the NSE, KGE. The NSE contains negative values in the interquartile range, suggesting that the mean observed discharge is a better predictor than the model for those stations. The model performs better in terms of KGE rather than NSE. The KGE is less influenced by time lags in the model prediction and more by good predictions over a longer period. A time lag or under/over estimations in predicted peak discharges could be an explanation for the large negative NSE values. The median Pbias is 31% and the interquartile range indicates that the model overestimates the mean discharge in general.
Figure 10: The distribution of NSE, KGE and Pbias model performance indicators calculated from the validation stations over the period 1991-2017. Not all outliers fit in the plot area.
The spatial distribution of the performance indicators are shown in Figure 10, Figure 12, and Figure 13, for NSE, KGE and Pbias, respectively. All indicators show more or less the same spatial pattern, where the model performs worse in the eastern part of Europe compared to other areas. Also, in a lot of cases, similar performance levels are present in clusters. This might indicate that there is a relation between particular spatial/basin characteristics and model performance.
Figure 11: The spatial distribution of NSE in the stations over the validation period (1991-2018).
Figure 12: The spatial distribution of KGE in the stations over the validation period (1991-2018).
Figure 13: The spatial distribution of KGE in the stations over the validation period (1991-2018).
Analysis of the model performance during extreme low and high flow events show that VIC-WUR captures the extreme high flows well, but underestimates the extreme low flow events in almost all cases (Figure 14). During extreme high flow events (>95th percentile), VIC-WUR tends to overestimate the smaller basins and underestimate the larger basins. For the low flow, however, there still is a moderate correlation between observed and measured low flows.
Figure 14: Scatter plot of observed and modelled 5th and 95th percentiles of long-term daily specific discharges at the validation stations over the period 1991-2018. The Pearson correlation is shown in the upper-left corner.
References
Beven K, Binley A (1992) The Future of Distributed Models: Model Calibration and Uncertainty Prediction. Hydrol. Process. 6 (3): 279–98.
Bohn, Theodore J., et al. (2013), 'Global evaluation of MTCLIM and related algorithms for forcing of ecological and hydrological models', Agricultural and Forest Meteorology, 176, 38-49.
Burek, P., Van der Knijff, J., and de Roo, A. (2013) LISFLOOD Distributed Water Balance and Flood Simulation Model. JRC Technical Reports. Doi:10.2788/24719
Donnelly, C., Andersson, J. C. M., & Arheimer, B. (2016) Using flow signatures and catchment similarities to evaluate the E-HYPE multi-basin model across Europe. Hydrol. Sci. J. 61(2), 255-273. doi:10.1080/02626667.2015.1027710
Droppers, B., W. H. P. Franssen, M. T. H. van Vliet, B. Nijssen, and F. Ludwig. 2020. Simulating human impacts on global water resources using VIC-5. Geosci. Model Dev. 13:5029-5052. https://doi.org/10.5194/gmd-13-5029-2020
Hamman, J. J., Nijssen, B., Bohn, T. J., Gergel, D. R., and Mao, Y. (2018) The Variable Infiltration Capacity model version 5 (VIC-5): infrastructure improvements for new applications and reproducibility, Geosci. Model Dev., 11, 3481-3496, https://doi.org/10.5194/gmd-11-3481-2018.
Hundecha, Y., Arheimer, B., Donnelly, C., and Pechlivanidis, I. (2016) A regional parameter estimation scheme for a pan-European multi-basin model. J. Hydrol.-Regional Studies. 6, 90-111, doi: 0.1016/j.ejrh.2016.04.002
Hundecha, Y., Arheimer, B., Berg, P., Capell, R., Musuuza, J., Pechlivanidis, I., & Photiadou, C. (2020). Effect of model calibration strategy on climate projections of hydrological indicators at a continental scale. Climatic Change, 163(3), 1287-1306.
Liang, X., D. P. Lettenmaier, E. F. Wood, and S. J. Burges (1994), A simple hydrologically based model of land surface water and energy fluxes for general circulation models, J. Geophys. Res., 99(D7), 14415–14428, doi:10.1029/94JD00483.
Lindstöm G, Pers C, Rosberg J, Strömqvist J, Arheimer B (2010) Development and testing of the HYPE (Hydrological Predictions for the Environment) water quality model for different spatial scales.
Hydrol Res 41(3-4):295-319. doi:10.2166/nh.2010.007
Lohmann, D., R. Nolte-Holube and E. Raschke (1996). "A large-scale horizontal routing model to be coupled to land surface parametrization schemes." Tellus A 48(5): 708-721.
Lohmann, D., E. Raschke, B. Nijssen and D. P. Lettenmaier (1998). "Regional scale hydrology: I. Formulation of the VIC-2L model coupled to a routing model." Hydrological Sciences Journal 43(1): 131-141.
Nijssen, Bart, et al. (2001), 'Predicting the Discharge of Global Rivers', Journal of Climate, 14 (15), 3307-23.
Ntegeka V, Salamon P, Gomes G et al (2013) EFAS-Meteo: A European daily high-resolution gridded meteorological data set for 1990–2011. Report EUR, 26408.
Pianosi F, Fanny S, Wagener T (2015) A Matlab Toolbox for Global Sensitivity Analysis. Environ Modell Softw 70: 80–85.
