HYDROLOGICAL PROCESSES
Hydrol. Process. 20, 2091– 2109 (2006)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hyp.6197
Assessment of climate-change impacts on alpine discharge
regimes with climate model uncertainty
Pascal Horton, Bettina Schaefli, Abdelkader Mezghani, Benoı̂t Hingray* and André Musy
Ecole Polytechnique Fédérale de Lausanne, Hydrology and Land Improvement Laboratory, CH-1015 Lausanne, Switzerland
Abstract:
This study analyses the uncertainty induced by the use of different state-of-the-art climate models on the prediction
of climate-change impacts on the runoff regimes of 11 mountainous catchments in the Swiss Alps having current
proportions of glacier cover between 0 and 50%. The climate-change scenarios analysed are the result of 19 regional
climate model (RCM) runs obtained for the period 2070–2099 based on two different greenhouse-gas emission
scenarios (the A2 and B2 scenarios defined by the Intergovernmental Panel on Climate Change) and on three
different coupled atmosphere-ocean general circulation models (AOGCMs), namely HadCM3, ECHAM4/OPYC3 and
ARPEGE/OPA. The hydrological response of the study catchments to the climate scenarios is simulated through
a conceptual reservoir-based precipitation-runoff transformation model called GSM-SOCONT. For the glacierized
catchments, the glacier surface corresponding to these future scenarios is updated through a conceptual glacier surface
evolution model.
The results obtained show that all climate-change scenarios induce, in all catchments, an earlier start of the snowmelt
period, leading to a shift of the hydrological regimes and of the maximum monthly discharges. The mean annual runoff
decreases significantly in most cases. For the glacierized catchments, the simulated regime modifications are mainly
due to an increase of the mean temperature and the corresponding impacts on the snow accumulation and melting
processes. The hydrological regime of the catchments located at lower altitudes is more strongly affected by the changes
of the seasonal precipitation. For a given emission scenario, the simulated regime modifications of all catchments are
highly variable for the different RCM runs. This variability is induced by the driving AOGCM, but also in large part
by the inter-RCM variability. The differences between the different RCM runs are so important that the predicted
climate-change impacts for the two emission scenarios A2 and B2 are overlapping. Copyright  2006 John Wiley &
Sons, Ltd.
KEY WORDS
climate change; regional climate models; hydrological modelling; snowmelt modelling; glacier hydrology;
Alps
INTRODUCTION
The Alps are a key element of the hydrological regime of the Swiss river network and they are the source of
many of Europe’s major river systems. The specific mean annual discharge in mountainous areas is generally
higher than in catchments located at lower altitudes in the same climatic region. This results essentially
from higher precipitation amounts induced by orographic effects, but also from low evapotranspiration rates
induced in large part by the relatively low mean temperatures. Additionally, the hydrological regime of such
environments is strongly influenced by water accumulation in the form of snow and ice and the corresponding
melt processes, resulting in a pronounced annual cycle of the discharge. Therefore, a modification of the
prevalent climate, and especially of the temperature, can considerably affect the hydrological regime and
induce important impacts on the water management (e.g. Burlando et al., 2002; Jasper et al., 2004; Schaefli,
2005).
* Correspondence to: Benoı̂t Hingray, Hydram-ISTE, Station 2, EPFL, 1015 Lausanne, Switzerland. E-mail: [email protected]
Copyright  2006 John Wiley & Sons, Ltd.
Received 18 May 2005
Accepted 17 November 2005
2092
P. HORTON ET AL.
The quantification of potential climate-change impacts on a water resources system is conditioned by the
availability of local or regional climate-change predictions of different key meteorological variables, such as
precipitation and temperature. Currently available climate-change projections are mainly based on the results
of coupled atmosphere–ocean general circulation models (AOGCMs) or on the results of regional climate
models (RCMs) driven by outputs of the former. RCMs have a higher spatial resolution than AOGCMs and
are usually expected to describe the regional-scale climate better.
Regional climate-change projections based on such climate model outputs are, however, highly uncertain,
mainly due to the unknown future greenhouse-gas emissions, but also due to the highly simplified representation of reality encoded in these models. As a consequence, different AOGCM experiments, or different RCM
experiments driven by the same AOGCM outputs, generally yield different climate evolutions for the same
emission scenario (e.g. Arnell and Hulme, 2000; Frei et al., 2003; Räisänen et al., 2004). The uncertainty
introduced by the RCM is generally considered to be substantially smaller than the one inherited by the driving
AOGCM (Jenkins and Lowe, 2003). An analysis of the data of the EU project PRUDENCE (Prediction of
regional scenarios and uncertainties for defining European climate change risks and effects; Christensen et al.,
2002) suggests, however, that RCM inter-model variability cannot be neglected (Hingray et al., 2006a).
Different studies have assessed climate-change impacts on hydrological processes in the Swiss Alps (e.g.
Braun et al., 2000; Etchevers et al., 2002; Jasper et al., 2004; Zierl and Bugmann, 2005). All of these studies
show that the temperature increase strongly affects the temporal evolution of the snowpack and, accordingly,
the runoff regimes of the catchments studied. Few studies have analysed different greenhouse-gas emission
scenarios or have taken into account the prediction uncertainty due to the climate model used. For alpine
catchments without glaciers, Jasper et al. (2004) and Zierl and Bugmann (2005) show, however, that these
two sources of climate-change prediction uncertainty lead to a wide range of possible future states.
This study focuses on the assessment of climate-change impacts on the discharge regime of 11 catchments
in the Swiss Alps, considering the climate predictions of multiple state-of-the-art regional climate models.
Some of the catchments selected have glacier-covered areas, decreases of which are expected to enhance
the climate-change-induced modifications of the hydrological regime (Braun et al., 2000; Schaefli, 2005).
For each catchment, the climate-change impact analysis is conducted through the simulation for an observed
control period (1961–1990) and for a future period (2070–2099) having a variety of modified (predicted)
climates. These climate predictions were derived from a suite of RCM experiments undertaken within the
framework of the EU PRUDENCE project (Christensen et al., 2002), for the two greenhouse-gas emission
scenarios A2 and B2, as defined by the Special Report on Emission Scenarios (SRES; Nakićenović and Swart,
2000), of the Intergovernmental Panel on Climate Change.
A description of the case studies, the corresponding datasets, and the models used, both for the discharge
simulation and for the production of small-scale meteorological time-series scenarios, is presented next. Then
the climate-change impact prediction results are discussed, followed by the main conclusions of this study.
CASE STUDIES: DATA AND MODELS
Catchment characteristics
The catchments analysed were selected to represent the seven different hydro-climatic zones of the Swiss
Alps (Laternser and Schneebeli, 2002) and the different mountainous hydrological regimes occurring in these
zones (Figure 1). The choice of catchments was also conditioned by data availability. Coincident series of
observed discharge and meteorological data are necessary for model calibration and validation, and observed
meteorological data are required for the simulation of the 30-year reference (control) period (1961–1990).
The catchments selected have different proportions of glacier cover (between 0 and 50%) and altitude ranges
(Table I). Their total areas vary between 39 and 185 km2 , and their mean altitudes vary from 1340 m a.s.l.,
for the Minster catchment, to 2940 m a.s.l., for the Drance de Bagnes catchment. The catchments represent
a large range of reference hydrological regimes, as defined by Aschwanden and Weingartner (1985) for the
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2093
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
Figure 1. Location of the 11 case study catchments in the seven hydro-climatic regions of the Swiss Alps (SwissTopo, 1997)
Table I. Characteristics of the case study catchments and corresponding discharge regimes according to the regime definitions
given by Aschwanden and Weingartner (1985) for the Swiss Alps (proportion of glacier cover, annual precipitation and
hydrological regime correspond to the control period)
Catchment
Area
(km2 )
Glacier
(%)
Mean
altitude
(m a.s.l.)
Altitude
range
(m a.s.l.)
Lapse
rate
(° C/100 m)
Annual
precipitation
(mm)
Hydrological
regimea
Drance de Bagnes
Saaser Vispa
Lonza
Rhône at Gletsch
Weisse Lütschine
Minster
Tamina
Vorderrhein
Dischmabach
Rosegbach
Verzasca
166Ð6
65Ð2
77Ð8
38Ð9
164Ð0
59Ð2
147Ð0
158Ð0
43Ð3
66Ð5
185Ð2
39Ð0
33Ð2
32Ð8
50Ð0
16Ð4
—
1Ð4
2Ð8
2Ð6
27Ð2
—
2940
2840
2600
2710
2150
1340
1810
2150
2360
2710
1650
1960–4310
1710–4190
1520–3900
1760–3610
650–4170
880–2290
550–3220
750–3310
1670–3130
1760–4010
480–2880
0Ð50
0Ð46
0Ð57
0Ð55
0Ð60
0Ð54
0Ð61
0Ð58
0Ð56
0Ð43
0Ð57
1620
1627
2187
2262
1774
2197
1554
1748
1460
1412
2175
‘a-glacial’
‘b-glacial’
‘a-glacial’
‘a-glacial’
‘a-glacio-nival’
‘nival de transition’
‘nival alpin’
‘nivo-glacial’
‘b-glacio-nival’
‘a-glacial’
‘nivo-pluvial meridional’
a Aschwanden
and Weingartner (1985).
Swiss Alps (Table I). These were determined on the basis of dimensionless Pardé coefficients (Pardé, 1933),
which relate the mean monthly discharge of a given month to the total annual discharge.
All catchments selected (except the Verzasca catchment) show typical discharge patterns characterized by
a single peak occurring between May and August. This peak is due to snowmelt-induced high flows. For
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2094
P. HORTON ET AL.
catchments with glaciers, the peak is sustained by usually significant ice-melt-induced flows occurring later in
the melt season. The amplitude of the monthly flows is high between the low flows during the winter season
and the high flows in late spring or summer. The amplitude, as well as the occurrence date of the discharge
peak, depends strongly on the elevation range covered by the catchment (Aschwanden and Weingartner, 1985).
The higher the mean elevation of the catchment is, the later the peak occurs and the higher the amplitude is (see
the regimes simulated for the control period, Figure 3). Meanwhile, the Verzasca River has a typical regime
of the southern Alps, with a second discharge peak occurring in the autumn, due to the heavy precipitation
events that are usually observed in this period.
System models
The simulation of catchment behaviour for the different time periods was carried out using a hydrological
model, for the precipitation–runoff transformation, and a land-cover evolution model. The only land-cover
change that is explicitly accounted for is the modification of the glacier-covered surface. Land-use changes
related to the vegetation could presumably have a significant impact on future hydrological regimes, but
they are not considered in this study. Zierl and Bugmann (2005), however, found that, for five Swiss alpine
catchments, such land-use changes only have a small impact on discharges under similar climate-change
scenarios to the ones used here.
Hydrological model. The hydrological discharge simulation is carried out at a daily time-step through a
conceptual reservoir-based model called GSM-SOCONT (Schaefli et al., 2005). The model has two levels
of discretization, with the ice-covered part of the catchment first separated from the not ice-covered part,
and then both parts subdivided into 10 elevation bands of same surface. Each of the resulting spatial units
is characterized by its surface and its mean elevation, and is assumed to have a homogeneous hydrological
behaviour.
For each spatial unit, the meteorological data series are computed from data observed at neighbouring
meteorological stations. For temperature, regional altitudinal gradients are estimated based on the observed
temperature series (Table I). Based on these gradients, the temperature time-series for a given spatial unit is
interpolated according to its mean elevation. To account for the enhancement of precipitation with altitude, a
mesoscale annual precipitation gradient of 80 mm per 100 m, estimated by Kirchhofer and Sevruk (1991) for
the entire Swiss alpine region, is applied. The resulting area-average daily precipitation amounts are corrected
by a constant multiplicative factor in order to respect the mean annual precipitation amounts estimated for
each catchment by Schädler and Bigler (2002) based on a long-term water balance analysis.
Based on the meteorological time-series, the temporal evolutions of the snowpack and of the glacier melt
are computed. The aggregation state of precipitation (liquid, solid or mixed) is determined through a fuzzy
temperature threshold approach (Klok et al., 2001). Snow and ice melt are simulated through a simple degreeday approach (Rango and Martinec, 1995). Ice melt occurs only when the glacier surface of the spatial unit
being considered is free of snow.
For each hydrological unit, a reservoir-based modelling approach is used to simulate the rainfall and meltwater–runoff transformation. For ice-covered spatial units, this transformation is completed through two linear
reservoirs, one for ice melt and one for the equivalent rainfall, defined as the sum of snowmelt and rainfall.
For spatial units that have no ice, the equivalent rainfall is transformed into runoff through the conceptual
model SOCONT (Consuegra and Vez, 1996). It is composed of two reservoirs, i.e. a linear reservoir for the
slow contribution (accounting for soil infiltration processes) and a non-linear reservoir for direct, or quick,
runoff. Actual evapotranspiration is estimated as a function of the potential evapotranspiration (PET) and the
filling rate of the slow component reservoir.
The model has seven parameters to calibrate (four for catchments without a glacier): two degree-day factors
for the ice and snowmelt computation, four time constants of the reservoirs, and the maximum storage capacity
of the slow reservoir.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
2095
The calibration procedure is based on the multicriteria calibration procedure presented by Schaefli et al.
(2005). It aims to minimize the Nash–Sutcliffe model efficiency (NSE) coefficient (Equation (1); Nash and
Sutcliffe, 1970) and the bias, estimated between observed and simulated daily discharges. Special attention is
also paid to the reproduction of the observed mean annual and the mean monthly discharges over a calibration
period. Calibration and validation were undertaken for two consecutive 10-year periods (1971–1980 and
1981–1990 respectively, for the majority of the catchments).
The calibrated model performs well for all catchments for both the calibration and the validation period.
NSE coefficients of 0Ð85 or higher were obtained for the modelled daily discharge for most catchments. Only
the two ice-free catchments of Minster and Verzasca had lower NSE values, both being close to 0Ð75. For
both calibration and validation periods, NSE values for monthly discharges exceeded 0Ð9 for all catchments.
n
NSE D 1 Qo,i Qs,i 2
iD1
n
1
2
Qo,i Qo iD1
where Qo,i and Qs,i are the observed and simulated discharges of the ith time step, Qo is the mean observed
discharge over the calibration or validation period, and n is the number of time steps simulated. An NSE
coefficient value of unity indicates a perfect model.
Glacier surface evolution model. The glacier area was considered to be constant during the control simulation
period. For this period, the glacier area was obtained from vector-based topographic maps (SwissTopo, 1997).
It corresponds approximately to the middle of the control period, the exact year depending on the date of the
field surveys. The glacier area was also assumed to be constant during the future simulation period, but using
a different (simulated) constant value that accounts for the glacier retreat.
The response of a glacier system to a climate modification could be simulated through a dynamical ice-flow
model (e.g. Oerlemans et al., 1998). Such a physical modelling approach is, however, highly data intense
and can only be applied to well-investigated glacier systems. In this study, the glacier surface for future time
periods is updated through a conceptual modelling approach based on the so-called accumulation area ratio
(AAR) method.
The AAR is defined as the ratio between the accumulation area of the glacier and its entire surface
(Anonymous, 1969). The AAR method assumes that the steady-state accumulation area of a glacier occupies
some fixed proportion of the total glacier area (e.g. Meier and Post, 1962; Gross et al., 1976). The concept is
classically used in the palaeoclimatic reconstruction of glacier surfaces (e.g. Porter, 1975; Benn and Lehmkuhl,
2000). For a given glacier, these studies assume the existence of a constant steady-state AAR value that can
be estimated and used to predict the steady-state glacier area for any period according to
Aice,s D
Aacc,s
AARs
2
where Aice,s km2 is the total steady-state glacier area, AARs is the steady-state AAR and Aacc,s km2 is the
estimated steady-state accumulation area for the climatic conditions of the period considered.
A glacier is defined as being in a steady state if its dimensions remain constant over the period considered.
This is a theoretical concept that is rarely or never encountered in practice (Paterson, 1994) and the steadystate AARs value is unknown a priori. In this study, the mean annual AAR value AARm was used, on the
assumption it is a good estimator of the relationship between the climate and the glacier area if estimated over
a long enough time period. It was further assumed that this relationship remains constant for future periods.
In other words, the mean annual AAR value for the present situation was assumed to remain constant for the
future period.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2096
P. HORTON ET AL.
The annual AAR value can be estimated based on the hydrological model outputs for the control period:
for a given hydrological year (starting on the 1 October), the AAR is computed as the sum of spatial units
that experience net snow accumulation divided by the total (known) glacier surface.
Based on AARm,cont , the estimated AARm value for the 1961–1990 control period, the future glacier surface
Aice,fut km2 can be simulated according to
Aice,fut D
Aacc,fut
AARm,cont
3
where Aacc,fut km2 is the mean annual accumulation area of the future period as simulated by the hydrological
model under future climate conditions (the annual accumulation area is estimated as the sum of spatial units
that experience net snow accumulation for a given hydrological year under the future climate conditions). For
a further discussion of the glacier surface evolution model, refer to Schaefli (2005).
Data
Observed periods. The glacio-hydrological model needs three input time-series, namely daily precipitation,
daily PET and mean daily temperature. For all catchments, the precipitation and temperature data are obtained
from nearby meteorological stations of the Swiss Meteorological Institute. The PET data are derived from
the monthly PET series estimated by New et al. (2000) for the 1961–1990 control period, according to the
Penman–Monteith version given by Burman and Pochop (1994). The series of mean daily discharges for
model calibration and validation are provided by the Swiss Federal Office for Water and Geology.
The spatial discretization of the catchment is completed based on a 25 m resolution digital elevation model
(SwissTopo, 1995) and on 1 : 25 000 digital, vector-based, topographic maps (SwissTopo, 1997).
Future periods. Each RCM experiment consists of a simulation for the period 1961–1990 (control run)
and a simulation for the period 2070–2099 (future run). For each RCM experiment, the boundary conditions
were obtained from one of the three AOGCMs used in PRUDENCE (Christensen et al., 2002): ARPEGE/OPA
(Royer et al., 2002), HadCM3 (Gordon et al., 2000; Pope et al., 2000) and ECHAM4/OPYC3 (Roeckner et al.,
1999). In the case of HadCM3, a global model of the atmosphere alone (HadAM3H) was used between the
global coupled model and the RCMs. The use of this intermediate model resulted in a much better simulation
of the present-day climate (Hulme et al., 2002). The AOGCM experiments were completed for the two SRES
emission scenarios A2 and B2 (Nakićenović and Swart, 2000). The A2 scenario assumes a heterogeneous
world where regional cultural identities are strengthened, with a high population growth but with less concern
for rapid economic development (Nakićenović, 2000). The B2 scenario also assumes a heterogeneous world,
but with local solutions to enhance economic, social and environmental sustainability. Compared with A2,
the technological progress in B2 is less rapid, and the greenhouse-gas emissions are expected to be smaller
(Nakićenović, 2000). Hence, a smaller global temperature increase is associated with B2 than A2.
In this study, nine RCMs are included (Table II). One of these, ARPEGE, is not a regional model; rather, it is
a global atmospheric model with variable horizontal resolution, from 50 km in the centre of the Mediterranean
Sea to 450 km in the southern Pacific Ocean (Gibelin and Déqué, 2003). In PRUDENCE, the boundary
conditions for this model (sea-surface temperatures, sea ice) were obtained from ARPEGE/OPA and HadCM3
(Déqué, personal communication). For the 11 catchments analysed in this study, the horizontal resolution of
ARPEGE is comparable to the other RCMs and, for the purposes of the study, ARPEGE was considered as
an RCM. Outputs from 19 RCM experiments are available in total: 12 for the A2 scenario and seven for B2.
A synthesis of these experiments is given in Table III.
The RCMs are reported to reproduce well the regional features of meteorological surface variables such
as precipitation and temperature (e.g. Frei et al., 2003). Their reliability, however, is not the same for each
individual RCM grid cell, particularly in mountainous regions. As a result, instead of considering only the
grid cell containing a single given case study, it was decided to use data from several grid cells. In this study,
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2097
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
Table II. The three AOGCMs and the nine RCMs used in the PRUDENCE project (Christensen et al., 2002) and the
corresponding institutions
Institution
AOGCM
ARPEGE/OPA
HadCM3
ECHAM4/OPYC3
RCM
ARPEGE
HIRHAM
CHRM
CLM
HadRM3H
RegCM
REMO
RCAO
PROMES
Reference
Centre National de Recherches Météorologiques, Toulouse, France
Hadley Centre for Climate Prediction and Research, Bracknell, UK
Max-Planck-Institut für Meteorologie, Hamburg, Germany
Royer et al. (2002)
Gordon et al. (2000), Pope
et al. (2000)
Roeckner et al. (1999)
Centre National de Recherches Météorologiques, Toulouse, France
Danish Meteorological Institute, Copenhagen, Denmark
Institute for Atmospheric and Climate Science, Zurich, Switzerland
Institute for Coastal Research, Geesthacht, Germany
Hadley Centre for Climate Prediction and Research, UK
International Centre for Theoretical Physics, Trieste, Italy
Max-Planck-Institut für Meteorologie, Hamburg, Germany
Swedish Meteorological and Hydrological Institute, Norrköping, Sweden
Universidad Complutense de Madrid, Toledo, Spain
Gibelin and Déqué (2003)
Christensen et al. (2001)
Vidale et al. (2003)
Doms and Schättler (1999)
Hulme et al. (2002)
Giorgi et al. (1993a,b)
Jacob (2001)
Räisänen et al. (2004)
Arribas et al. (2003)
Table III. RCM experiments according to the driving AOGCM and to the
emission scenarios (all RCM experiments were conducted in the framework
of PRUDENCE)
Scenario name
A2
B2
AOGCM name
RCM name
No.
HADCM3
HADCM3
HADCM3
HADCM3
HADCM3
HADCM3
HADCM3
HADCM3
HADCM3
ARPEGE/OPA
ECHAM4/OPYC3
ECHAM4/OPYC3
HADCM3
HADCM3
HADCM3
HADCM3
ARPEGE/OPA
ECHAM4/OPYC3
ECHAM4/OPYC3
CHRM
CLM
HadRM3H
HIRHAM
PROMES
RCAO(H)
RegCM
REMO
ARPEGE
ARPEGE
HIRHAM
RCAO(E)
HadRM3H
PROMES
RCAO(H)
ARPEGE
ARPEGE
HIRHAM
RCAO(E)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
for each catchment analysed, the regional changes predicted by a given RCM are thus averaged over nine
grid cells encompassing the catchment of interest. The uncertainty in the regional climate-change scenarios
associated with the number of aggregation grids (9, 16 or 25) has been estimated for the Rhône catchment at
Gletsch. It is much smaller than the uncertainty resulting from the inter-model differences and, therefore, is
neglected.
The local-scale time-series of temperature and precipitation for a given future scenario are obtained by
perturbing the observed time-series for a control period, based on the regional climate changes predicted by
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2098
P. HORTON ET AL.
the climate models. The method used for temperature perturbation is the one presented by Shabalova et al.
(2003). It preserves the changes in mean and variability given by the climate-change scenario for each season.
For a given RCM experiment, the climate-change scenario is defined in terms of the absolute changes of the
mean temperature XMTs between the control and the future runs and in terms of the relative changes of the
standard deviation of daily temperature XSDTs (defined as the difference between the values obtained from
the RCM output for the future period and the control period, divided by the value for the control period),
where s refers to the season (s D 1: December–February (DJF); s D 2: March–May; s D 3: June–August
(JJA); s D 4: September–November). The future scenario temperature Tscen is calculated according to
Tscen,s t D [Tobs,s t MTobs,s ]XSDTs C 1 C MTobs,s C XMTs
4
where Tscen,s t (° C) is the local-scale scenario temperature on day t in the season s, Tobs,s t is the local
temperature on day t observed during the control period in the season s, and MTobs,s is the corresponding
mean daily temperature.
Note that the standard deviations of daily temperatures SDT are not available from PRUDENCE experiments. Instead, PRUDENCE has produced grids of 90-day temperature variances for each season. The change
in the daily standard deviations was thus set equal to the change in the 90-day standard deviations produced
in PRUDENCE. It can be easily shown that, in this case, the perturbation of the observed daily series based
on Equation (4) gives a future daily series that preserves the relative change in the 90-day standard deviations. Note also that these 90-day statistics have been estimated in PRUDENCE directly from the RCM
output series. A significant temperature trend has been simulated for the 30-year period of 2070–2099 that
could potentially influence the variance estimation. PRUDENCE data were thus preprocessed according to
the method presented in Hingray et al. (2006b).
For precipitation, the use of a comparable perturbation method is not possible because this would generate
negative rainfall values. The future local-scale precipitation time-series are estimated based on the simple
proportional relationship
MPfut,s
Pscen,s t D Pobs,s t
5
MPcont,s
where Pscen,s t (mm) is the local-scale scenario precipitation on day t of the season s (s D 1–4), Pobs,s t is
the precipitation observed on day t during the control period, and MPfut,s and MPcont,s are respectively the
mean seasonal precipitations for the future and the control run for a given RCM experiment.
Note that, for the future scenarios, the PET is estimated as a function of the perturbed temperature based on
the observed linear relationship for the control period, assuming that this relationship remains constant in the
future. In fact, Ekström et al. (2006) have shown that PET values calculated from the outputs of HadRM3H
using a Pennman–Monteith method reach unrealistically high values for both the control and the future period.
RESULTS
Climate changes
The ARPEGE/OPA, HadCM3 and ECHAM4/OPYC3 AOGCMs predict global-mean warmings for scenario
B2 of C2Ð4 ° C, C2Ð4 ° C and C2Ð8 ° C respectively, and of C3Ð0 ° C, C3Ð3 ° C and C3Ð6 ° C for scenario A2
(Roeckner et al., 1999; Gordon et al., 2000; Gibelin and Déqué, 2003). The regional mean annual temperature
increases predicted by the 19 PRUDENCE RCM experiments are similar for all catchments and are higher
than the global-mean warming (Table IV), at around C3Ð0 ° C for B2 and C4Ð0 ° C for A2. This enhanced
regional warming is consistent with the temperature increases that have already been observed in the Swiss
Alps over the 20th century (e.g. Beniston et al., 1994; Weber et al., 1997). The predicted regional warming
for the summer is higher than for the other seasons (e.g. the maximum summer warming predicted for A2
for the Rhône catchment is as large as C8Ð1 ° C, being around C5Ð2 to C6 ° C for the other seasons).
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
Copyright  2006 John Wiley & Sons, Ltd.
Drance de Bagnes
Saaser Vispa
Lonza
Rhône at Gletsch
Weisse Lütschine
Minster
Tamina
Vorderrhein
Dischmabach
Rosegbach
Verzasca
Median
Catchment
A2 scenario
B2 scenario
A2 scenario
Annual change in precipitation (%)
B2 scenario
DJF
A2 scenario
B2 scenario
JJA
Seasonal change in precipitation (%)
A2 scenario
2Ð5
2Ð5
2Ð5
2Ð4
2Ð5
2Ð4
2Ð4
2Ð4
2Ð4
2Ð4
2Ð4
2Ð4
3Ð0
3Ð0
3Ð0
3Ð0
3Ð1
3Ð1
3Ð0
3Ð0
2Ð9
3Ð0
3Ð1
3Ð0
4Ð7
4Ð6
4Ð6
4Ð6
4Ð7
4Ð7
4Ð6
4Ð6
4Ð7
4Ð6
4Ð5
4Ð6
3Ð3
3Ð3
3Ð3
3Ð2
3Ð2
3Ð2
3Ð2
3Ð2
3Ð2
3Ð2
3Ð2
3Ð2
4Ð2
4Ð0
4Ð0
3Ð9
3Ð9
3Ð8
3Ð9
3Ð9
4Ð0
4Ð0
4Ð0
4Ð0
6Ð1
6Ð1
6Ð1
6Ð2
6Ð2
6Ð3
6Ð2
6Ð1
6Ð2
6Ð1
6Ð1
6Ð1
11
12
9
10
13
13
13
12
13
15
16
13
6
6
5
4
6
5
8
5
9
8
8
6
0
0
1
0
1
0
1
0
2
2
1
0
21 8
22 8
19 6
20 6
25 7
25 5
24 7
22 6
24 7
26 11
26 11
24 7
0
1
2
2
3
2
1
0
0
2
2
1
7
3
3
1
1
0
2
0
3
8
3
3
10
12
12
15
15
16
15
15
15
17
13
15
23
23
23
20
18
22
20
21
21
24
29
22
0
1
1
6
5
4
1
4
1
11
6
1
14
15
15
16
14
14
21
19
22
27
24
16
28
28
28
28
26
24
27
28
29
37
30
28
46
43
43
39
39
37
39
40
37
42
46
40
23
26
26
32
31
28
31
32
26
30
34
30
3
2
2
1
2
0
1
0
3
0
1
1
52
48
48
51
49
55
50
52
45
50
54
50
27
27
27
25
27
22
22
23
21
23
26
25
10
9
9
9
12
8
6
7
3
5
6
8
Min. Med. Max. Min Med Max Min. Med. Max. Min. Med. Max. Min. Med. Max. Min. Med. Max. Min. Med. Max. Min. Med. Max.
B2 scenario
Annual change in temperature (° C)
Table IV. Minimum, median and maximum regional climate changes predicted by the PRUDENCE RCM experiments for 2070–2099 for B2 scenario (seven
experiments) and A2 scenario (12 experiments): a) annual changes of mean temperature and precipitation, b) seasonal changes of mean precipitation amounts
for winter (DJF) and summer (JJA)
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
2099
Hydrol. Process. 20, 2091– 2109 (2006)
2100
10
Winter (DJF)
9
8
7
6
5
4
3
2
1
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
XMT [°C]
XMT [°C]
P. HORTON ET AL.
10
Summer (JJA)
9
8
7
6
5
4
3
2
1
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Global ∆T [°C]
Global ∆T [°C]
60
60
Winter (DJF)
40
20
XMP [%]
XMP [%]
40
0
Summer (JJA)
20
0
−20
−20
−40
−40
−60
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
−60
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Global ∆T [°C]
Global ∆T [°C]
HADCM3-A2
ARPEGE/OPA-A2
ECHAM4/OPYC3-A2
HADCM3-B2
ARPEGE/OPA-B2
ECHAM4/OPYC3-B2
Figure 2. Regional seasonal changes of precipitation (XMP) and temperature (XMT) as a function of the annual global mean warming T
as predicted by the PRUDENCE RCM experiments for the Rhône catchment for the period 2070– 2099. Top: regional temperature changes;
bottom: regional precipitation changes. Each symbol corresponds to one RCM experiment; large symbols identify the RCAO experiments
There is a clear gradation of the projected regional warming between the B2 scenario and the A2 scenario:
every percentile of the estimated temperature-change distribution based on the A2 RCM experiments has
a higher value than that estimated for B2. The spread of regional changes obtained from the different
AOGCM/RCM experiments for a given emission scenario is, however, considerable (Table IV and Figure 2).
This variability is induced by the driving AOGCMs and by the RCMs. Considering a given RCM, the driving
AOGCM induces a variability of the same order of magnitude as the one induced by the emission scenario
(Figure 2). Considering a given AOGCM, the inter-RCM variability represents an important additional source
of prediction uncertainty. Whereas for each AOGCM the predicted global-mean warming is clearly distinct
for the B2 and A2 scenarios, the ranges of regional changes predicted by the different RCMs for a given
AOGCM are often overlapping.
Except for a few experiments and catchments, a decrease of the annual precipitation is predicted (Table IV);
there is no clear gradation between the projected changes of the annual precipitation for B2 and A2. Most
RCMs predict an increase of winter precipitation and a decrease of summer precipitation for both scenarios
and for all catchments (Table IV and Figure 2). For spring, the predictions of the different RCMs do not
have a clear tendency towards an increase or a decrease (for all catchments and both scenarios). For autumn,
precipitation is predicted to decrease for the A2 scenario, but no tendency is detectable for B2. The different
levels of prediction variability highlighted for temperature changes are also observed for precipitation changes
(Figure 2).
Glacier surface decrease
The predicted decrease of glacier surfaces between the control and the future period is considerable
(Table V), with almost no ice-covered areas left in the future. For the highest catchment, the Drance catchment,
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2101
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
Table V. Predicted change of the annual runoff and predicted glaciation rate for the B2 and A2 scenarios (2070–2099): the
minimum, median and maximum values of the 7 and 12 simulations based on the PRUDENCE RCM experiments available
for the scenario B2 and A2
Catchment
Discharge modification (%)
B2 scenario
Drance de Bagnes
Saaser Vispa
Lonza
Rhône at Gletsch
Weisse Lütschine
Minster
Tamina
Vorderrhein
Dischmabach
Rosegbach
Verzasca
Median
Glaciation rate (% of total catchment area)
A2 scenario
B2 scenario
Min.
Med.
Max.
Min.
Med.
Max.
36
21
19
28
33
26
24
22
24
28
27
26
26
13
14
20
23
14
16
12
17
18
16
16
19
5
7
15
12
6
5
6
4
6
3
6
49
34
32
40
47
42
39
35
39
41
40
40
30
17
16
23
25
16
16
14
16
22
19
17
24
9
8
16
20
11
10
8
10
13
10
10
A2 scenario
Min.
Med.
Max.
Min.
Med.
Max.
0.8
0.6
1.0
0.0
0.5
—
0.0
0.0
0.0
0.0
—
—
1.8
3.7
2.6
0.3
2.2
—
0.0
0.0
0.0
1.4
—
—
6.7
7.1
5.1
6.7
6.2
—
0.1
0.0
0.0
2.3
—
—
0.1
0.0
0.1
0.0
0.1
—
0.0
0.0
0.0
0.0
—
—
1.0
1.4
1.6
0.0
0.9
—
0.0
0.0
0.0
0.3
—
—
1.7
3.4
2.3
0.0
2.6
—
0.0
0.0
0.0
1.3
—
—
the median future predicted glacier-covered area corresponds to 2% of the total catchment area for B2 (1%
for A2), compared with the 39% glacier cover during the control period. This result is primarily due to the
large predicted temperature increases. The periods during which precipitation falls in the form of snow are
also predicted to be shorter in the future. Despite the expected additional winter precipitation, this leads to less
snow accumulation on the glacier surface. Owing to the predicted high summer temperatures, the accumulated
winter snow disappears during the following melt season at much higher altitudes than for the control period.
The accumulation area of glaciers, thus, decreases considerably, which leads to significant glacier retreat. For
the Rhône catchment, the predicted median increase of mean temperature of C3Ð9 ° C for A2 corresponds to
an upward shift of around 710 m of the 0 ° C isotherm. This shift cannot be directly translated into a reduction
of the glacier surface because there is no linear relationship between the glacier surface and the 0 ° C isotherm
elevation. Nevertheless, it is worth noting that the mean altitude of the glacierized area for the control period
is around 2940 m a.s.l. and the highest point of the glacier is only about 670 m higher, at 3610 m a.s.l. The
simulated complete disappearance of the Rhône glacier is in line with the results obtained by Wallinga and
van de Wal (1998) with a physically based one-dimensional flow model. They concluded that the glacier
would vanish by 2100 under a 4Ð4 ° C temperature increase (without precipitation modification).
Annual discharge
The simulated annual discharges show a significant decrease compared with the control period, for all
catchments and for all climate experiments (Table V). The median decrease ranges between 26% (Drance
de Bagnes) and 12% (Vorderrhein) under the B2 scenario, and between 30% (Drance de Bagnes) and
14% (Vorderrhein) under scenario A2.
It should be noted that the simulated mean discharge decrease may be overestimated as a result of the
highly simplified simulation of the glacier retreat. The glacier surface estimate for a future climate scenario
is based on the assumption that the glacier has reached a stationary state for the future climate situation
considered. This will presumably not be the case due to the reaction time of the glaciers (from several years
to a few decades depending on the glacier size and configuration). The glacier area for the future period that
would result from a given climate scenario will thus presumably be slightly larger than the one simulated
with our model. The glacier retreat for the A2 and B2 scenarios, however, is predicted to be so large that the
contribution of glacier melt to the future hydrological regime will be negligible.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2102
P. HORTON ET AL.
Table VI. Simulated actual evapotranspiration for the control and the future period for the B2 and A2 scenarios; the values
are specific values for the ice-free part of the catchment. Note: the area of the ice-free catchment part varies between the
climate experiments (see Tables I and V)
Catchment
Evapotranspiration (mm)
Control
Drance de Bagnes
Saaser Vispa
Lonza
Rhône at Gletsch
Weisse Lütschine
Minster
Tamina
Vorderrhein
Dischmabach
Rosegbach
Verzasca
Median increase
296
243
269
294
391
507
274
274
255
229
380
—
B2 scenario
A2 scenario
Min.
Med.
Max.
Min.
Med.
Max.
360
299
331
365
478
593
333
323
308
266
447
62
377
305
338
373
502
610
341
333
313
276
462
69
445
333
375
426
551
662
388
374
349
303
511
114
385
308
349
385
501
614
342
331
313
276
461
80
416
328
371
414
533
646
376
363
341
298
494
102
491
353
408
455
574
689
418
400
378
328
551
144
The considerable decrease of the mean annual discharges has several reasons, the most evident being the
decrease in mean annual precipitation. Another significant part of the discharge reduction is due to the predicted
increase of evapotranspiration. For the currently glacierized catchments, this increase is partly a consequence
of the significant glacier retreat, which results in an increase of the ice-free portion of the catchment where
evaporation is subtracted from precipitation (in the ice-covered areas, evaporation is fed by ice and firn that is
pre-existent to the simulation period considered). It is noteworthy that the increase in total evapotranspiration
in the simulation is not linked to a potential recolonization by plants of the newly exposed glacier-free surface.
It is induced by the increase of areas covered by bare soils and rocks that are also affected by evaporation. The
results obtained, however, rely on the important assumption that the relationship between evapotranspiration
and the hydrological model parameters governing it remains constant in the future. The model parameters are
calibrated for a given land-cover configuration, which for the high mountainous catchments studied includes
large surfaces comparable to those that would appear through glacier retreat. A significant land-cover change
would, nevertheless, require a recalibration. However, this is not possible in the context of the present study.
Additionally, the substantial temperature increase throughout the seasons enforces the total evapotranspiration on ice-free areas; accordingly, all catchments show a strong increase of total evapotranspiration
(Table VI). This increase is expected to be less important in summers than in winters: even if there is
a pronounced temperature increase in summer, summer evapotranspiration will be limited due to reduced
precipitation.
The important variability of annual precipitation changes simulated by the different climate models
(Table IV) and the significant variability of simulated changes in evapotranspiration (induced by the variability
of changes in seasonal temperatures, and thus in PET) explain the different magnitudes of mean annual runoff
decrease. Note that if the maximum predicted discharge decrease is significantly higher for the A2 scenario
than for the B2 scenario, then the median and minimum decreases obtained for both scenarios are quite
similar. The variability highlighted is mainly due to inter-model differences.
Hydrological regime modifications
All simulations under the future regional climate scenarios show the same significant changes of the
hydrological regimes and the same trends of these changes. Summer discharge is found to reduce significantly,
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
2103
winter discharge increases, and the snowmelt-induced peak is shifted to earlier in the year and decreases in
most cases (Figure 3). Owing to the substantial reduction of the glacierized areas, pure glacial discharge
regimes tend to disappear.
The changes are very similar for catchments having the same present regime, independently of their
geographical situation. A gradation of changes is observed between the B2 and the A2 scenario, and the
changes predicted for B2 tend to be enhanced for A2. The median simulated shift of the maximum monthly
or semi-monthly discharge is, for example, around half a month for B2 and an entire month for A2. The
results are, however, highly variable for the different RCM experiments.
For each future climate experiment, an attempt was made to identify the corresponding discharge regime on
the basis of the official Swiss classification (Aschwanden and Weingartner, 1985). However, future regimes
appear to differ from the regimes currently observed in Switzerland, and for many catchments the main
difference compared with the present regime types was the appearance of a small secondary peak in autumn
that does not exist in present intra-alpine regimes, but only in south-alpine regimes.
Figure 3 illustrates the simulated regime modifications for the Rhône catchment (‘a-glacial’ regime), the
Weisse Lütschine (‘a-glacio-nival’ regime), the Verzasca (‘nivo-pluvial meridional’ regime) and the Minster
catchment (‘nival of transition’ regime). For both scenarios, the Rhône catchment and the Weisse Lütschine
catchment tend to have a future regime essentially driven by spring snowmelt. The variability of the results
obtained for the different climate models is, however, high and the corresponding regimes can be quite
different. The snowmelt-induced peak occurs, for example, between late May and July. The attenuation of
this maximum monthly discharge is also highly variable from one scenario to another. The changes obtained
for the Verzasca catchment are slightly less variable. The shift of snowmelt is about half a month for all
experiments. The rainfall response peak in autumn does not shift in time, but does vary in amount. This
catchment is, thus, still expected to exhibit the same ‘nivo-pluvial meridional’ regime, as it does presently.
For the Minster catchment, the peak observed for the control run in May–June shifts to April–May under B2
and to even earlier for A2. The seasonality of discharges is still significant, but much less pronounced than
for the control period; for A2, the seasonality tends to disappear. For both emission scenarios, the predicted
changes vary significantly between the RCM experiments. For the two extreme climate experiments (RCMs
driven by ECHAM4/OPYC3), the Minster regime even becomes exclusively driven by rainfall.
It should be noted that the shift of a glacier- or snowmelt-driven hydrological regime to a future regime
more driven by snowmelt and/or rainfall will be associated with a modification of the year-to-year variability
of the discharges (Figure 4). For high-elevation catchments (catchments with current proportions of glacier
cover higher than 15%), the variation coefficient of annual discharges is, for example, predicted to increase by
around 85% for scenario A2 (from 45% to 90% for the Weisse Lütchine catchment and from 115% to 150%
for the Saaser Vispa catchment, depending on the RCM). The interannual variability of seasonal discharges
also changes significantly due to an important modification of the temporary water storage properties of these
mountainous catchments.
The large range of climate-change impacts predicted by the 19 RCM experiments is worthy of further
analysis. First, the way this variability acts on the discharge regime prediction has to be highlighted.
The sensitivity of the hydrological regimes to temperature and precipitation variability in the PRUDENCE
RCM experiments was analysed. For catchments with high elevations, the regime modifications are highly
conditioned by the changes of the seasonal temperatures (Figure 5). The simulated maximum monthly
discharge is, for example, highly correlated to the change in mean spring temperature (Figure 6). The main
reason is twofold. First, the seasonal temperature changes are highly correlated. Consequently, the higher the
future spring temperature is, the higher is the future winter temperature. Accordingly, the period during which
snow accumulates is shorter and the snowpack at the beginning of the snowmelt period is smaller. Second, the
future snowmelt peak is still predicted to occur in early summer; but, with higher mean spring temperatures,
the volume of snowmelt during spring is higher and, as a consequence, the volume of the snowmelt at the
beginning of the summer is reduced. The high variability of the regional temperature prediction, therefore,
has a strong influence on the range of predicted future regimes (Figure 5). For catchments at lower altitudes,
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2104
P. HORTON ET AL.
20
Rhone at Gletsch
18
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
20
16
14
12
10
8
6
4
2
0
J
F
M
A
M
J
J
A
S
O
N
12
10
8
6
4
J
F
M
A
M
J
J
A
S
O
N
D
J
F
M
A
M
J
J
A
S
O
N
D
0J
F
M
A
M
J
J
A
S
O
N
D
F
M
A
M
J
J
A
S
O
N
D
12
Weisse Lütschine
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
14
0
D
10
8
6
4
2
10
8
6
4
2
J
F
M
A
M
J
J
A
S
O
N
Verzasca
12
0
D
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
16
2
12
0
18
10
8
6
4
2
12
10
8
6
4
2
0
J
F
M
A
M
J
J
A
S
O
N
D
11
Minster
10
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
11
9
8
7
6
5
4
3
10
9
8
7
6
5
4
3
2
2
1
1
0
J
F
M
A
M
J
J
Control
Median of RCMs
B2 HadRM3(H)
B2 PROMES(H)
B2 RCAO(H)
A
S
O
N
B2 ARPEGE(H)
B2 ARPEGE(A)
B2 HIRHAM(E)
B2 RCAO(E)
D
0
J
Control
Median of RCMs
A2 HadRM3(H)
A2 CLM(H)
A2 RCAO(H)
A2 REMO(H)
A2 HIRHAM(H)
A2 CHRM(H)
A2 PROMES(H)
A2 RegCM(H)
A2 ARPEGE(H)
A2 ARPEGE(A)
A2 HIRHAM(E)
A2 RCAO(E)
Figure 3. Changes of the hydrological regimes obtained by the 19 PRUDENCE RCM experiments for the period 2070– 2099. Left:
B2 scenario (seven experiments); right: A2 scenario (12 experiments); regime simulated for the control period (1961– 1990) in bold
continuous line (results obtained for the other catchments are presented in Horton et al. (2005)). This figure is available in colour online at
http://www.interscience.wiley.com/hyp
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2105
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
30
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
30
25
20
15
10
5
0
J
F
M
A
M
J
J
A
S
O
N
D
25
20
15
10
5
0
J
F
M
A
M
J
J
A
S
O
Annual monthly discharge
25th percentile of the monthly discharge
Median of the monthly discharge
75th percentile of the monthly discharge
N
D
Figure 4. Interannual variability of monthly discharges of the Rhône catchment for the control period (left) and the period 2070– 2099 (A2
emission scenario) with RCAO(E) climate experiment (right); each line corresponds to 1 year of the 30-year simulation period. This figure
is available in colour online at http://www.interscience.wiley.com/hyp
the influence of precipitation modification is more pronounced and the variability of the climate-change
impacts is mainly due to the large range of predicted regional precipitation changes (Figure 5). For a rainfalldriven regime (the Verzasca catchment in Figure 6), the relationship (previously highlighted for high-elevation
catchments, between the maximum monthly discharge and the increase of spring temperature) is still significant
but much weaker. Conversely, the changes of spring precipitation become a more relevant explanatory variable
(Figure 6).
In terms of impact on the hydrological regime, this substantial contribution of the inter-RCM variability to
the total prediction variability can also be pointed out, as seen in the plots of the maximum monthly discharges
versus regional changes in spring temperature and precipitation (Figure 6). The results suggest that the driving
AOGCM may induce more impact prediction variability than the inter-RCM variability for a given AOGCM.
For example, for the RCAO regional climate model (Table II; indicated with large symbols in Figures 2
and 6), the maximum monthly discharge predicted for the Verzasca catchment for A2 with HadCM3 as a
driving model is around twice as high as with ECHAM4/OPYC3 as a driving model. Differences of results
obtained for two RCMs driven by the same AOGCM under the same emission scenario are often smaller.
Nevertheless, the results show that the inter-RCM variability is far from being negligible, e.g. as illustrated in
Figure 6 by the spread of the predicted maximum monthly discharges for HadCM3 under scenario A2. Note
that a quantitative evaluation of the uncertainty induced by each of the two climate modelling levels would
require more AOGCM and RCM experiments.
CONCLUSIONS
The climate-change impact study presented on mountainous discharge regimes is based on a classical
simulation approach using the outputs of a climate model to drive a hydrological model and a glaciercover evolution model. The analysis focuses on the climate-change prediction uncertainty induced by the use
of different climate models. Nineteen experiments were considered. They are based on two greenhouse-gas
emission scenarios (SRES A2 and B2), three coupled AOGCMs and nine different RCMs. The catchments
studied are representative of the different hydro-climatic regions of the Swiss Alps; accordingly, some general
conclusions can be drawn for the alpine area studied.
Considering the various hydrological regimes and the predictions associated with the different climate
models, there is a general trend in the simulated climate-change impact for both emission scenarios: the
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2106
20
20
18 Rhone at Gletsch
18
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
P. HORTON ET AL.
16
14
12
10
8
6
4
2
0
J
F
M A M
J
J
A
S
O N
Monthly mean runoff [mm/d]
Monthly mean runoff [mm/d]
8
6
4
2
F
M A M
12
10
8
6
4
2
J
F
M A M
J
J
A
S
O N
D
J
F
M A M
J
J
A
S
O N
D
12
10
J
14
0
D
12 Verzasca
0
16
J
J
A
S
O N
D
10
8
6
4
2
0
Control
Median of RCMs
A2 HadRM3(H)
A2 CLM(H)
A2 REMO(H)
A2 HIRHAM(H)
A2 CHRM(H)
A2 PROMES(H)
A2 RCAO(H)
A2 RegCM(H)
A2 ARPEGE(H)
A2 ARPEGE(A)
A2 HIRHAM(E)
A2 RCAO(E)
Figure 5. Sensitivity of the hydrological regimes to temperature and precipitation variability of the PRUDENCE RCM experiments (for
the Rhône and the Verzasca catchments, scenario A2, period 2070– 2099). Left: regime variability induced by the precipitation variability
under median predicted seasonal temperature increase; right: regime variability induced by temperature variability under median predicted
precipitation change. This figure is available in colour online at http://www.interscience.wiley.com/hyp
predicted climate change results in a significant decrease of the total annual discharge and in a shift of the
monthly maximum discharge to earlier periods of the year due to the temperature increase and the resulting
impact on the snow melting processes. An increase of the discharge year-to-year variability is also expected.
For a given RCM experiment, the hydrological regimes associated with each of the two emission scenarios
are clearly distinct. Considering all RCM experiments, the resulting ranges of hydrological regimes for both
emission scenarios are overlapping. Accordingly, an unambiguous prediction of the climate-change impacts
for a given greenhouse-gas emission scenario is not possible. The large prediction variability induced by the
19 RCM experiments is partly due to the underlying driving AOGCMs. This source of climate-change impact
uncertainty has already been highlighted by Zierl and Bugmann (2005) for alpine catchments. The results
of the present study clearly show that the use of several RCM experiments with the same driving AOGCM
can result in impact differences that are comparable to the results from the same RCM driven by different
AOGCMs. This leads to the conclusion that the quantitative assessment of hydrological changes based on a
small number of regional climate-change scenarios may yield misleading results.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2107
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
12
20
Rhone at Gletsch
18
Weisse Lütschine
10
Q max [mm/d] .
Q max [mm/d] .
16
14
12
10
R2 = 0.902
8
6
8
6
4
R2 = 0.829
4
2
2
0
0
0
1
2
3
4
5
XMT [°C]
6
7
0
8
14
2
3
4
5
XMT [°C]
6
7
8
14
Verzasca
Verzasca
12
12
10
10
Q max [mm/d] .
Q max [mm/d] .
1
8
6
R2 = 0.579
4
2
R2 = 0.857
8
6
4
2
0
0
1
2
3
4
5
XMT [°C]
6
7
8
0
−50
−30
−10
10
XMP [%]
HADCM3-A2
ARPEGE/OPA-A2
ECHAM4/OPYC3-A2
HADCM3-B2
ARPEGE/OPA-B2
ECHAM4/OPYC3-B2
Control period
30
50
R2 Coefficient of determination of the linear regression
Figure 6. Maximum mean monthly discharge Qmax as a function of regional changes of either mean spring (March, April, May) temperature
(XMT) or mean spring precipitation (XMP) (large symbols identify the RCAO experiments)
It is also noteworthy that additional prediction uncertainty may arise from the method used for the
development of local-scale scenarios. In this study, a simple perturbation methodology was used, but it
would be interesting to compare the present results with those obtained through local-scale climate models
or statistical downscaling models that connect the local meteorological variables to synoptic-scale variables
(e.g. see Wilby and Wigley (1997) and Xu (1999) for a review of downscaling approaches).
The present analysis does not consider the modelling uncertainties associated with the hydrological model.
Schaefli (2005) showed that the uncertainty induced by the parameter uncertainty and the total modelling error
for a given model structure is far less important than the uncertainty due to the climate predictions. However,
as for the climate models, it is not possible to design a unique best hydrological model for a given catchment.
The one that has been retained for this study represents just one possible model structure. The analysis
of several equivalent hydrological models would complete the current results. The multi-model approach
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
2108
P. HORTON ET AL.
presented by Schaefli (2005) suggests that different hydrological models could increase the uncertainty of the
future discharge predictions considerably. An important additional source of uncertainty is the estimation of
the PET and the simulation of the corresponding actual evapotranspiration through the hydrological model.
Further research in this area is crucial for a detailed analysis of the climate-change impact uncertainty inherent
in the system response modelling.
ACKNOWLEDGEMENTS
Part of this work has been funded by the Swiss Federal Office for Energy. We also wish to thank the Swiss
Federal Office of Hydrology and Geology, for providing the discharge data, and the national weather service,
MeteoSwiss, for providing the meteorological time-series. We are grateful to the two anonymous reviewers
and the co-editors for their comments on the manuscript.
REFERENCES
Anonymous. 1969. Mass-balance terms. Journal of Glaciology 8(52): 3–7.
Arnell NW, Hulme M. 2000. Implications of climate changes for large dams and their management . Thematic Review II.2 prepared as an
input to the World Commission on Dams, Cape Town.
Arribas A, Gallardo C, Gaertner M, Castro M. 2003. Sensitivity of the Iberian Peninsula climate to a land degradation. Climate Dynamics
20: 477–489.
Aschwanden H, Weingartner R. 1985. Die Abflussregimes der Schweiz . Publikation Gewässerkunde Nr. 65, Geographisches Institut der
Universität Bern.
Beniston M, Rebetez M, Giorgi F, Marinucci MR. 1994. An analysis of regional climate-change in Switzerland. Theoretical and Applied
Climatology 49(3): 135– 159.
Benn DI, Lehmkuhl F. 2000. Mass balance and equilibrium-line altitudes of glaciers in high-mountain environments. Quaternary International 65–66: 15–29.
Braun LN, Weber M, Schulz M. 2000. Consequences of climate change for runoff from alpine regions. Annals of Glaciology 31: 19–25.
Burlando P, Pellicciotti F, Strasser U. 2002. Modelling mountainous water systems between learning and speculating looking for challenges.
Nordic Hydrology 33(1): 47–74.
Burman R, Pochop LO. 1994. Evaporation, Evapotranspiration and Climatic Data. Elsevier: Amsterdam.
Christensen JH, Räisänen J, Iversen T, Bjørge D, Christensen OB, Rummukainen M. 2001. A synthesis of regional climate change
simulations—a Scandinavian perspective. Geophysical Research Letters 28: 1003– 1006.
Christensen JH, Carter TR, Giorgi F. 2002. PRUDENCE employs new methods to assess European climate change. Eos, Transactions of
the AGU 83: 147.
Consuegra D, Vez E. 1996. AMIE: Analyse et Modélisation Intégrées du cheminement des eaux en zones habitées, modélisation hydrologique,
Application au bassin versant de la Haute Broye. IATE/HYDRAM, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland.
Doms G, Schättler U. 1999. The nonhydrostatic limited-area model LM (Lokal-Modell) of DWD, part I: scientific documentation. