Climate change: hydrological impact studies
Hans Middelkoop Department of Physical Geography, University of Utrecht, the Netherlands
1. Introduction
Because rivers are part of the hydrological cycle, climate has been an important control of river regimes and fluvial patterns in the past, and anticipated climate changes are expected to have major impacts on modern rivers in the future. Climate influences rivers directly by controlling the hydrological inputs of a river basin, but it also controls rivers in a more indirect way by influencing vegetation cover, erosion rates and soil characteristics of river basins. In this section we will focus on the potential hydrological consequences of future climate change on river regimes. For this purpose the chapter deals with the basics of runoff generation and examples of hydrological models first. Then the link with climate model output is discussed. After some general remarks on the impacts of climate change on river regimes, two case studies are presented, one from the Rhine basin and one from sub-arctic river basins. The Rhine basin is an example of a temperate-climate river with a combined rainfall-snowmelt regime with a densely populated river basin. Subarctic rivers are important since climate models predict relatively large climate changes for these regions, while these regions are considered very susceptible to climate change because of the role of snow and permafrost. Furthermore, the resulting changes in runoff may have impacts not only at regional scale, but major increases in meltwater runoff to the Arctic Ocean even may even affect the global climate (e.g. Arnell, 2005).
2. Runoff generation
In most climate impact studies the hydrological regime of a river is regarded as the result of time-varying changes in the general water balance of the upstream basin: runoff Q can be expressed as Q = P – E S, where P = amount of precipitation, E = amount of evapotranspiration and S = change in storage. These storages include water storage in vegetation, surface detention, storage in snow and glaciers, soil and groundwater storage and storage in lakes and channels. The importance of each of these components differs in time and in space, depending on season and local climate conditions, topography, soil characteristics, vegetation and land management, and land and river management.
Figure 2.1 shows schematically the main inputs, losses, transfers and storages of water from which runoff is generated. Most of these are controlled or affected by climate. Input is directly dependent on amounts of precipitation, including seasonal variations, intensity of precipitation events. Depending on the temperature, precipitation occurs in the form of snow, and remains stored on the ground surface until it melts, while there is some loss through sublimation. Evaporation losses depend on a combination of temperature, wind speed, radiation, and vapor pressure of the air. Canopy interception and transpiration losses depend on the vegetation cover. When considering climate change over longer periods, changes in vegetation may have a considerable impact on these transfers. The amount of water losses for carbon dioxide assimilation through stomata may vary considerably, depending on the vegetation or crop type. Permeability and water holding capacity of the subsoil are determined by the substrate characteristics such as lithology, degree of weathering, organic matter content and type, surface litter end crusting, and rooting by plants. The transfers from one storage component into another may also be influence by climate. The amount of water infiltrating into the soil not only depends on soil characteristics, but also on the flux of water available for infiltration, depending on precipitation intensities and snowmelt rate. In arctic regions seasonal or permafrost may hinder infiltration.
1 -2 -1 Pr Total Precipitation (kg m s ) -2 -1 Pn Snowfall (liquid equiv.) (kg m s ) -2 -1 Ps Precipitation reaching soil (kg m s -2 -1 Pv Precipitation intercepted by vegetation (kg m s ) -2 -1 Pg Infiltration (kg m s ) -2 -1 Qr Surface runoff (kg m s ) -2 -1 D1 Surface/deep soil soil water diffusion (kg m s ) -2 -1 K2 Gravitational drainage (kg m s ) -2 -1 En Sublimation (kg m s ) -2 -1 Er Bare-soil evaporation (kg m s ) -2 -1 Eg Evaporation from interception (kg m s ) -2 -1 Etr Transpiration (kg m s ) -2 Wr Canopy water store (kg m ) -2 Wn Snow pack SWE (Snow Water Equiv.) (kg m ) 3 -3 wg Surface soil water reservoir (m m ) 3 -3 w2 Bulk soil water reservoir (m m )
d1 Surface reservoir soil depth (m)
d2 Total soil depth (m)
Figure 2.1 Flows and storages of water in the water balance
Precipitation may arrive in the stream channel by one of the following flow paths: (1) direct precipitation onto the water surface, (2) overland flow, (3) shallow subsurface flow, and (4) groundwater flow. Snowfall and snowmelt water will follow one of these paths as well. The first two and part of the shallow subsurface flow may respond quickly to precipitation events, and represent the major runoff contribution during precipitation storm periods. This component is often referred to as quickflow. Baseflow is the sustained runoff that is fed by subsurface flow and groundwater flow, and forms the delayed part of river runoff. It is less variable in time than quickflow, and forms the major part of river runoff during dry periods. A hydrograph thus comprises both components, which usually are separated on the basis of the shape of the recession limb of the hydrograph (figure 2.2). Thus, as it affects the water balance components, climate change may lead to a shift in the relative proportions of quickflow and baseflow in rivers.
Figure 2.2 Quick flow and delayed runoff (baseflow) in a hydrograph (source: Ward & Robinson, 1990)
2 3. Hydrological models
3.1 Model types
To analyze the impacts of changes in climate (and land cover) on the runoff regime of rivers, hydrological models have been developed. These models simulate the transfers and storages of water in river basins, depending on meteorological inputs and river basin characteristics, such as topography, vegetation cover, geology and soil characteristics, and drainage density.
Models can be categorized in different types according to the model structure and degree of causality of the processes implemented in the model (black-box or empirical models, models based on physical laws, conceptual models based on simplified representations of these fundamental laws), spatial discretization (lumped or distributed parameters), temporal discretization (hourly, daily, monthly time step), and spatial coverage (point, catchment, regio, world). Leavesley (1994) gives a review of model types, of which the main categories are summarized here. Empirical models consider only statistical relations among the water balance components, for example between precipitation, temperature and runoff. They do not explicitly consider the physical laws that determine the processes behind these relations, but instead use a sort of transfer function that is empirically estimated. These are generally referred to as black box models. Empirical models do not consider spatially distributed parameters or variables, but are lumped models that consider larger areas (basins, sub-catchments) as a whole. Water balance models essentially keep track of the components in the water balance equation, thereby describing the storage components in varying degree of complexity. These models calculate the water balance components for consecutive time steps, varying from daily to annual, but mostly monthly. Key parameters in such models are the fraction of precipitation that contributes directly to runoff, the maximum storage capacity in the basin, and the time lag for converting water available for discharge (in particular from the groundwater reservoir) to streamflow (Arnell, 1992). These models must be calibrated using time series of input and runoff data observed in the past. Conceptual models use approximations or simplified schemes of the physical processes that control runoff generation, and may include empirical relations for parts of the model. These models attempt to consider the essential interactions and non-linearities in the water balance components, and account for the time it takes for precipitation to become streamflow. These models may consider vertical processes, such as interception, soil moisture storage, evapotranspiration, groundwater recharge and snow cover built-up and snowmelt. Lateral processes that may be included are surface runoff, subsurface flow and streamflow routing. Model parameters and inputs are usually spatially explicit, but sometimes are lumped over smaller subcatchments that are assumed to be homogeneous. When compared to water balance models, the times step used is generally smaller (i.e. hourly to 10-daily). These models, too, must be calibrated using time series of input and runoff data observed in the past. Physically based models consider the full sets of fundamental laws that describe the hydrological processes, transfers and storages involved in the runoff generation. Consequently, they are entirely distributed to facilitate the required detailed process description. A major disadvantage of these models is that they comprise a large number of distributed parameters, while the availability and quality of basin and climate data at the spatial and temporal resolution is often inadequate to estimate the model parameters and validate the results at this level of detail. Furthermore, calibration sessions often result in equifinality: different combinations of parameters sets yield equally good model performance, making it impossible to determine these parameters (Beven, 2001). This type of models is only applied for smaller catchments.
3 When applying a hydrological model to assess the hydrological implications of climate (and land use) changes, it is essential that the model is able to predict the relations between climate inputs, basin characteristics and runoff correctly under changed conditions. In this respect, empirical, black models should be treated with extreme caution, because these have been calibrated using empirical data of the past, but considering the interactions and non-linearities in the factors that control runoff, it is unlikely that the same empirical relations remain valid under changed environmental conditions. Ideally, one would apply a purely physically-based model, that describes all the water balance components and their spatial variability in a deterministic way. Unfortunately, it is often impossible to determine all the required parameters for such models for future conditions. Therefore, most studies have used a conceptual modeling approach, with distributed parameters and input variables.
3.2 Examples
Examples of hydrological models that have been applied to various river basins are the water balance model concept used in RHINEFLOW (Kwadijk, 1993) and variants of this model, the HBV model (Bergström, 1995) and the TOPMODEL (Beven et al., 1995).
Water balance models Kwadijk (1993) and Van Deursen & Kwadijk (1993) developed a raster-based water balance model concept for hydrological modeling of large river basins. The first model was made for the Rhine basin (named RHINEFLOW), and was based on a 3 km x 3 km raster grid, and a time step of one month. Later versions of this model and models applied to other river basins used a finer grid (1 km x 1 km) and time steps of 10 days to 1 day. Examples of these models are the subarctic models USAFLOW (Van der Linden, 2002), applied to a tributary of the Pechora River in European-Siberia, and TANAFLOW and BARENTSFLOW applied in North Scandinavia and the Barentz region (Dankers, 2002). These water balance models simulate the water balance of a river basin as a series of storages of reservoirs (figure 2.3). For each grid cell, the model calculates on monthly basis the storages and transfers from precipitation to runoff, using the major water storage compartments snow, soil, groundwater and lakes. For each time step in a model run, these reservoirs are updated, representing the temporal behaviour of these storages. Basin stream flow is obtained by adding the net water production for all cells located in a catchment. Input data of the model are: temperature and precipitation (and for later versions for E-pot: reference evapotranspiration or the set of input parameters for the Penman equation) time series from measurement stations, geographical data (soil characteristics, vegetation, terrain elevation). Assuming that all water available for runoff leaves the catchment within one time step, the model produces month-to-month or 10- daily runoff for the main river and for its main tributaries. The models were developed in the PCRaster modeling language (www.pcraster.nl) using generic functions for e.g. determining the drainage network or accumulating water originating from upstream cells. The models were calibrated for three parameters: (1) separation between runoff and groundwater discharge, (2) flow recession, and (3) snowmelt rates per degree temperature rise. Calibration occurred against time series of observed river discharge at different stations within the basin, and snow cover data. Evapotranspiration was determined in the earliest models using the Thornthwaite-Mather approach, while later versions used an empirical relation between temperature and reference evapotranspiration (Brandsma, 1995), and a full Penman-Monteith model.
Although the spatial resolution of this model type was mostly 1x1 km2, and thus individual cells can be evaluated at this level, the calibration and validation do not assure a correct estimate for individual cells, and thus local assessments cannot be made based on the regional model. Nevertheless, at the level of subcatchments of tributaries with a size of several hundreds of square km the models produce reliable results.
