Supplementary Material for

Rapid screening of operational freshwater availability using global models

M.W. Straatsmaa,d, P.T.M. Vermeulenb, M.J.M. Kuijperb, M. Bontec, F.G.M. Nielec, M.F.P.Bierkensa,b

a Faculty of Geosciences, Department of Physical Geography, Utrecht University, PO Box 80115, 3508 TC, Utrecht, The Netherlands

b Deltares, PO Box 85467, 3584 CB, Utrecht, The Netherlandsc Shell Global Solutions, PO Box 60, 2280 AB, Rijswijk, The Netherlands

d corresponding author m.w.straatsma@uu.nl

This pdf includes:

1. Study area

2. Methods

3 Validation and well yield comparison between a global and national groundwater model

Figs. S1 tot S5

Tables S1 to S2

Full reference list

1

2

3

456789

10

11

12

13

14

15

16

17

18

19

1. Study area

In this research, we considered two study areas. The Lower Rhine Meuse Estuary (LRME) in The Netherlands, and the Lower Buffalo Bayou area in the United States of America (Fig. S1; Table S1). The gauging stations for LRME are located 151 and 85 km upstream for Lobith and Lith, respectively. The closest gauging station for the LBB is along the West Fork of the San Jacinto River, 70 km upstream from the confluence with the Buffalo Bayou.

Fig. S1 Study area around the asset of five by five cells, sized 5 arcminutes. The asset outlines (Fig. 1) are skewed due to the geographic coordinate system of the figure. Gauging stations are located outside of the map extent; see Table S1 for northing and easting of the stations.

Table S1 River runoff gauging stations of the GRDC and USGS closest to the study areas

ID Source River StationShell Asset

Nor-thing

Eas-ting

basin size (km2) start end

Q mean

(m3/s)a

6421100 GRDC Meuse Lith MP 51.8 5.5 29000 1911 1995 3256435060 GRDC Rhine Lobith MP 51.8 6.1 160800 1901 2012 22208068090 USGS San Jacinto Porter, TX DP 30.1 -95.3 2491 1984 2013 20a Annual average river runoff

2. Methods

2.1 Computation of Operational Freshwater Availability

To compute Operational Freshwater Availability (OFWA), we based our modeling choices on the objective to obtain a central or ‘middle of the road’ estimate of the OFWA. That meant that we did not use extreme climate models, extreme climate change scenarios, or extreme socioeconomic scenarios for the future. We combined two hydrologic models: PCR-GLOBWB for the surface and shallow subsurface, and MODFLOW for the deep groundwater. The PCRaster Global Water Balance and water resources model (PCR-GLOBWB) was used to simulate water storages and fluxes of the terrestrial part of the hydrological cycle. PCR-GLOBWB is a grid-based global hydrological model (Van Beek et al. 2011) with five arcminute (5’) cells in latitude and longitude. For each grid cell, PCR-

20

2122232425

26

272829

30

31

3233

34

3536373839404142

GLOBWB simulates moisture storage in two vertically stacked upper soil layers, as well as the water exchange between the soil, the atmosphere and the underlying shallow groundwater reservoir. This last reservoir is a renewable or non-fossil groundwater resource and is represented as a linear reservoir (Kraijenhof van de Leur 1958). PCR-GLOBWB includes water demands from irrigation, livestock, industry and households. By comparing the demands to the available water supply, the initial water gap is computed. The water demands can be fulfilled from three sources: (1) renewable (often shallow) groundwater, (2) surface water, and (3) fossil groundwater. If sources 1 or 2 can not meet the total demand, the usage of fossil groundwater is assumed. PCR-GLOBWB is described in detail in Sutanudjaja et al. (2014) and Van Beek et al. (2011). PCR-GLOBWB was used to compute the OSWA.

PCR-GLOBWB is limited to the upper part of the soil and simulates shallow groundwater as a store with vertical fluxes to the surface layer and deep groundwater. It can not simulate the flow patterns of fossil, non-renewable groundwater, or drawdowns due to the extraction fossil groundwater. Therefore, PCR-GLOBWB was coupled to a global groundwater model (GGWM) built in MODFLOW (Harbaugh et al. 2000), which simulates lateral shallow and deep groundwater flows. De Graaf et al. (2014) have developed this global groundwater model, which is based on a single layer representation of the subsurface. The base data for the aquifer properties were the high resolution Global Lithological Map (GLiM) (Hartmann and Moosdorf 2012) and a global map of permeability (Gleeson et al. 2011). OGWA was based on computations with this global groundwater model. The steady-state global groundwater model was imported into iMOD (Vermeulen and Minnema 2015) modeling software to facilitate the implementation of the set of ten drawdowns. The imported steady state model was converted to a transient model using entries for recharge and surface water elevations from PCR-GLOBWB on a monthly timescale (1044 time steps between 1st of January 1980 and 31st of December 2065). After OSWA and OGWA were computed, OFWA was derived in a postprocessing step.

2.2 Operational Surface Water Availability from PCR-GLOBWB

The application of PCR-GLOBWB to compute Operational Surface Water Availability (OSWA) comprised three steps: (1) selection of climate and water demands inputs, (2) calibration of PCR-GLOBWB using discharge time series, and (3) calculation of OSWA.

2.2.1 Selection of climate forcing and sectoral water demand. We required a meteorological forcing covering the full period to determine the OFWA between 1980 and 2050. We used data from the Coupled Model Intercomparison Project (CMIP5) (Taylor et al. 2012). In CMIP5, 42 climate modelling groups participated to determine the climate change of four scenarios from the fifth assessment report (AR5) of the International Panel on Climate Change of the IPCC (2013). We chose the RCP6.0 climate change scenario as this is not one of the extreme scenarios (IPCC 2013), which leads to a 2.2 °C change in global average surface temperature. To date, no analysis has been carried out to determine which of the 42 climate models that participated in the CMIP5 give a central estimate of the climate change. In order to select one of the 42 climate models, we assessed a set of five climate models (GFDL-ESM2M, Hadgem2-ES, IPSL-CMA5-LR, MIROC-ESM-CHEM, NorESM1-M) that were delivered early to the CMIP5. These models were bias-corrected within the ISIMIP I project using the WATCH observational climate data of the time period

43444546474849505152

535455565758596061626364656667

68

69

707172

737475767778798081828384

between 1961 and 1999 (Hempel et al. 2013). Out of these five models, we selected the IPSL-CM5A-LR climate model (Dufresne et al. 2013), because it ranked third out of the five models considered with respect to the accumulated river runoff over the period between 2006 and 2099 for our study areas. The coefficient of variation of the accumulated river runoff in 2099 range (between GCMs) was 0.08 for the LRME basin and 0.17 for the area around LBB.

The socioeconomic scenario affects the industrial and domestic water demands. Because these scenarios do not provide spatially explicit projections of water demands, we used the methodology developed by Wada et al. (2014), which uses population and GDP changes derived from the socioeconomic scenarios. Here, we used the SSP2 scenario (Vuuren et al. 2014), which has been labeled as ‘middle of the road’, indicating that it is not one of the extreme scenarios. The combination of SSP2 and RCP6.0 can be seen as reasonable as it combines a ‘middle of the road’ SSP with a ‘stabilizing’ RCP (Vuuren et al. 2014). Agricultural and irrigation water demands depend on temperature and rainfall, not on the SSP scenario. They were computed at runtime by PCR-GLOBWB.

2.2.2 Rainfall correction using discharge time seriesGlobal hydrological models, including PCR-GLOBWB, are generally not calibrated with measured river runoff data from gauging stations. This is permissible for global assessments, but for operational water availability, accurate mean values of modelled river runoff are crucial. As the largest source of uncertainty to global streamflow modelling is the precipitation forcing (Biemans et al. 2009), we used a simple scheme to correct the rainfall, based on measured river runoff available from the Global Runoff Data Centre (GRDC) and the US Geological Survey (USGS) (main text, Table 1). The scheme essentially calculates a rainfall correction factor applied to rainfall data to ensure that simulated mean river flow matches the observed river flow. Here, we provide details on the implementation of the procedure.

We applied the following methodology on the ERA-Interim_GPCP rainfall fields as well as on the IPSL-CM5A-LR rainfall fields. First, we ran PCR-GLOBWB between 1980 and 2010 to determine the average river runoff at each gauging station. An initial estimate of the rainfall correction factor (P cor,

init) was obtained from the reference river runoff (Qref,mean) and the modeled river runoff (QPCR,mean):

Pcor, init = Qref,mean / QPCR,mean (1)

Pcor,init can be above, or below 1 depending on the catchment, or equal to one in case of a perfect match. Straightforward application of Pcor, init to the rainfall fields resulted in an overcompensation in terms of predicted discharge. To find the final precipitation correction factor (Pcor,final), we established the relationship between the correction factor and the model error (Fig. S2), where the error in the mean river runoff (EQmean) is:

EQmean,i = QPCR,mean,i – Qref,mean (2)

A set of five, linearly changing, correction factors (Pcor,i with i=1,..5) was created for each gauging station that ranged from Pcor,init to 1. These five correction factors were applied to the rainfall fields, and subsequently used to force PCR-GLOBWB. This resulted in a nonlinear relation between the correction factor and the model error. From the relationship between EQmean,i and Pcor,i, the final rainfall correction factor (Pcor,final) was determined as the point where EQmean = 0 (Fig. S2).

8586878889

90919293949596979899

100101102103104105106107108109110111112113

114

115116117118119

120

121122123124125

The correction factors are linked to the basins upstream of the gauging stations, which leaves out the areas downstream of the stations, and other ungauged basins. The area downstream of the stations was assigned a weighted average of basin area upstream of the stations. The correction factor for the ungauged basins was interpolated using inverse distance weighting. The area over open sea was assigned a value of one. This led to a correction factor maps for each study area. For example for the LRME asset, two gauging stations were used in the river basin. One for the Rhine River at Lobith, and one for the Meuse River at Lith. For each station a set of five correction factors was determined, and mapped over the whole study area.

.500 .600 .700 .800 .900 1.000-400

-300

-200

-100

0

100

200

300Station 1 Station 2

Precipitation correction factor (-)

mea

n er

ror (

m3/

s)

Fig. S2 Example of the relationship between the rainfall correction factor (Pcor) and predictive error in the mean runoff (EQmean). Pcor,init is the lowest correction factor for each station. The final correction factor (Pcor, final) is indicated by the dashed vertical lines.

2.2.3 Operational surface water availability calculationWe computed the OSWA components, river runoff and environmental flow requirements, with PCR-GLOBWB using the corrected IPSL-CM5A-LR forcing over the period between 1980 and 2065. To avoid double counting of these components, outflow cells were determined in the 5 by 5 cell window around the asset. Only the river runoff in these outflow cells was used, while the whole contributing catchment was taken into account for runoff generation. The river runoff was based on a run that included water demands, contrary to environmental flow, which was based on a pristine run. OSWA was computed as all river runoff minus environmental flow, and set to zero if the environmental flow was larger than the runoff.

2.3 Operational groundwater availability from the global groundwater modelThe application of the GGWM to compute OGWA comprised the downscaling of the model forcing (groundwater recharge and drainage levels), and the calculation of well yields, increased infiltration, and OGWA.

2.3.1 Downscaling of groundwater recharge and drainage levelsWe forced the GGWM (de Graaf et al. 2014) with output from PCR-GLOBWB, specifically net groundwater recharge and surface water levels. However, for studies with operational relevance, surface water level at five arcminute resolutions is too coarse a representation. To ensure as much

126127128129130131132133

134

135136137

138139140141142143144145146147

148149150151152

153154155156

spatial detail in the GGWM forcing as possible, we downscaled surface water levels from 5’ to 3 arcseconds (3’’). This is a factor 100 in cell length, and a factor 10,000 in cell area, which greatly improved channel positioning and water level variations. Downscaling of groundwater recharge was implemented as a refinement of the 5’ grid, using a nearest neighbor interpolation. We followed the method outlined in De Graaf et al. (2014) and Sutanudjaja et al. (2014) for the downscaling part of river runoff and discharge, but we implemented it at 3’’ resolution.

Firstly, the bankfull river discharge (Qbkfl), also known as the channel forming river discharge was computed. The bankfull river discharge is the river discharge that occurs once every 1.5 years, and was derived by downscaling the once every 1.5 years runoff from PCR-GLOBWB from 5’ to 3’’, and accumulating the downscaled runoff over the 3’’ flow direction map. Channel width (Wch) was calculated using Lacey’s morphometric relationship with bankfull river discharge (Lacey 1930):

W ch≈ Pbkfl=4.8√Qbkfl ,3 ' ' (3)

Combining equation 3 with Manning’s roughness equation gives the channel depth (Dch, bkfl), assuming a rectangular cross section:

Dc h ,bkfl=( n∗Qbkfl0.5

4.8∗Sl0.5 )3 /5

(4)

with n = Manning’s roughness coefficient (set to 0.02) and Sl = river slope. Monthly fields of surface water levels (HRIV) were subsequently computed by applying equation (5).

HRIV=DTM−Dc h ,bkfl2

+Dc h ,bkfl∗Qm

Qbkfl(5)

in which DTM is the HydroSHEDS Digital Terrain Model elevation (Lehner et al. 2008), Dch, bkfl = the depth of the channel at bankfull river runoff, Qm = river runoff in month m, and Qbkfl = the bankfull river runoff.

The first two terms in eq. 5 give the bottom of the river bed. The channel depth is divided by 2 as the terrain height from the HydroSHEDS DTM does not necessarily represent the bankfull water level as it is a snapshot of the heights at the moment of the data acquisition. The heights at the locations of the rivers, therefore more likely represent the median water level. The third term computes the ratio of the river runoff verses the bankfull river runoff and multiplies with the channel depth at bankfull river runoff. This gives a time series of spatially distributed surface water levels that vary around the DTM. The monthly river runoff is computed by routing the monthly runoff over the 3’’ flow direction map. Given the 3’’ resolution, this step is computationally intensive for large catchments.

2.3.2 Deriving operational groundwater availabilityOGWA depends on the maximum allowed depth of the drawdown. For this study, we assigned a range of ten increasingly deeper drawdowns to the site and ran the GGWM between 1980 and 2065 at a monthly time step. The maximum drawdown level is limited to one third of the saturated part of the aquifer, following Gefell et al. (1994). They stated that for a typical well design the drawdown at the well does not exceed the depth of one third of the initial aquifer depth due to the required

157158159160161162

163164165166167

168

169170

171

172173

174

175176177

178179180181182183184185186

187188189190191192

seepage face of the well to deliver the water at the pump. The drawdown level is an exponential function of the step that ranges from one to ten (Table S2). At step 1, the drawdown level is set to one third of the saturated thickness of the aquifer. In each subsequent step, the drawdown is divided by two. The exponential decrease in drawdown levels was chosen, as additional regulatory constraints on the groundwater level generally limit deep drawdowns, and this choice provides most information for small drawdowns.

The drawdowns were implemented with a ‘drain’ boundary condition, which ensured that the groundwater level was lowered over the full spatial extent of the asset. In practice a series of groundwater wells or a well field would need to be installed to lower the groundwater level over asset area. However, a well field would not have exactly the same resulting groundwater level as a drain, as the groundwater level would be slightly higher in between the wells. The difference between the groundwater level from a drain and from a well field depends on the spacing of the wells. Model output comprised: (1) simulated groundwater heads, (2) simulated drain fluxes, which represent the well yields, and (3) increased surface water infiltration (induced river bed recharge). Heads and fluxes were subtracted from the results of an undisturbed simulation to obtain the effects caused by the implemented drawdown at the asset.

Table S2 Overview of groundwater lowering and groundwater levels for each of the 10 drawdown steps

Asset LRME LBB

Terrain height (m +msl) 2.3 10.1Aquifer bottom height (m +msl ) 698.3 608.4Groundwater level (m +msl) 0.5 6.6Aquifer thickness (m) 700.5 618.5Aquifer saturated depth (m) 696.5 611.8

Groundwater level loweringa dz (m) dz (m)step 10 0.5 0.4

step 9 0.9 0.8step 8 1.8 1.6step 7 3.6 3.2step 6 7.3 6.4step 5 14.5 12.7step 4 29.0 25.5step 3 58.0 51.0step 2 116.0 102.0step 1 232.2 203.9

a Step 8 was used in main text Figs. 3 and 4

3 PCR-GLOBWB validation and well yield comparison between a global and a national groundwater modelPCR-GLOBWB has been validated against runoff measured at gauging stations by Van Beek et al. (2011). They showed that PCR-GLOBWB simulations of monthly river discharge compared well with observations from most of the large rivers. Exceptions are basins subject to large extractions for irrigation purposes, where simulated discharge exceeds the observations even when water demand is taken into account. To validate PCR-GLOBWB for our study areas, we chose the ERA-Interim

193194195196197198

199200201202203204205206207208

209

210

211

212213214215216217218

reanalysis data (Dee et al. 2011) as the climate forcing. ERA-Interim reanalysis data were rescaled using the Global Precipitation Climatology Project (GPCP) (Adler et al. 2003; Huffman et al. 2009). We will therefore refer to this forcing as ERA-Interim_GPCP. The reason for the rescaling is that the vast majority of the observations included in this ERA-Interim reanalysis are satellite-based. The GPCP correction is based on daily precipitation measurements from 6700 quality controlled meteorological stations that are interpolated to a global monthly-averaged dataset (Balsamo et al. 2010). We applied the same rainfall correction method to the ERA-Interim_GPCP rainfall fields as to the IPSL-CM5A-LR to derive the correct average discharge.

Time series of predicted discharge were compared against the gauge data, and PCR-GLOBWB performance was assessed by the Nash-Sutcliffe model efficiency (Nash and Sutcliffe 1970). Fig. S3 shows the time series of the measured river runoff and the corrected PCR-GLOBWB output as well as the regime curves of the gauge data plus PCR-GLOBWB output. The regime curves show a clear delay in the peak of the river runoff of up to two months. For the West fork of the San Jacinto River, a few large peaks were not captured by the ERA-Interim_GPCP rainfall fields, due to the small size of the catchments.

Fig. S3 Validation of PCR-GLOBWB. The right side shows the gauge data and the PCR-GLOBWB modelled river runoff based on the corrected rainfall fields. The left side shows the regime graphs of

219220221222223224225226

227228229230231232233

234

235236

the gauge data, and three PCR-GLOBWB runs: no correction (with water demands, but without correction), Pristine (corrected, but without water demands) and PCR-GLOBWB (corrected, and with water demand.

3.2 Comparison of GGWM well yields with a detailed national scale groundwater model

The GGWM was validated against local groundwater heads, and showed good model performance (de Graaf et al. 2014). No comparison was made between actual and modeled well yields because of lacking well yield data. To obtain a rough estimate of the error made by our global modeling approach, we compared the well yields of the GGWM with well yields from a detailed local groundwater model, the Netherlands Hydrological Instrument (NHI) (De Lange et al. 2014). NHI is a seven-layer groundwater model, at 250x250 meter resolution, using relevant groundwater and surface water information from the Netherlands. The subsoil parameters of the NHI model were re-projected to 3’’ to substitute the subsoil parameters of the GGWM model. We will refer to this model as GGWM-NHI. The following MODFLOW files were converted: (1) boundary conditions (-) for all model layers, describing an accurate layout of the North sea shore and outlet of the river Rhine, (2) starting heads (m above ordnance datum) for all model layers describing the lateral in- and outflow across the model boundary of the LRME asset towards the outside area, (3) transmissivities (m2/day) for all model layers describing the lateral conductance, and (4) vertical resistance (day) in between all model layers describing the vertical resistance caused by high resistance material (clay) in between the sandy aquifers represented by the model layers. For GGWM-NHI to imitate the full penetration of a well in the one-layer GGWM model and make the yield comparable, we put drains at the required drawdown level within a given aquifer as well as drains at the bottom of any aquifer above it., seven levels of drains were implemented on top of each aquitard in order to connect the aquifers in between the aquitards (Fig. S4; Table S3). We compared the mean well yields of the GGWM and GGWM-NHI for each of the ten drawdown levels to assess its performance (Fig. S5).

237238239

240

241242243244245246247248249250251252253254255256257258259260

Fig. S4 Cross section through the subsurface at LRME with the depth of the seven additional drainage levels in the NHI model. The dotted lines show the drains at different levels and these lines are extended over the full area location of LRME. The yellow colored layers represent aquifers, the other layers represent aquitards (clay layers). These layers are related to the geological formations as given in (REGIS 2015; Vernes and Van Doorn 2005).

Table S3 Subsurface description of LRME for the NHI (local groundwater model at LRME)

Height (m +OD)

Thickness Aquifer (m)

Thickness Aquitard (m)

Transmisivity(m2/day)

VerticalResistance(day)

Horizontal Permeability(m/day)

Vertical Permeability(m/day)

3.2 5.0 5.6 1.1-1.8 18.1 511.0-19.9 5.9 276.0 46.8 0.036-25.8 0.0 1.0-25.8 4.7 39.0 8.3 0-30.5 15.6 372.0-46.1 5.0 96.0 19.2 0.042-51.1 11.2 153.0-62.3 15.8 291.0 18.4 0.073-78.1 11.8 181.0-89.9 61.1 693.0 11.3 0.065-151.0 0.0 1.0-151.0 63.2 712.0 11.3 0-214.2Total: 2,112.6

261

262263264265266

267

268

269

0.5 1.8 7.3 29116

0

0.5

1

1.5

2

2.5

Drawdown (m)

Wel

l yie

ld fa

ctor

(-)

GGW

M- N

HI /

GGW

M

Fig. S5 Well yield factor (=GGWM-NHI / GGWM; [m3/s]) per drawdown level. If the well yield factor is below one, the GGWM overestimates, above one the GGWM underestimates well yields compared to the GGWM-NHI.

References

Adler RF et al. (2003) The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–Present) Journal of Hydrometeorology 4:1147-1167 doi:10.1175/1525-7541(2003)004<1147:tvgpcp>2.0.co;2

Balsamo G, Boussetta S, Lopez P, Ferrantie L (2010) Evaluation of ERA-Interim and ERA-Interim-GPCP-rescaled precipitation over the U.S.A. ECWMF, Reading

Biemans H, Hutjes RWA, Kabat P, Strengers BJ, Gerten D, Rost S (2009) Effects of Precipitation Uncertainty on Discharge Calculations for Main River Basins Journal of Hydrometeorology 10:1011-1025 doi:10.1175/2008JHM1067.1

de Graaf IEM, Sutanudjaja EH, van Beek LPH, Bierkens MFP (2014) A high resolution global scale groundwater model Hydrol Earth Syst Sci Discuss 11:5217-5250 doi:10.5194/hessd-11-5217-2014

De Lange WJ et al. (2014) An operational, multi-scale, multi-model system for consensus-based, integrated water management and policy analysis: The Netherlands Hydrological Instrument Environmental Modelling & Software 59:98-108 doi:http://dx.doi.org/10.1016/j.envsoft.2014.05.009

Dee DP et al. (2011) The ERA-Interim reanalysis: configuration and performance of the data assimilation system Quarterly Journal of the Royal Meteorological Society 137:553-597 doi:10.1002/qj.828

Dufresne JL et al. (2013) Climate change projections using the IPSL-CM5 Earth System Model: from CMIP3 to CMIP5 Clim Dyn 40:2123-2165 doi:10.1007/s00382-012-1636-1

Gefell MJ, Thomas GM, Rossello SJ (1994) Maximum Water-Table Drawdown at a Fully Penetrating Pumping Well Ground Water 32:411-419 doi:10.1111/j.1745-6584.1994.tb00658.x

Gleeson T et al. (2011) Mapping permeability over the surface of the Earth Geophysical Research Letters 38:L02401 doi:10.1029/2010gl045565

Harbaugh AW, Banta ER, Hill MC, McDonald MC (2000) MODFLOW-2000, the U.S. Geological Survey modular ground-water model: User guide to mukularization concepts and the ground-water flow process. United States Geological Survey, Reston, Verginia

270

271272273

274

275

276277278279280281282283284285286287288289290291292293294295296297298299300301302

Hartmann J, Moosdorf N (2012) The new global lithological map database GLiM: A representation of rock properties at the Earth surface Geochemistry, Geophysics, Geosystems 13:Q12004 doi:10.1029/2012gc004370

Hempel S, Frieler K, Warszawski L, Schewe J, Piontek F (2013) A trend-preserving bias correction - the ISI-MIP approach Earth Syst Dynam 4:219-236 doi:10.5194/esd-4-219-2013

Huffman GJ, Adler RF, Bolvin DT, Gu G (2009) Improving the global precipitation record: GPCP Version 2.1 Geophysical Research Letters 36:L17808 doi:10.1029/2009gl040000

IPCC (2013) Climate change 2013: The physical science basis. contribution of working group I to the fifth assessment report of the Intergovernmental Panel on Climate Change. In: Stocker TF et al. (eds)

Kraijenhof van de Leur DA (1958) A study of non-steady groundwater flow with special reference to a reservoir coefficient De Ingénieur 70:87-94

Lacey G (1930) Stable channels in alluvium Proceedings of the Institution of Civil Engineers, London 229:259-292

Lehner B, Verdin K, Jarvis A (2008) New Global Hydrography Derived From Spaceborne Elevation Data Eos Trans AGU 89:93-94 doi:10.1029/2008eo100001

Nash JE, Sutcliffe JV (1970) River flow forecasting through conceptual models part I — A discussion of principles Journal of Hydrology 10:282-290 doi:http://dx.doi.org/10.1016/0022-1694(70)90255-6

REGIS (2015) Regis2 II v2.1 (2009) Dinoloket TNO. https://www.dinoloket.nl/ondergrondmodellen. Accessed 24-07-2015

Sutanudjaja EH, van Beek LPH, de Jong SM, van Geer FC, Bierkens MFP (2014) Calibrating a large-extent high-resolution coupled groundwater-land surface model using soil moisture and discharge data Water Resources Research 50:687-705 doi:10.1002/2013wr013807

Taylor KE, Stouffer RJ, Meehl GA (2012) An Overview of CMIP5 and the Experiment Design Bulletin of the American Meteorological Society 93:485-498 doi:10.1175/bams-d-11-00094.1

Van Beek LPH, Wada Y, Bierkens MFP (2011) Global monthly water stress: 1. Water balance and water availability Water Resources Research 47 doi:10.1029/2010wr009791

Vermeulen PTM, Minnema B (2015) iMOD user manual. Version: 3.01. Deltares, Utrecht, The Netherlands

Vernes RW, Van Doorn THM (2005) Van gidslaag naar hydrologische eenheid - toelichting op de totstandkoming van de dataset - REGIS II. TNO, Utrecht

Vuuren DP et al. (2014) A new scenario framework for Climate Change Research: scenario matrix architecture Climatic Change 122:373-386 doi:10.1007/s10584-013-0906-1

303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336

337

338