4 RCM Meteo. stats (HadCM4)
P T Rad. RH Wind
TANA Penman- AE Rain Snow PE SNOW Monteith
Soil Water
Q
Ground Water
Figure 2.3 Model scheme of the water balance model TANAFLOW (source: Dankers, 2002)
HBV A second example of a widely used modelling framework is the HBV model, developed by the Swedish Meteorological and Hydrological Institute (SMHI) (Bergström & Forsman, 1973; Bergström, 1995; www.smhi.se/sgn0106/if/hydrologi/hbv.htm). The HBV model is a conceptual hydrological model of which the scheme is shown in figure 2.4. The model consists of subroutines for meteorological interpolation, snow accumulation and melt, evapotranspiration estimation, a soil moisture accounting procedure, routines for runoff generation and finally, a simple routing procedure between subbasins and in lakes. HBV can be used as a semi-distributed model by dividing the catchment into subbasins. Each subbasin is then divided into zones according to altitude, lake area and vegetation. Input data are observations of precipitation, air temperature and estimates of potential evapotranspiration. The time step is usually one day, but it is possible to use shorter time steps. The evaporation values used are normally monthly averages although it is possible to use daily values. Air temperature data are used for calculations of snow accumulation and melt. It can also be used to adjust potential evaporation when the temperature deviates from normal values, or to calculate potential evaporation.
Figure 2.3 Model scheme of the HBV model (source: Bergatröm,1995)
The model is used for flood forecasting in the Nordic countries, and many other purposes, such as spillway design floods simulation, water resources evaluation, nutrient load estimates
5 and climate impact studies (Bergström et al., 2001; Gardelin et al., 2001). In different model versions HBV has been applied in more than 40 countries all over the world, comprising different climatic conditions. The model has been applied for scales ranging from lysimeter plots to the entire Baltic Sea drainage basin (Bergström & Carlsson, 1994; Graham, 1999).
TOPMODEL TOPMODEL (http://www.es.lancs.ac.uk/hfdg/topmodel.html; Beven & Kirby 1979; Beven et al., 1995) is a physically based watershed model that simulates the variable-source-area concept of streamflow generation. TOPMODEL predicts the relative amount and spatial distribution of subsurface, infiltration excess, and saturation excess overland flow based on surface topography and soil properties. TOPMODEL makes use of a topographic index of hydrological similarity based on an analysis of the topographic data, which can be described as follows: ln(α / tan β), where α is the area draining through a point from upslope and tan β is the local slope angle. The index identifies areas with greater upslope contributing area, α, and lower gradients, β, as being more likely to be saturated than areas with lower α and higher β. The model assumes a spatially uniform recharge rate and quasi-steady subsurface response to derive a function relating local soil moisture storage or water table depth to the topographic index (ln(a/tanß)) of a catchment:
Si = S + m {λ – ln(α / tan β)I – (δ – ln (Ki), where Si is the local soil moisture deficit, S is the mean soil moisture deficit of the catchment, m is a parameter that characterizes the decrease in hydraulic conductivity with soil depth, a is the drainage area per unit contour length, ß is the slope, Ki is the lateral transmissivity of the soil profile when the water table just intersects the surface, and λ and δ are the mean values of ln(a/tanß) and ln(K) for the catchment. Many applications ignore the soil transmissivity terms because the spatial pattern of soil transmissivity is seldomly known and is therefore assumed to be constant over the catchment. Si represents a negative soil moisture deficit so that Si = 0 at complete saturation and Si > 0 when a soil moisture deficit occurs. The mean soil moisture deficit of a catchment at time t, St, is calculated by: St = St-1 – (qt-1 – ri)Δt, where q is the total catchment runoff at time t - 1 divided by the catchment area, r is the net recharge rate into the soil column, and Δt is the time interval used for the model simulation. The soil moisture deficit at every point (grid cell) in the catchment is then computed using the first equiation to calculate Si and water is routed to the catchment outlet via: (1) subsurface runoff in areas with a soil moisture deficit larger than the precipitation added during a time step; (2) subsurface and infiltration excess overland flow in areas with rainfall intensities greater than the infiltration capacity, and (3) subsurface and saturation excess overland flow in areas with either a soil moisture deficit smaller than the incremental precipitation in a unit time step or that were saturated during the previous time step. The subsurface flow rate qb of the catchment is calculated by
qb = exp( -λ – δ)) exp(-St / m), and the saturation excess runoff q0, which is the sum of the excess soil moisture and direct precipitation that falls on the saturated areas, is calculated by
q0 = (1/At) ∫As {-Si / t + r} dA, where As is the area of the catchment with surface saturation (Si ≤0) and At is the total area of the catchment. This approach means that predicted soil moisture patterns will follow the outline of the topographic index and the predicted saturated source area will expand and contract as the water balance of the model changes. Total runoff q at each time step is the sum of subsurface and surface runoff.
4. Linking output of climate models to hydrological models
To assess the climate-induced changes in hydrological regime, a reference period or ‘baseline’ scenario must be defined. In many climate impact studies 30 to 40-year time periods between the 1950s and 1990s are used as reference. The hydrological characteristics
6 for this reference period are then compared to those resulting from a changed climate projected to a 20 to 30-year time slice in the future. Projection times often used are 2050 (plus or minus a decade) and 2080 – 2100. For these periods, different climate scenarios may be evaluated, for example assuming different emission and global warming scenarios or resulting from different global climate models (General Circulation Models – GCMs). Although using 20 to 30-year time slices as a basis may be adequate to assess the average changes and changes in year-to-year variability, the considered periods are too short to assess the implications for extreme events or for the design of engineering works (which usually are based on events with 100-year or longer recurrence times). Relatively few studies have considered transient runs, in which the hydrological response under gradually changing climate conditions is evaluated.
Ideally, impact studies using hydrological models use information on climate change at the time scale and spatial resolution of the hydrological model. Although GCMs work at a temporal resolution of less than one hour, the raw outputs of climate change experiments from are an inadequate basis for assessing the effects of climate change on land-surface processes at regional scales. This is because the spatial resolution of GCMs is too coarse (in the order of 2 latitude x 3 longitude) to adequately represent the spatial variation in surface topography, snow cover and vegetation, or to resolve important sub-grid scale processes. Consequently, GCM output is often unreliable at individual and sub-grid box scales. While GCMs may produce adequate simulations of large-scale (i.e. continental-scale) atmospheric circulation patterns and temperatures, regional-scale simulations of precipitation and estimates of precipitation extremes derived from GCM output are unreliable. This requires an additional step to link GCM output to hydrological models, which is referred to as downscaling.
There are several ways to downscale GCM output to the spatial resolution required in hydrological studies of river basins. These methods essentially address two key issues: downscaling the GCM output to the hydrological modeling scale, and using GCM output (T, P, etc.) directly as inputs for hydrological models (which assumes that the GCM simulates the future climate well), or using GCM output to define changes (anomalies) in climate that may be applied to historical data.
The simplest method of downscaling is to perturb time series of observed climate data from meteorological stations within or near the investigated river basin in the past using GCM output. For this purpose, a GCM run is carried out for the present (baseline) climate and for a time slice with increased greenhouse forcing in the future, and for each GCM grid cell the difference between the these runs is determined. These differences are usually expressed as average 10-daily or monthly changes in climate parameters (such as P, T, and other parameters needed by the hydrological model). By taking the difference between two GCM runs it is assumed that systematic bias in the GCM output is cancelled out. The monthly climate anomalies are subsequently interpolated between the GCM grid cells to the locations of the meteorological stations from which the observed climate records were obtained. These observed records are then adapted according to the anomalies determined using the GCM. Temperature changes are usually applied as an absolute change:
Tsc(t) = Tobs(t) + (TGCM-sc – TGCM-ref), with Tsc(t) = scenario time series to be used as input in the hydrological model, Tobs(t) = a baseline observed climate series, TGCM-sc = average climate values for changed climate calculated with the GCM, TGCM-ref = average climate values for the baseline climate calculated with the GCM. Precipitation changes are usually implemented as relative changes by applying the ratio (PGCM-sc / PGCM-ref) to the observed time series of precipitation: Psc(t) = Pobs(t) x (PGCM-sc / PGCM-ref). The disadvantage of this method is that it is impossible to change the length of wet or dry periods, or to distribute the changes in e.g. precipitation over either previously dry days or over the wettest days (increasing high- intensity precipitation). In a modeling study for the Rhine basin Shabalova et al. (2003) therefore also estimated the changes in variance of T and P using the GCM, and applying these as well to the observed time series:
7 Tsc(t) = (Tobs(t) - Tobs-avg) x GCM-sc / GCM-ref + Tobs-avg + (TGCM-sc – TGCM-ref), where Tobs-avg = n-year averages of the observed 10-daily of monthly temperatures. This transformation changes the mean of Tobs(t) as in the previous model, but also changes the standard deviation of Tobs(t) by the ratio GCM-sc / GCM-ref. Negative values that would arise when the same procedure is applied to precipitation were reset to zero values. In this way the spatio-temporal resolution of the climate series as contained in the observed records from the meteorological stations within the study area remains preserved. Also, spatial correlations between climate variables remain preserved.
Another method of bridging the spatial difference between GCM output and hydrological models is statistical downscaling using empirical relationships between grid-box scale climate (such as atmospheric circulation indices) and sub-grid scale surface predictands (such as precipitation). This approach assumes these large-scale atmospheric features are well simulated by a GCM, and that there are stable relations between these features and local detail, not only under present-day climate but also under future climate forcing. Various examples of statistical downscaling results for daily precipitation and potential evaporation as input for hydrological models were evaluated by Wilby (e.g., Wilby et al., 1998; Diaz-Nieto & Wilby, 2005). Used predictors for precipitation in UK catchments were e.g.: mean sea-level pressure, surface zonal velocity, surface vorticity, 850-hPa geopotential height. Many downscaling techniques apply to single point estimates of climate parameters. However, in larger river basins, also the spatial correlation between climate variables should be considered. This requires a major extension of the downscaling method. For example, Bardossy & Plate (1992) and Stehlik & Bardossy (2002) incorporated spatial correlation in their estimates of precipitation in the Rhine basin and Greece using atmospheric circulation patterns derived from GCM output. Table 4.1 summarizes relative strengths and weaknesses of using climate anomalies or statistical downscaling for climate scenario generation.
Table 4.1 Relative strengths and weaknesses of using climate anomalies or statistical downscaling for climate scenario generation (source: Diaz-Nieto & Wilby, 2005) Scenario technique Strengths Weaknesses Climate anomalies - Station-scale scenarios - Depends on realism of the climate - Computationally straightforward model providing the change factors and quick to apply - Temporal structure is unchanged - Local climate change scenario is for future climate scenarios directly related to changes in the - Step changes in scaling at the regional climate model output monthly interface - Restricted to time-slice scenarios Statistical - Station-scale scenarios - Depends on realism of the climate downscaling - Ensembles of climate scenarios model providing the forcing permit uncertainty analyses - Requires high quality observations - Delivers transient climate change and climate model output scenarios at daily time-scale - Predictor-predictand relationships - Allows exploration of temporal are not always stationary sequencing of meteorological events - Choice of predictor variables and transfer function affects results
The final method of downscaling is using a high-resolution regional climate model (RCM) nested within the grid of the GCM. The spatial resolution of the RGM is much higher (in the order of 0.5 x 0.5 or less), which is fine enough to consider sub-grid detail such as topography, land cover patterns or coast lines, and to give more realistic representations of subgrid-scale weather systems. However, models of higher resolution cannot practically be used for global simulation of long periods of time. To overcome this, RCMs are constructed for limited areas and run for shorter periods (20 years or so). RCMs take their input at their boundaries and for sea-surface conditions from the global GCMs. Thus, there is only a one- way linking from GCM to RCM, as the RCM does not pass information back to the GCM.
8 RCMs have greatly improved climate information for use in hydrological modelling, but regional climate models do not yet provide all the solutions for generating climate change scenarios. There will be errors in their representation of the climate system and their resolution will not be sufficient for some applications. Particularly precipitation estimates still may not replicate observed precipitation amounts from meteostations. In addition, there are two main limitations to their use in conjunction with GCMs. Predictions from an RCM are dependent on the realism of the global model driving it; any errors in the GCM predictions will be carried through to the RCM predictions. Because different GCMs represent the climate system in different ways, predictions that they make at a regional scale can be very different. Furthermore, interfacing RCMs with GCMs is a complex technical issue and running a RCM requires computing resources (although PC-versions are becoming available). Finally, the data requirements of RCMs are very substantial. In recent years RCMs are increasingly applied in regional climate impact studies. Most climate centres currently have regional climate models operational. Examples of RGMs are the PRECIS project of the UK Hadley Centre (http: //www.meto.gov.uk/research/hadleycentre/models/PRECIS.html), the REMO model of the Max-Planck-Institut für Meteorologie, Hamburg (http://www.mpimet.mpg.de/en/depts/dep1/reg/), the The Canadian Regional Climate Model (CRCM) of the Canadian Centre for Climate Modelling and Analysis, Victoria (http://www.cccma.bc.ec.gc.ca/models/crcm.shtml).
5. Hydrological impacts
5.1 Changes in water balance components
Precipitation is the most important driving force of fluvial systems. Changes in precipitation will therefore be the primary causes of runoff changes. Although main attention may be focused to mean climate, variability of the climate inputs to the hydrological system may change as well. Because of non-linearities in the hydrological response, both the mean and variability of runoff may be affected by either a change in mean climate or variability of climate (figure 5.1). This is important, since many fluvial processes and river socio-economic functions of modern rivers are determined by the magnitude and occurrence of extremes.
Figure 5.1 Schematic illustration of the effect of changing the mean and variance of climate inputs in the distribution of hydrological output (source: Arnell, 1996)
9 Temperature is the second important climate control, as it affects evaporation and snow melt. Various sensitivity studies have evaluated the effects of changes in annual temperature and actual precipitation on the annual runoff. By plotting sensitivity surfaces of the changes in annual runoff to changes in precipitation and temperature, an indication is obtained of the sensitivity of a river to climate change (figure 5.2). The slope and the distance between the lines are measures for this sensitivity. The steeper the lines are, the greater the sensitivity to precipitation relative to temperature. The closer the spacing between the lines, the more sensitive the system is to climate change. Table 5.1 summarizes for different rivers the sensitivity of annual runoff for different changes in precipitation and a temperature rise of 2 °C.
Figure 5.2 Sensitivity of annual runoff to changes in precipitation and temperature (from: Arnell, 1996)
Table 5.1 Percentage change in annual runoff for different changes in precipitation and a temperature rise of 2 °C (source: Arnell, 1996)
5.2 Seasonal changes in runoff
The change in annual runoff depends on how temperature and precipitation change over the year. Runoff changes may therefore vary considerably over the seasons, depending on the relative changes in climate inputs as well as the present intra-annual variations of these parameters. Particularly snowmelt may considerable influence the seasonal distribution of
10 runoff. This is illustrated in figure 5.3 that shows the monthly discharges of high-elevation catchments where snow storage and snowmelt have a significant contribution in the annual runoff cycle. Under warmer climate conditions, the amounts of snow storage decrease, resulting in higher winter runoff, and reducing the snowmelt peak in late spring. Here, temperature dominates the changes in flow regime.
Figure 5.4 shows monthly runoff for catchments in Belgium (Bultot et al., 1988) and Switzerland (Gellens, 1998) for present conditions and under a climate with more winter precipitation, drier summers and increased potential evaporation. The Swiss catchment (Murg) shows a decrease in March runoff due to reduced snow storage, and an increase in January due to both higher temperature and larger precipitation. The hydrograph for the Semois (Belgium) shows an increase in winter runoff due to higher precipitation amounts and a reduced summer flow due to enhanced evaporation. However, the Aa basin comprises large aquifers, which are recharged in winter with extra precipitation, and which provides sufficient baseflow in summer to maintain runoff as at present, in spite of the drier climate.
Figure 5.4 Monthly runoff from three Belgian and a Swiss catchment (source: Arnell, 1996)
In the next sections the hydrological impacts on the water balance components and the seasonal variations ins runoff is elaborated for the Rhine basin and for different sub-arctic basins.
11 6. Impact of climate change on the hydrological regime of the Rhine
This section summarizes the results of a series of climate impact studies on the Rhine basin in which Utrecht University participated since the 1990s (Kwadijk, 1993; Kwadijk & Middelkoop, 1994; Kwadijk & Rotmans, 1995; Grabs et al., 1997; Middelkoop et al., 2001, Middelkoop et al., 2002; Middelkoop et al., 2004), and results published by KNMI (Shabalova et al., 2003).
6.1 Rhine basin
The Rhine basin (figure 6.1) covers an area of 185,000 square kilometres. On the basis of its geographical and climatological characteristics, the Rhine basin can roughly be subdivided into three parts: the Alpine area upstream Basel, the German middle mountains between Basel and Köln, and the lowland area. The Alpine mountains comprise more than 16,000 km2, with maximum elevations of more than 4000 m a.s.l., about 400 km2 of which are covered with glaciers. The main tributaries in this area are the Aare, Reuss, Limmat and Thur rivers. The German middle mountains comprise the Vogesen and Black Mountains in the south, the Schwabische and Fränkische Alb along the eastern boundary of the basin, and the Rheinische Schiefergebirge in the central-northern area. Maximum elevations range from more than 1000 m a.s.l. in the south to around 600 m a.s.l. towards the north. The main tributaries within the middle part of the basin are the Neckar, Main, Mosel, Lahn and the Sieg. The lowland part comprises extensive sedimentary areas, including loess, (fluvio)glacial deposits, cover sands, and fluvial deposits of the lower river Rhine delta.
Figure 6.1 The Rhine basin and the investigated subcatchments
12 Climatic characteristics of the basin vary considerably for the three major parts of the basin. Within the Alpine area, large differences in precipitation occur, associated to both orographic and convective precipitation. Maximum annual precipitation in the mountains can be as much as 3000 mm, whilst in valleys at the lee side annual precipitation is only 600 mm. A substantial part of the precipitation is temporarily stored in a snow cover. Within the middle mountain area, climate parameters and their spatial variability are increasingly being determined by the site elevation. Whilst average temperatures decrease with elevation, high temperatures occur on sheltered valley slopes. Precipitation generally increases with elevation, with considerably larger annual precipitation at the west-exposed sides of mountain ranges, and low precipitation at the lee sides. In summer, convective precipitation is important within the lower areas. The climate of the lowland part is maritime in character, with lower annual and daily amplitude of temperature than the upstream part of the basin. Annual average precipitation is about 750 mm.
The discharge of the river Rhine is determined by the amount and timing of precipitation, snow storage and snow melt in the Alps, the evapotranspiration surplus during the summer period, and changes in the amount of groundwater and soil water storage. Figure 6.2 shows the present hydrograph of the river Rhine for different gauging stations along its course. The Alpine rivers are governed by a snowmelt regime, with a pronounced maximum in summer. This maximum is generated by storage of precipitation in the snow cover in the winter, and its melting in spring and summer, amplified by summer rains. Retention of water in the Alpine border lakes has a smoothing effect on the Rhine discharge fluctuations. Downstream of Basel, the pluvial regime gradually starts to dominate the Rhine discharge. At the Mosel confluence, the discharge maximum is moved to the winter season, maintaining however a considerable discharge in summer from the Alpine region. In dry periods, like the summer of 1976, the proportion of the discharge coming from the Alps can be as much as 95%. The summer discharge minimum in the central and lowland areas is due to high evapotranspiration during the growing season exceeding the contribution of precipitation to the runoff, in spite of the precipitation maximum in the summer period. During the winter half-year, precipitation falls in the lower parts of the basin predominantly as rain, and eventual snowfall usually melts quickly. Further downstream, the declining contribution of the tributaries to the mean annual runoff is mainly due to the regression of precipitation in the lower parts of the basin.
Figure 6.2 Hydrographs for different gauging stations along the Rhine
The average discharge of the Rhine at the German-Dutch border is about 2,300 m3/s. Important aspects for water management of the lower river reaches are flood protection, inland navigation, water supply to regional water systems and for industry, agriculture and domestic water use. Furthermore, the lower Rhine functions as a major nature area. All these
13 user functions depend on the discharge regime, and their design and demands have been adjusted to the present-day regime. For example, flood protection is based on a design water level that is associated with a flood with a 1250-year recurrence time. However, if climate change might induce an increase in peak flows, this design discharge will increase, and hence flood reduction measures would be required.
6.2 The Rhine Basin Study
In 1989 the International Commission for the Hydrology of the Rhine basin (CHR) initiated a research project for the development of a water management model for the entire Rhine basin. The objective of the study was to assess the impact of climate change on the river Rhine and to investigate the consequences for the user functions of the river, i.e. flood protection, navigation, water availability, using a set of hydrological models. Regarding the difficulties envisaged when developing a detailed model for a basin as large as the Rhine basin, the following approach was chosen. Along a bottom-up line several detailed models were developed for several sub-catchments, while along a top-down line a coarse water balance model was developed for the entire Rhine basin. Using this set of models, the effects of climate change on the discharge regime in various parts of the Rhine basin were calculated for different climate scenarios. This section presents the impact of selected climate scenarios on the hydrological regime of the river Rhine and discusses similarities and differences between the model results. The results of this study have been extensively reported in Grabs et al. (1997) and were published in Climatic Change (Middelkoop et al., 2001).
6.2.1 Method and key areas
Detailed hydrological models with a physical basis that use a daily or shorter time step have been developed for representative subcatchments (<5,000 km2) within each of the main parts of the Rhine basin (figure 6.1). These models are suitable to analyse the effects of changes in climate and land use on average, low and peak discharges in the sub-catchments. A coarse scale water balance model, RHINEFLOW, was developed for the entire Rhine basin. This model enables investigation of the effects of climate changes on monthly average discharges for the entire river Rhine and its main tributaries.
The catchments for which the detailed models were developed are shown in Figure 1 and their main characteristics are summarised in table 6.1. The alpine catchments include the Alpine and pre-Alpine parts of the Rhine basin, with an altitude range between 300 and 2500 m. In these areas, snow storage and snowmelt strongly influence the annual cycle of runoff. Precipitation intensities show a high spatial variability, associated with the large differences in elevation. The Middle Mountain catchments are part of the Mosel basin, and cover an altitude range between 150 and 700 m. The Vecht catchment in the lowland part of the Rhine basin has only minor elevation differences. Here, the sub-soil consists of permeable sedimentary deposits, so that groundwater is an important component in the water balance of the catchment.
14 Table 6.1 Characteristics of the investigated sub-catchments in the Rhine basin. The location of the subcatchments is shown in figure 6.1 Area Altitude range Land use type coverage (%) Annual Annual Annual (m a.s.l.) precip evapotr runoff (km2) Max Avg Min Forest Pasture Meado Urban (mm) (mm) (mm) w + arable Thur 1700 2504 769 356 29 9 52 8 1450 560 890 Murg 212 1035 580 390 29 0 62 8 1220 600 620 Ergolz 261 1169 590 305 40 4 51 5 1080 640 440 Broye 392 1514 710 441 25 2 67 5 1300 710 590 Prüm 150 700 435 150 33 0 55 5 900 460 440 Blies 205 545 330 205 53 0 41 6 930 590 340 Vecht 3800 110 30 5 20 0 75 5 780 495 285
Hydrological models
For the analysis of the detailed catchments, existing rainfall-runoff models were applied, because their concepts and performances had been proven adequate in earlier applications. In view of their application for changed climate conditions all these models were conceptual or physically based, with a varying degree of detail, and they have been adapted to the specific physical conditions of the catchments. The detailed models comprise about 10 to 15 parameters each. These have been calibrated separately for different hydrological components, such as snowmelt (Alpine models) or ground water storage (Lowland model). All models have been validated and calibrated in split-sample tests using observed time series of input data and output discharge (cf. Klemes, 1986). Efficiency coefficients and correlations between observed and simulated discharge varied between 0.8 and over 0.9 for the verification periods, which indicated that all models performed well for the area and at the resolution they were designed for. A more extensive description of the models is given in Grabs et al. (1997). Table 6.2 summarizes the models used.
6.2.2 Climate change scenarios
Climate change scenarios have been provided by the Climatic Research Unit, University of East Anglia with the assistance of the Institute of Hydrology, Wallingford. The construction of these scenarios is based on two General Circulation Models (GCM), the Hadley centre’s high-resolution 11-layer atmospheric GCM (UKHI), and the Canadian CCC model (referred to as XCCC) (Hulme et al., 1994). Using each model, a control integration for present day greenhouse gas concentrations was made, as well as a run with doubled CO2-concentrations. From the results, climate change fields that indicate climate changes per degree global warming were generated. These have been rescaled according to the global warming resulting from a doubling of CO2 concentrations, assuming the IPCC emission scenario IS92a (IPCC, 1996) with a global climate sensitivity of 2.5° Celsius, while ignoring the effect of sulphate aerosols. For each scenario, anomalies of mean monthly temperature, precipitation, wind speed, radiation, and vapor pressure have been determined for the year 2100. These were interpolated down to a grid resolution of 0.5 - 0.5 longitude/latitude. Climate anomalies for the year 2050 have been obtained by linear scaling of the results obtained for 2100.
15 Table 6.2 Summary of the model characteristics Area Thur Murg, Ergolz, Saar Vecht Rhine Broye Model name WaSiM-ETH IRMB HSPF Vecht RHINEFLOW Reference Gurtz et al. Bultot et al. Bicknell et De Laat Kwadijk (1997) (1994). al. (1993) (1992); (1993) Warmerdam et al. (1993) Temporal 1 h 1 day 1 day 1 day 1 month resolution Spatial resolution 100 x 100 m 100 x 100 m 100 x 100 250 x 250 m 3 x 3 km (partly lumped) m No. of land use 10 8 12 12 5 types Number of meteo 22 1 15 1 27 stations Climate data T,P,e,u,G T,P,Ts,e,u,G,Hs T,P,e,u,G T,P,A,e,S,u T,P Reference period 1981–1995 1981–1993 1961–1990 1965–1990 1956–1980 Snow Radiation and Energy balance Energy Not included Linear T accumulation wind balance model correction and Temp-index Evapotranspiration Penman– Modified Modified Penman– Thornthwaite Modified Penman– Penman– Monteith Monteith Monteith Groundwater TOPMODEL Multiple Percolation 2-D steady Recession infiltration and approach storages model to deep state ground term recharge ground water model water is lost from system Runoff TOPMODEL, Separation of Separation Separation Water Separation of overland flow of overland of overland balance, overland flow and base flow flow and flow and separation of and base flow base flow base flow overland flow and base flow Flow routing Translation- Unit Pulse model Muskingum Not included diffusion hydrogrammes T = temperature; P = precipitation; e = water vapour pressure; u = wind speed; G = global radiation; A = air pressure; Hs = snow depth
The monthly climate anomalies have been applied to the available base-line climate series in a straightforward way. Temperature changes were added as absolute changes to the base line series; the other climate parameters were adapted according to their relative changes. Table 6.3 shows the changes in P and T, projected to the year 2050 according to the UKHI and XCCC climate scenarios for different parts of the Rhine basin. All scenarios envisage an increase in annual precipitation, due to an increase of winter precipitation. The temperature rise according to the UKHI scenario is in the order of 2 C, with a greater rise in winter than in summer. The XCCC scenario gives a temperature increase by about 1–1.5 C. The UKHI scenario is drier than the XCCC scenario in terms of atmospheric vapor saturation. Overall, the XCCC experiment yielded the more moderate changes of the two.
16 Table 6.3 Changes in temperature and precipitation in different parts of the Rhine basin according to the UKHI and XCCC experiments, projected to the year 2050 Alpine area Central Germany Lowland area
winter = Nov-Apr; summer = May-Oct
6.2.3 Hydrological changes at the Rhine basin scale
Figure 6.3 shows the hydrologic responses obtained using the RHINEFLOW model for the UKHI and XCCC scenarios for different stations along the Rhine and its main tributaries. The hydrograph for the entire basin (station Rees) shows a rise of winter flow and a reduction of summer flow. Within the basin, the largest increases in winter flow are found for the Alpine area (Brugg, Rheinfelden). When going from the Alps downstream along the Rhine, the winter increase is damped because of the smaller increase in winter flow from the tributaries (Neckar, Main, Mosel) in Germany, but still is present in Rees. The reduction in summer flow is largest in the Alps, too. In the central part of the basin, the RHINEFLOW model indicates a small decrease of summer flow, such that this reduction is still present at the Rees station downstream. The UKHI scenarios generally resulted in greater changes than the XCCC scenarios. Nevertheless, both climate models indicate a shift of the hydrological regime in the entire Rhine basin. In the upper Alpine area the intra-annual difference between low winter flow and high summer flow decreases (and even may be inverted), while in the lower parts the existing summer-winter differences are amplified.
Figure 6.3 Monthly average discharge at different gauging stations in the Rhine basin for the UKHI and XCCC scenarios, calculated using the RHINEFLOW model. Alpine area: Brugg (Aare) and Rheinfelden (Rhine); main tributaries in Germany: Rockenau (Neckar), Kleinheubach (Main), Cochem (Mosel); entire basin: Rees (Rhine).
17 Figure 6.3 - continued -
6.2.4 Hydrological changes at the catchment scale
Figure 6.4 shows the changes in monthly discharge according to the UKHI2050 scenario for different parts of the Rhine basin, based on the detailed models. The main trends that were found per subregion within the Rhine basin can be summarized as follows.
Alpine Area In the Alpine area, higher temperatures will reduce the amount of snow accumulation during winter. This results in higher winter discharge, and lower summer discharge. In addition, winter precipitation increases, while precipitation may decrease in some summer months. Higher temperatures will intensify evapotranspiration, particularly during summer. On an annual basis, this increase is larger than the precipitation increase, resulting in a reduction of annual runoff. When comparing responses of the different catchments in more detail, major differences show up. Depending on the altitude ranges of the catchments, the maximum daily flow may either increase or decrease. Winter peak flows in the high-alpine area generally increase, especially for floods with a return period of more than 10 years. In pre-Alpine areas, however, this increase is less significant. Changes in summer peak flows could not be well determined by the models, since these are largely generated by convective storms, which demands a much finer temporal and spatial modeling scale. Summer minima decrease in all cases.
German Middle Mountains Area In the German Middle Mountains, the investigated catchments demonstrate only a minor seasonal shift in river flow. The changes in runoff are controlled by the balance between increased precipitation on the one hand, and increased evapotranspiration rates due to higher temperatures on the other hand. This balance depends both on the expected climate changes and on the present climate and land use. In the investigated cases, the accelerated evapotranspiration seems to counterbalance the higher precipitation, resulting in a slight reduction of average runoff during winter, and a much greater reduction during summer. Depending on the severity of net precipitation shortage in summer, the soil water deficit at the end of summer becomes larger, and results in a considerable time lag (weeks to months) until it is recharged by precipitation. Peak flows resulting from heavy rainfall and convective thunderstorms, however, are expected to increase. The differences in response between catchments are considerably smaller than in the Alps.
Lowland Area In the lowland area, increased winter precipitation will cause higher winter discharge and winter peak flows. Under conditions of the UKHI2050 scenario, annual peak flows increase by the order of 20%. During summer, higher evapotranspiration levels cause a net
18 precipitation deficit, reducing discharge in late summer by about 5%. It may take several weeks before the deficit in groundwater storage is replenished by precipitation.
Broye catchment - Monthly Eact and discharge 140 Eact, present Eact, UKHI2050 120 Q present Q UKHI2050 100 h t n
o 80 m / 60 m m 40 20 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month A Broye catchment - monthly snow storage 60 h t n o 50 present m /
m 40 m
t UKHI2050 n e
l 30 a v i
u 20 q e
r e
t 10 a w 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month B
Prum catchment - monthly Eact and discharge 140 Eact, present Eact, UKHI2050 120 Q present Q UKHI2050 100 h t n
o 80 m / 60 m m 40 20 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month C
Vecht catchment - monthly Eact and discharge 140 Eact, present Eact, UKHI2050 120 Q present Q UKHI2050 100 h t n
o 80 m / 60 m m 40 20 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month D Figure 6.4 Hydrological changes in different parts of the Rhine basin according to the UKHI scenario projected to the year 2050, based on the detailed models. (A) Alpine area – Broye catchment, monthly actual evapotranspiration and discharge (mm); B: monthly snow storage (water equivalent, in mm). (C) German Middle Mountain area – Prüm catchment, monthly actual evapotranspiration and discharge (mm); D: Lowland area – Vecht catchment, monthly actual evapotranspiration and discharge (mm).
19 6.2.5 Discussion: comparison of modeling results
In addition to the general trends described above, this study also examined the effects of different modeling resolutions, differences in the level of detail in which the models represented the hydrological processes, different climate scenarios, and the characteristics of the investigated catchments. These are discussed below.
Model results at different scales It should be emphasized here that the comparison does not aim at identifying model errors, or bad performance of a model. Each of the models has been tested and has proved to be adequate for the catchment and the scale it was developed for. The comparison, therefore, focuses on analyzing to what extent and why different model concepts and modeling scales influenced the model results. When considering the changes in river flow on a monthly basis, the general trends found using the high-resolution models are similar to those obtained using the coarse RHINEFLOW model. The degree of the changes, however, is sometimes different. In figure 6.5 a comparison is made on the basis of the UKHI2050 scenario.
For the Alpine area, the RHINEFLOW model envisages a larger increase of the winter discharge than the detailed models. This may reflect a larger contribution of changes in the amount of snow storage during winter. The area on which the RHINEFLOW results were based included the highest parts of the Alps, containing large volumes of snow. In contrast, the highest points of the detailed models are considerably lower, resulting in a different importance of snow storage. Part of the differences found between the Ergolz and the other catchments can be explained from differences in snow storage during winter as well. The high degree of spatial variability of radiation and the role of temperature inversions inherent to mountain areas all affect snow melt and evapotranspiration, and therefore they were well represented in the detailed alpine models. This was not possible in the coarser RHINEFLOW model. Using monthly averages of temperature and precipitation in the RHINEFLOW model may occasionally result in different estimates of the amounts of snow storage and snowmelt. For example, when the average temperature in a month is below zero, RHINEFLOW stores all precipitation in that month as snow. However, if a cold month ends with a week of thaw and rain, the average temperature may be below zero, but there is runoff to the river. The differences for the German Middle Mountains (Saar basin) seem to be caused by the representation of evapotranspiration processes (cf. Table II). Generally, the detailed models suggest an overall runoff decrease between 5% in winter and 25% in autumn, while the RHINEFLOW model suggests an increase in winter runoff and a much smaller decrease during summer. The concept of Thornthwaite used in the RHINEFLOW model to calculate evapotranspiration may have underestimated the effect of higher temperatures, and did not take the changes in air vapor pressure into account. Differences between the RHINEFLOW results for the area downstream of Andernach and the Vecht basin in the lowland part of the basin can be both explained by the physical differences between these areas, and by different ways of representing evapotranspiration. The effects seem the strongest for the summer period. This might be caused by the role of groundwater. In the Vecht basin, groundwater flow contributes substantially to the runoff in this river. This ground water flow is well represented by the Vecht model, while in the RHINEFLOW model groundwater flow is represented simply by a linear recession equation. In general, the catchments of the detailed models are considerably smaller than the subsections of the Rhine basin from the RHINEFLOW model that were considered for comparison with the detailed models. As a result, typical characteristics and local conditions, such as elevation, geology, land use within a sub-catchment can be different from the average situation in the larger area evaluated by the RHINEFLOW model. Examples are karst phenomena within the Ergolz catchment; large forest coverage in the Blies catchment; the importance of ground water storage in the flat sedimentary Vecht catchment.
20 A: Alpine + pre-Alpine area; UKHI2050 80 Rheinfelden (RHINEFLOW) 60 Limmat (RHINEFLOW) 40
e Thur g
n 20
a Ergolz h c 0 Broye % -20 -40 -60 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month
B: German Middle Mountains UKHI2050 15 10 5 0 e
g -5 n
a -10 h c -15 Prum % -20 Blies -25 Cochem (RHINEFLOW) -30 Mettlach (RHINEFLOW) -35 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month
C: Lowland area UKHI2050 25 20 15 10 e
g 5 n
a 0 h c -5 % -10 Lowland (RHINEFLOW) -15 Vecht -20 -25 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec month
Figure 6.5 Comparison of the results of the RHINEFLOW model with the detailed models for the UKHI2050 scenario – changes in monthly average discharge (%)
Different scenarios The greater changes observed for the UKHI scenario when compared to the XCCC scenario (figure 6.3) are in agreement with the larger changes in the climate variables resulting from the UKHI experiment. The higher winter discharge according to the UKHI scenario is due to a higher temperature rise causing a larger reduction of snow storage, and a greater increase of winter precipitation. The lower summer flow under UKHI conditions is due to the smaller snowmelt contribution, and to a more intensified evapotranspiration due to the higher temperature rise and drier atmospheric conditions. Nevertheless, the trends of the responses are the same for both climate models.
21 6.2.6 Estimation of peak flows
From the model results, it is difficult to achieve reliable estimates of peak flows under changed climate conditions. Peak flows in small areas depend very much on precipitation characteristics, such as convective storms and length of wet periods. In this study changes in precipitation were implemented in a rather simplified way, as the percentage of precipitation increase has been evenly distributed over the whole range of present day precipitation. This may lead to inconsistent estimates of precipitation extremes under changed climate conditions. In addition, the method assumes that the number of days with precipitation remains unchanged. Under changed climate conditions with higher temperatures, it may be expected that convective high intensity precipitation may occur more frequently. However, as the size of such storms is small, estimations of floods in larger catchments are less sensitive for individual events. Estimating effects in larger Alpine areas from the response in small catchments is a precarious task. Floods in larger catchments occur mainly in winter as result of large-scale frontal rainfall. In the Alpine area, a decrease in the number of flood days (discharge larger than the p95 fractile) is foreseen in summer, but an increase may occur in winter. Overall, a tendency to more contrasted streamflow regimes with more abundant winter flooding and summer flood due to convective storms seems to be produced. Annual peak flows with return periods of over 50 years may increase by about 10% until the year 2050. Average discharge in the German Middle Mountains reduces due to accelerated evapotranspiration, and the model results suggest a reduction in the frequency of peak flows. Nevertheless, the magnitude of peak flows resulting from heavy rainfall and convective thunderstorms are expected to increase. Since in winter both rainfall and the melt water runoff contribution from the Alps are expected to increase, peak flows in the middle and lower Rhine River will increase. Unfortunately, changes in discharge extremes could not be determined directly with the RHINEFLOW model, because peak flows are masked by the low temporal resolution of the model. Alternatively, a statistical method, based on the relationship between monthly average discharge and peak discharge, was applied to achieve estimates of peak flows in the downstream part of the Rhine basin (Kwadijk and Middelkoop, 1994). These estimates suggest that peak flows of the lower river Rhine with recurrence times in the order of 100 to 1000 years may increase by about 5–8% by the year 2050.
6.2.7 Low flows
The reliability of the simulation results is higher for low flow conditions, because periods of low flow are characterised by a lower temporal variability (in the order of weeks), which is more in accordance with the temporal resolution of the climate scenarios. A major uncertainty for estimates of low flow is caused by uncertainty in evapotranspiration. The changes in transpiration by plants and the effect of increased CO2 concentrations on the biomass production as well as on the stomatal resistance and transpiration efficiency of the plant leaves require further attention. A reduction of low flows was found for all applied scenarios and in all catchments. In the Alpine catchments the decrease varied between about 10% and 30% for the XCCC2050 scenario and 20% to 40% for the UKHI2050 scenario. In the German Middle Mountains low flows decreased by about 10% under XCCC2050 conditions, and up to 20% under UKHI2050 conditions. In the lowland catchment the decrease of low flows was only a few percent. For the entire Rhine basin, summer low flows reduce by about 5% for the XCCC2050 scenario and 12% under conditions of the UKHI2050 scenario.
6.3 Recent impact studies for the Rhine
As a follow-up the first Rhine basin study, several climate impact studies were carried out for the Rhine basin using a 10-daily version of RHINEFLOW, with a resolution of 1 km x 1 km, and using a wider range of climate scenarios, and projections until the year 2100 (Middelkoop & Kwadijk, Wengen; Shabalova et al., 2003). These scenarios included a maximum annual temperature rise in the Rhine basin up to about 5°C, and precipitation changes varying from
22 10% increases in the northern parts to a 10% decrease in the Alps. Precipitation decreases mainly occur in summer. While these studies again used climate anomalies calculated by GCMs to adapt a baseline time series of climate data, Shabalova et al. (2003) also included changes in climate variability in the scenarios. These studies confirmed the general pattern of changes resulting from the previous studies, but gave a more precise quantification of changes in extreme flows. Under conditions of extreme climate change, mean winter discharges of the Rhine might increase by up to 30% by the year 2100, while discharge in late summer might decrease by about 30%. Design flows for flood protection (1250-yr recurrence time) works might increase by about 10-15%.
6.4 Conclusions and implications for water resources and water management
The bandwidth of the simulation results is wide. This is primarily the result of uncertainty in the climate scenarios. A second concern is the way regional climate scenarios (in particular for precipitation) can be derived from downscaling GCM results. Different ways of downscaling the input climate scenarios might give different estimates of the changes. Finally, a minor part of the uncertainty is caused by disagreements among the model results. Nevertheless, all studies indicate similar impacts. Due to climate change the river Rhine is expected to shift from a combined rainfall-snowmelt regime to a more rainfall dominated regime. This coincides with a seasonal change in the discharge regime: winter discharge will increase, and summer discharges decrease. The frequency and height of peak flows will increase. During summer, periods of low flow will occur more frequently and last longer. These changes can have considerable socio-economic implications:
Flood Defense The frequency and magnitude of peak flows is expected to increase. Due to the rise of the 0ºC-line in the Alps and the resulting degradation of the alpine permafrost, mass movements and rockslides may occur over larger areas. In the Alpine area peak flows may increase by over 10%. Along the Rhine in the Netherlands the design discharge for flood protection might increase by about 5–8% over the next 50 years, and by over 10% until the year 2100.
Inland navigation An increased frequency of flood periods will hamper inland navigation on the Rhine more often. Longer periods of low flow will also increase the average annual number of days during which inland navigation is hampered or stagnates. When the Rhine discharge drops below about 1000 to 1200 m3/s, ships on the major transport route Rotterdam-Germany-Basel cannot be fully loaded, and transporting cost rise. The average annual number of days that the Rhine discharge at Lobith is below 1000 m3/s may increase from 19 (under present day conditions) to 25 – 35 days, depending on the applied scenario. Current projects on channel improvements can only partly alleviate these problems.
Hydropower Generation Due to the increased winter discharge hydropower generation is expected to increase during this season. In Switzerland, the water availability for power generation may increase over the entire year, whilst further downstream the annual production decreases.
Water Availability for Industry, Agriculture and Domestic Use Low flow periods during summer reduce water availability for industrial use and for drinking water production. During these periods the water demand for agriculture is expected to increase due to higher temperatures. Also, the use of river water for cooling purposes may be limited, not only because of a reduced river flow, but also because of higher water temperatures. In general, climate change will increase the water demand by various sectors, particularly during summer when water availability is low, and will require an even more balanced water-resources management.
23 7. Water Balance Modelling of (Sub-)Arctic Rivers and Freshwater Supply to the Barents Sea Basin
Shortened version of a paper with the same title, by E.A. Koster, R. Dankers and S. Van der Linden from the Department of Physical Geography of Utrecht University, published in: Permafrost And Periglacial Processes (2005), DOI: 10.1002/ppp.510.
7.1 Introduction
In the framework of two EU-funded research projects, “Barents Sea Impact Study” (BASIS) (Lange & The BASIS Consortium, 2003) and “Tundra Degradation in the Russian Arctic” (TUNDRA) hydrological modelling studies of two river catchments of intermediate size (104- 105 km2), that both drain into the Barents Sea Basin (figure 7.1; table 7.1), were executed. It concerned the Tana River in northern Fennoscandia (Dankers, 2002; Dankers & Christensen, 2004) and the Usa River, which is a tributary of the Pechora river in northeast European Russia (Van der Linden, 2002; Van der Linden & Woo, 2003; Van der Linden & Christensen, 2003; Van der Linden et al., 2003). The Pechora River catchment occupies a significant part (324,000 km2) of the total Barents Sea Basin catchment and consequently supplies a major portion of the freshwater discharge to the Basin. The mean annual inflow to the Barents Sea of the Pechora River is only known approximately and is estimated at c. 130 to 160 km3 (Peterson et al., 2002), whereas the total inflow of all rivers to the Barents Sea Basin is estimated at c. 478 km3 (Ivanov, 1999). The Tana River catchment is one of the largest, northwards draining catchments (16,000 km2) in Fennoscandia.
The main goal of this study was to assess potential changes in freshwater supply to the Barents Sea/ Arctic Ocean, based on water balance modeling and climate scenarios. Specific objectives were to: (a) adjust an existing water balance model to the specific (Sub-)Arctic conditions found in the Tana and Usa River Basins, and (b) assess the impact of climate change scenarios on the freshwater supply of these rivers to the Barents Sea.
Table 7.1 Catchment characteristics and runoff regimes Data Tana river Usa river Catchment size 16,000 km2 93,000 km2 Hydrograph character subarctic nival river regime subarctic nival river regime Mean annual discharge 166 m3/sec, highly variable 1091 m3/sec, highly variable Snowmelt peak runoff 1,500 – 3,000 m3/sec 6,000 - 15,000 m3/sec Mean annual air temperature -0.5 to -3°C -3 to -7°C Mean annual precipitation 340-460 mm 400-800 mm Permafrost (predominantly) sporadic to discontinuous discontinuous to continuous
7.2 Usa basin: river runoff and climate impact
The catchment of the Usa River (Van der Linden, 2002; Van der Linden et al., 2003) has an area of 93 000 km2 (table 7.1) and is bordered by the Ural Mountains in the east. The catchment consists mainly of lowland areas and is located for the major part in the discontinuous permafrost zone (defined as 20–95% frozen ground) with continuous permafrost (>95%) in the northernmost part and in the Ural Mountains, and seasonal frost (<20%) to the southwest. Permafrost thickness may reach up to 700 m. Unconsolidated deposits in the lowlands of the Usa catchment mainly consist of thick sequences of glacio- lacustrine, fluvial and aeolian sediments. Mean annual temperature ranges from -3°C in the south to -7°C in the northernmost regions. The annual amplitude is very large: winter temperatures as low as -55°C have been measured, and summer temperatures up to 35°C occur. In the lowland areas mean annual precipitation ranges from about 400 to 800 mm, whereas precipitation values up to 950 mm are reached in the Ural Mountains. Vegetation in the southern part of the Usa Basin varies from extensive spruce forests on the taiga foothills
24 of the Ural mountains, to taiga lowlands with sphagnum bogs and pine and birch forests and occasionally willows along rivers. To the north, the taiga zone gives way to a forest tundra with spruce and birch trees, and subsequently to a shrub tundra covered by dwarf birch and heath and a moss–lichen tundra on the driest parts of the plains.
Figure 7.1 Locations of the Tana and Usa river catchments
Methodology
The sensitivity of the Usa River to changes in climate was evaluated with a spatially- distributed water balance model, USAFLOW, which is an adapted version of the based on the RHINEFLOW model developed by Kwadijk (1993) for the River Rhine (figure 7.2). The spatial resolution was 1 by 1 km. The temporal resolution was maintained at monthly timesteps, but in order to examine potential changes in peakflow, the model was also applied on a 5-day basis to two sub-basins of the Usa catchment. USAFLOW has been adapted for permafrost conditions by introducing variable (in space and time) separation coefficients that effectively control the extent of infiltration and percolation to the groundwater reservoir, the remainder being considered as rapid runoff. In areas of continuous permafrost it was assumed that infiltration would be impossible, and that 100% of the excess water is discharged as rapid runoff. In areas with discontinuous permafrost, sporadic permafrost and seasonal frost this ratio was 90%, 80% and 60% respectively, meaning that even in the latter only 40% of the effective precipitation was allowed to drain to the groundwater reservoir and contribute to the
25 baseflow runoff. In the scenario run it was assumed that the rise in temperature of the near- surface permafrost would be similar to the increase in air temperature, resulting in a shift of the permafrost zones. In the control run, the USAFLOW model was forced with observed climatology. When comparing the results with observations on river discharge, the model appeared to simulate both the amount and the seasonal distribution of the Usa River reasonably well (figure 7.3A). Climate change scenarios were derived from the GCM HadCM3 (Gordon et al., 2000). From this model the mean monthly anomalies between control and scenario runs were used to adjust observation records. The considered scenario was based on the climate forcing according to the SRES A2 emission scenarios, which results in the largest climate changes when compared to the other IPCC marker scenarios (Houghton et al., 2001). The scenario runs correspond to a 30-year period between 2070 and 2100.
Figure 7.2 Flow diagram of the water balance models (TANAFLOW and USAFLOW), adapted after Kwadijk (1993)
A B Figure 7.3 Observed and simulated discharge of the Usa River at the Adzva station (A) and of the Tana River at the Polmak, Norway station (B). The observations are averages over 1951-1980 for the Usa and over 1961-1990 for the Tana; the simulated discharge is based on a 30-year control run corresponding to a greenhouse gas forcing of 1961 (after Van der Linden, 2002 and Dankers, 2002).
26 Impact of Climate Change
Under the A2 scenario, the temperature and precipitation in the Usa Basin are projected to increase dramatically by the end of the twenty-first century. The rise in basin-averaged mean annual temperature amounts to +6.8°C, ranging from +5.2°C in the summer months (June– August) to more than 10°C in December. The annual precipitation increases by 28%, with the largest changes again occurring in the winter months (+42% in December–February). These changes have a significant impact on the water balance (table 7.2).
Table 7.2 Water balance of the Tana and Usa River Basins in the control and scenario runs. All quantities are annual averages over 30 years and expressed in mm/year. The observed discharge is the discharge measured at Polmak (Norway) in the period 1961-1990 for the Tana River, and at Makarikha (Russia) in the period 1941-1970 for the Usa River. Makarikha drains only 71% of the entire Usa Basin, which explains the difference between observed and simulated river discharge (here normalized by drainage area) Tana Usa Control Scenario % change Control Scenario % change Precipitation 508 634 +25 600 768 +28 Sublimation 90 63 –30 N/A N/A N/A Evapotranspiration 59 63 +7 154 210 +36 Discharge, observed 368 503 Discharge, simulated 361 502 +39 443 554 +25
The evaporative sum as calculated by USAFLOW increases by 36%, which is due not only to the higher summer temperatures, but also to the decrease in permafrost extent in the scenario run, leading to more water available from soil storage. Particularly in areas of discontinuous permafrost that are expected to change into sporadic permafrost and thawed ground, the infiltration capacity increases significantly. In the winter season, the additional precipitation is temporarily stored as snow and mostly discharged in spring and late autumn (figure 7.4A). Consequently, on an annual basis the river runoff increases considerably, and the total outflow of the Usa River changes from 42 to 52 km3/year. The changes in the seasonal patterns in river discharge were analyzed in more detail in two sub-basins. The Kosyu catchment has its headwaters in the Ural Mountains (up to 1900 m high). Although snowmelt advances more rapidly in the scenario run, the timing of the peakflow remains largely unchanged (figure 7.5A). However, since most of the snow cover has disappeared, discharge is much lower in June. In the lowland tundra part of the Kosedayu river catchment the rise in temperature results in a major shift in snowmelt runoff by about one month, from early May to early April. In the scenario run the peak discharge is some 20% higher as well (Figure 5B). Over the whole of the Usa Basin, spring discharge increases by 20%, and the maximum runoff shifts from June to May (figure 7.4A). In summer the percentage changes are rather small, but in autumn and winter the runoff shows a dramatic increase (+67% in September– November, and +119% in December– February), although the absolute values remain very low. In other words, the period of low flows in winter begins about one month later and ends one month earlier, with a major increase in discharge in early spring at the expense of runoff in June.
27 A B Figure 7.4 Simulated mean monthly discharge of the Usa river at the Adzva station (A) and of the Tana river at the Polmak station (B) under current and changed climate conditions of the SRES scenario A2. A: Change in the Usa hydrograph according to a GCM HADCM2 scenario projection for 2070-2099. In this case the climate model anomalies were used to adjust observation records. B: Control run (30 years corresponding to a greenhouse gas forcing of 1961) and scenario run (2071-2100) used climate data derived from the RCM HIRHAM 4
A
Figure 7.5 Present and simulated (for the period 2070-2099) mean 5-daily discharge for the Kosyu (A) and Khosedayu (B) river, tributaries of the Usa river, using the HADCM3 A2 scenario
28 7.3 Tana basin: river runoff, snowmelt and evaporation
The second study that is discussed here (Dankers, 2002; Dankers & Christensen, 2005) focused on the Tana Basin with an areal extent of ca. 16 000 km2. The Tana River catchment is located in the northernmost part of Fennoscandia (figure 7.1) and has a physiography distinctly different from the Usa Basin (table 7.1). Although much smaller than the Usa, the Tana River (Tenojoki in Finnish) is one of the largest rivers in Fennoscandia flowing into the Barents Sea, and it drains an extensive upland area. Mean annual temperatures in the Tana Basin range from well below -2°C at inland meteorological stations to slightly below 0°C near the Barents Sea. The annual precipitation is relatively low, between 340 mm and 460 mm, with most rain usually falling in the summer months. The vegetation consists predominantly of extensive forests of mountain birch although some isolated pine forests can be found in the river valleys. In places large peat bogs have been formed, while tundra heaths and bare mountaintops dominate the landscape above the treeline. Most of the Tana Basin is underlain by Precambrian bedrock, covered with glacial tills that were mainly deposited during the Weichselian glaciation. Permafrost occurs only sporadically, and is primarily found in palsa mires and on bedrock surfaces.
Methodology
The hydrological model used in the Tana Basin, TANAFLOW, is also based on the RHINEFLOW model, and was was applied on a 10-day basis with a spatial resolution of 1 by 1 km. As permafrost is of limited importance in the Tana Basin, the variable separation coefficients of USAFLOW were not used here; however, the influence of frozen soils on runoff generation in spring was taken into account by constraining infiltration into the soil during snow coverage. Attention was turned to developing physically-based descriptions of snowmelt and evapotranspiration. Snow accumulation and snowmelt in TANAFLOW are simulated using an energy balance module (TANASNOW) (Dankers, 2002), and evapotranspiration is calculated according to the Penman–Monteith model. All necessary meteorological input data (precipitation, temperature, radiation, humidity and wind speed) were derived from the regional climate model HIRHAM4 (Christensen et al., 1998; Christensen & Kuhry, 2000) that was driven by the global model ECHAM/OPYC (Roeckner et al., 1996). Here, the simulations encompass a control run of 30 years with a greenhouse gas forcing corresponding to the 1960s, and a scenario run according to the A2 emission scenario corresponding to the period 2070–2100. These allow comparing the results with those for the Usa Basin. Compared with meteorological observations in the Tana Basin, the HIRHAM estimates of temperatures in spring are slightly too low in the control run, which leads to a delay in the snowmelt simulated by TANASNOW. As a result, TANAFLOW simulates the annual discharge peak exactly one time step of 10 days too late (figure 7.3B). Nevertheless, the annual amounts and seasonal variations are represented satisfactorily.
Impact of Climate Change
In the Tana Basin the A2 scenario implies a significant temperature and precipitation increase. The basin-averaged mean annual temperature rises more than 5°C, from -3.0°C to +2.2°C, with the largest warming again in the winter months (+6.3°C in December– February), and slightly less in summer (+3.5°C in June–August). The annual precipitation is projected to increase by 25% and also in this case the largest increase occurs in the winter months (+38%). Looking at the annual water balance (table 7.2), the scenario run shows a decrease in sublimation, which is mainly due to a much shorter snow season. Annual evapotranspiration increases only slightly, and again, this is mostly because of a shorter snow season. In the rest of the summer the evaporation rates are very similar to the control run, in spite of the higher temperatures. The annual river discharge increases by almost 40%, which is proportionally even higher than the increase in precipitation. Again, this can be explained by the extra
29 precipitation in winter that contributes to snowmelt runoff in spring and late autumn, and not so much to more evaporation in summer. This means that the annual freshwater influx from the Tana River into the Barents Sea increases from ca. 5.3 to 7.3 km3. Remarkably, in the scenario run the annual discharge peak of the Tana River not only occurs 20 days earlier, but is also about 10% higher (figure 7.4B). This is due to a larger snow accumulation earlier in the season, which depends mainly on the amount of precipitation in spring. Other model experiments in the Tana Basin did show a decrease in the peak runoff; however, the shift in time by about 20 days was a common feature (Dankers, 2002). In relative terms, the changes in river discharge are even larger during the rest of the year. As the evaporation rates in summer are similar to those in the control run, the summer (July–September) discharge of the Tana River increases in the scenario run by 46%, and due to a higher baseflow the winter runoff is even 55% higher. Nevertheless, even under the A2 scenario, the overall pattern is still very much dominated by the snowmelt runoff peak in spring (figure 7.4B).
7.4 Discussion
Under the A2 scenario, both the Tana and the Usa show a significant increase in the annual freshwater flux to the Barents Sea by the end of the twenty-first century, by about 25% in the Usa River and almost 40% in the Tana. These figures are comparable to those obtained in previous modelling studies based on doubled atmospheric CO2 concentrations (for example, Vehviläinen & Lohvansuu, 1991; Miller & Russell, 1992; Van Blarcum et al., 1995), but higher than the estimate by Shiklomanov et al. (2000) for the total annual inflow into the
Arctic Ocean at doubled CO2 concentrations. Water balance calculations by Ivanov (1999) for doubled CO2 conditions also indicated an increase of the annual runoff of the Pechora River by 8–40%. Summarizing, it may be concluded that the results presented in this study correspond well with general assessments of discharge changes due to climate change.
When comparing the results for the Tana and Usa rivers, one should be aware that, although the models used started from the same premises, they are not exactly the same and, in each study, different aspects of the hydrological system were emphasized. In the Tana Basin, the total amount of sublimation was reduced by 30% in the scenario run, which is mainly due to a much shorter snow season. Only to a very limited extent, this reduction was balanced by an increase in evaporation (+7%), and the remainder contributed to the increase in spring runoff. On the other hand, in the Usa Basin sublimation was not accounted for at all. Instead, the annual amount of evapotranspiration increased by 36%, which is related to an increased soil storage in summer due to a decrease in permafrost extent. This underlines the relevance of permafrost degradation to the hydrological response of areas currently within the discontinuous and continuous permafrost zone. Monitoring permafrost temperatures since the 1970s in the Usa Basin has revealed that the permafrost is highly dynamic, even over relatively short-term climatic fluctuations (Oberman & Mazhitova, 2001). An increase in active layer depths by, on average, 11% to 15% has been measured in the period 1970–1995. This is ascribed to a simultaneous increase in mean annual air temperature of ca. 1°C. In the Tana Basin, having only a sporadic permafrost extent, a more plateau-like landscape and underlain by bedrock, changes in permafrost were assumed to have much less influence. Therefore, this factor was not accounted for in the current analysis. Also the effective constraint of seasonal frost on the infiltration of snow meltwater was not changed in the scenario run, the assumption being that under future conditions the soil under the winter snow pack will still be frozen—even under the A2 scenario the snow season in the Tana Basin still lasts 5 to 7 months. Therefore, future studies should strive to take account of all possible feedback effects related to snow, vegetation and soil dynamics, in order to come up with more accurate predictions of the impact of climate change on discharge characteristics of (Sub-)Arctic rivers and on freshwater supply to the Arctic Ocean.
30 An increase in the total annual freshwater inflow into the Arctic Ocean at doubled CO2 conditions in the order of at least 10% to 20% seems to be generally accepted (Shiklomanov et al., 2000). In the Tana Basin even a precipitation increase of 25% resulted in a more than proportional increase of runoff, as most of the extra snowfall in winter is discharged in spring, instead of evaporated during summer. In both river basins the higher temperatures in the scenario runs resulted in a spring flood occurring 20 days to 1 month earlier, as was also found by Ivanov (1999).
Under present as well as future conditions warm spring temperatures will melt the bulk of the snowpack within a few days or weeks, while the soil is still frozen. Frozen ground has a low storage capacity and infiltration of water is impeded, especially when the ground ice content is relatively high. Therefore, it is not surprising that the height of the discharge peaks in the simulation studies remained more or less equal or increased somewhat.
The results of the two studies discussed here should only be seen as tentative for a number of reasons. First, important feedback effects, for example caused by the vegetation response to changing environmental conditions, have not been taken into account (see Harding et al., 2002). Second, discharge values are ultimately determined by the difference between precipitation and evapotranspiration. Third, factors that will have a substantial impact on evapotranspiration include the type of vegetative cover, the length of the snow-free season, and the length of the growing season, and all of these can be expected to change under different climatic conditions. Finally, the A2 scenario that we used in our analyses should merely be considered as an extreme case and not as a forecast for future conditions. If any of the other IPCC marker scenarios is used, the magnitude of the changes is different from what is presented here, although the patterns remain more or less the same.
7.5 Conclusions
The hydrographs of the Tana and Usa Rivers under changed climate conditions show overall patterns that remain strongly dominated by snowmelt runoff peaks in spring. Discharge amounts during peakflow remain more or less the same or show a slight increase. However, the timing of the peak discharges reveal distinct differences between the scenario and control runs. An earlier occurrence of the peakflow events by about 20 days or more is a common feature. Concerning the annual discharge amounts, a strong increase of 25% for the Usa River and as much as 39% for the Tana River is simulated in conformity with projected increases in precipitation. Differences in response of the two investigated catchments are probably due mainly to differences in evapotranspiration, reflecting different permafrost conditions. Obviously, the resulting increases in the annual freshwater influx from the Tana River (from 5.3 to 7.3 km3) and from the Usa River (from 42 to 52 km3) into the Barents Sea are insignificant in absolute terms. But in relative terms they agree remarkably well with earlier estimates of changes in freshwater inflow by the other, very large (Sub-)Arctic rivers.
31 References
Arnell NW, 1992, Factors controlling the effects of climate change on river flow regimes in a humid temperate environment, Journal of Hydrology 132, 231-342. Arnell NW, 1996, Global Warming, River Flows and Water Resources. Wiley, Chichester. Arnell NW, 2005, Implications of climate change for freshwater inflows to the Arctic Ocean, Journal of Geophysical Research 110, D07105, doi: 10.1029/2004JD005348, 2005. Bárdossy A, Plate EJ, 1992, Space-time model of daily rainfall using atmospheric circulation patterns, Water Resources Research 28(5), 1247–1259. Bergström S, Carlsson B, 1994, River runoff to the Baltic Sea: 1950-1990, Ambio 23, 280-287. Bergström S, Carlsson B, Gardelin M, Lindström G, Pettersson A Rummukainen M, 2001, Climate change impacts on runoff in Sweden - assessments by global climate models, dynamical downscaling and hydrological modelling, Climate Research 16, 101-112. Bergström S, Forsman A, 1973, Development of a conceptual deterministic rainfall-runoff model, Nordic Hydrology 4, 147-170. Bergström, S, 1995, The HBV model. Computer Models of Watershed Hydrology. In: Signh VP (ed), Water Resources Publication, Colorado, USA. Beven KJ, 2001, Dalton Medal Lecture: How far can we go in distributed hydrological modelling?, Hydrology and Earth System Sciences, 5(1), 1-12. Beven KJ, Kirkby MJ, 1979, A physically based variable contributing area model of basin hydrology, Hydrol. Sci. Bull. 24(1), 43–69. Beven KJ, Lamb R, Quinn P, Romanowicz R, Freer J, 1995, TOPMODEL. In Singh VP (Ed) Computer Models of Watershed Hydrology, Chapter 18, Water Resource Publications: 627-668. Bicknell B R, Imhoff JC, Kittle JL, Donigian AS, Johanson RC, 1993, Hydrological Simulation Program – Fortran (HSPF): Users Manual for Release 10, U.S. EPA Environmental Research Laboratory, Athens, Georgia. Brandsma T, 1995, Hydrological impact of climate change: a sensitivity study for the Netherlands, PhD thesis, Technical University, Delft. Bultot F, Dupriez GL, Gellens D, 1988, Estimated Regime of Energy-Balance Components, Evapotranspiration and Soil Moisture for a Drainage Basin the Case of a CO2 Doubling, ClimaticChange 12, 39–56. Bultot F, Gellens D, Schädler B, Spreafico M, 1994, Effects of Climate Change on Snow Accumulation and Melting in the Broye Catchment (Switzerland), Climatic Change 28, 339–363. Chiew FHS, Whetton PH, McMahon TA, Pittock AB, 1995, Simulation of the impacts of climate change on runoff and soil moisture in Australian catchments. Journal of Hydrology 167, 121-147. Christensen OB, Christensen JH, Machenhauer B, Botzet M, 1998, Very high-resolution regional climate simulations over Scandinavia – Present climate, Journal of Climate 11, 3204-3229. Cohen SJ, 1991, Possible impacts of climatic warming scenarios on water resources in the Saskatchewan River sub-basin. Water Resources Bulletin 19, 291-317. Dankers R, 2002, Sub-arctic hydrology and climate change. A case study of the Tana River Basin in Northern Fennoscandia. Thesis Utrecht University, Netherlands Geographical Studies 304, 237 pp. Dankers R, Christensen OB, 2004, Climate change impact on snow coverage, evaporation and river discharge in the sub-arctic Tana Basin, Northern Fennoscandia, Climatic Change 69, 367–392. Dankers R, De Jong SM, 2004, Monitoring snow-cover dynamics in Northern Fennoscandia with SPOT VEGETATION images, Int. Journal of Remote Sensing 25, 2933-2949. De Laat PJM, 1992, MUST, a Pseudo Steady-State Approach to Simulating Flow in Unsaturated Media, ICID Bull. CIID 41, 49–60. Diaz-Nieto J, Wilby RL, 2005, A comparison of statistical downscaling and climate change factor methods: impacts on low flows in the river Thames, United Kingdom, Climatic Change 69, 245-268. Gardelin M, Bergström S, Carlsson B, Graham LP Lindström G, 2001, Climate change and water resources in Sweden - Analysis of uncertainties. In: Beniston M (Ed), Climatic Change: Implications for the Hydrological Cycle and for Water Management. Advances in Global Change Research. Kluwer Academic Publishers, Dordrecht. Gellens D, Roulin F, 1998, Streamflow Response of Belgian Catchments to IPCC Climate Change Scenarios, Journal of Hydrology 210, 242–258. Grabs W, Daamen K, Gellens D, Kwadijk JCJ, Lang H, Middelkoop H, Parmet BWAH, Schädler B, Schulla J Wilke K, 1997, Impact of Climate Change on Hydrological Regimes and Water Resources Management in the Rhine Basin, CHR-report I-16, CHR, Lelystad. Graham LP, 1999, Modelling runoff to the Baltic basin. Ambio 28, 328-334.
32 Gurts J, Baltensweiler A, Lang H, Menzel L, & Schulla J, 1997, Auswirkungen von klimatischen Variationen auf Wasserhaushalt und Abfluss im Flussgebiet des Rheins, Schlussbericht NFP 31, Zürich, Hochschulverlag AG. Harding R, Kuhry P, Christensen TR, Sykes MT, Dankers R, Van der Linden S, 2002, Climate feedbacks at the tundra-taiga interface, Ambio Special Report 12, 47-55. Houghton JT, Ding Y, Griggs DJ, Noguer M, Van der Linden PJ, Xiaosu D (eds.), 2001, Climate Change 2001: The Scientific Basis. Cambridge: Cambridge University Press. Hulme M, Conway D, Brown O, Barrow E, 1994, A 1961–1990 Baseline Climatology and Future Climate Change Scenarios for Great Britain and Europe. Part III: Climate Change Scenarios for Great Britain and Europe, Climate Research Unit, University of East Anglia, Norwich. IPCC, 1996, Climate Change 1995; The Science of Climate Change, Houghton JT, Meira Filho LG, Callander BA, Harris N, Kattenberg A, Maskell K (eds.), University Press, Cambridge. Ivanov VV, 1999, Effects of global changes on the hydrological regime, surface water (freshwater) and estuarian ecology of the Barents Sea: Consequences for water resources and environmental threats. In: Lange MA, Bartling B, Grosfeld K (eds.), Global changes and the Barents Sea region. Proceedings of the First International BASIS Research Conference St. Petersburg, Russia. February 22-25, 1998. An ENRICH Conference. Institute for Geophysics, University of Münster, Germany, 63-90. Klemes V, 1986, Operational Testing of Hydrological Simulation Models, Hydrol. Sci. 31, 13–24. Kwadijk JCJ, 1993, The impact of climate change on the discharge of the River Rhine. Utrecht: Koninklijk Nederlands Aardrijkskundig Genootschap / Faculty of Geographical Sciences, University of Utrecht (Nederlandse Geografische Studies 171). Kwadijk JCJ, 1993, The impact of climate change on the discharge of the River Rhine. Thesis Utrecht University, Netherlands Geographical Studies 168, 201 pp. Kwadijk JCJ, Middelkoop H, 1994, Estimation of Climate Change on the Peak Discharge Probability of the River Rhine, Climatic Change 27, 199–224. Kwadijk JCJ, Rotmans J, 1995, The Impact of Climate Change on the River Rhine: A Scenario Study, Climatic Change 30, 397–425. Lange M, The BASIS Consortium 2003. The Barents Sea impact study (BASIS): methodology and first results, Continental Shelf Research 23, 1673-1694. Leavesley GH, 1994, Modeling the effects of climate change on water resources – a review, Climatic Change 28: 159-177. Middelkoop H, Daamen K, Gellens D, Grabs W, Kwadijk JCJ, Lang H, Parmet BWAH, Schädler B, Schulla J Wilke K, 2001, Impact of climate change on hydrological regimes and water resources management in the Rhine Basin, Climatic Change 49, 105-128. Middelkoop H, Kwadijk JCJ, Van Deursen WPA, Van Asselt MBA, 2002, Scenario analyses in global change assessment for water management in the lower Rhine delta. In: Beniston M, (red.), Climatic Change: Implications for the Hydrological Cycle and for Water Management. Advances in Global Change Research 10, Kluwer, Dordrecht, pp. 445-463. Middelkoop H, Van Asselt MBA, Van ’t Klooster SA, Van Deursen WPA, Kwadijk JCJ, Buiteveld H, 2004, Perspectives on flood management in the Rhine and Meuse rivers, River Research and Applications 20, 327-342. Miller JR, Russell GL. 1992. The impact of global warming on river runoff, Journal of Geophysical Research 97/D3: 2757-2764. Mimikou M, Kouvopoulos Y, 1991, Regional climate change impatcs: 1. Impacts on water resources, Hydrological Sciences Journal 36, 247-258. Nash L, Gleick PH, 1991, ‘Sensitivity of streamflow in the Colorado Basin to climate changes’, Journal of Hydrology 125, 221-241. Nemec J, Schaake JC, 1982, Sensitivity of water resource systems to climate variations. Hydrological Sciences Journal 27, 327-343. Oberman NG, Mazhitova GG, 2001, Permafrost dynamics in the north-east of European Russia at the end of the 20th century, Norwegian Journal of Geography 55: 241-244. Peterson BJ, Holmes RM, McClelland JW, Vörösmarty CJ, Lammers RB, Shiklomanov AI, Shiklomanov IA, Rahmstorf S, 2002, Increasing river discharge to the Arctic Ocean, Science 298, 2171-2173. Roeckner E. Arpe K, Bengtsson L, Christoph M, Claussen M, Dümenil L, Esch M, Giorgetta M, Schlese U, Schulzweida U, 1996, The atmospheric general circulation model ECHAM-4: Model description and simulation of present-day climate, Hamburg: Max-Planck-Institute (MPI Report 218). Shabalova MV, Van Deursen WPA, Buishand TA, 2003, Assessing future discharge of the river Rhine
33 using regional climate model integrations and a hydrological model, Climate Research 23: 233-246. Shiklomanov IA, Shiklomanov AI, Lammers RB, Peterson BJ, Vörösmarty CJ, 2000, The dynamics of river water inflow to the Arctic Ocean. In: Lewis EL et al. (eds.) The freshwater budget of the Arctic Ocean. NATO Science Series 2. Environmental Security – Vol. 70, Kluwer Acad. Publ., 281-296. Stehlik J, Bardossy A, 2002, Multivariate stochastic downscaling model for generating daily precipitation series based on atmospheric circulation, Journal of Hydrology 256 (1-2): 120-141. Van Blarcum SC, Miller JR, Russell GL, 1995, High latitude runoff in a doubled CO2 climate. Climatic Change 30, 7-26. Van der Linden S, 2002, Ice rivers heating up. Modelling hydrological impacts of climate change in the (sub)arctic. Thesis Utrecht University, Netherlands Geographical Studies 308, 174 pp. Van der Linden S, Christensen JH, 2003, Improved hydrological modeling for remote regions using a combination of observed and simulated precipitation data, Journal of Geophysical Research 108/D2, 4072-4082. Van der Linden S, Virtanen T, Oberman N, Kuhry P. 2003. Sensitivity analysis of discharge in the Arctic Usa basin, East-European Russia, Climatic Change 57, 139-161. Van der Linden S, Woo M-k, 2003, Application of hydrological models with increasing complexity to subarctic catchments, Journal of Hydrology 270, 145-157. Van Deursen WPA, Kwadijk JCJ, 1993, Rhineflow: An Integrated GIS Water Balance Model for the River Rhine, In Kovar K, Nachtnebel HP (eds.), Application of Geographic Information Systems in Hydrology and Water Resources Management, Proceedings of the Conference on HYDROGIS, Vienna, Austria, IHAS publication 211, pp. 507–519. Vehviläinen B, Lohvansuu J, 1991,. The effects of climate change on discharges and snow cover in Finland, Hydrological Sciences Journal 36, 109-121. Warmerdam PMM., Kole JW, Stricker JNM, 1993, Rainfall-Runoff Modelling in the Research Area of the Hupselse Beek, The Netherlands, Proceedings of the International Symposium on Runoff and Sediment Yield Modelling, Warsaw, 1993. Ward RC, Robinson M, 1990, Principles of Hydrology, Third edition, McGraw-Hill, London. Wilby RL, Wigley TML, Conway D, Jones PD, Hewitson BC, Main J, Wilks DS, 1998, Statistical downscaling of General Circulation Model output: A comparison of methods, Water Resources Research 34, 2995-3008. Wolock DM, McCabe GJ, Tasker GD, Moss ME, 1993, Effects of climate change on water resources in the Delaware River basin. Water Resources Bulletin 29, 475-486.
34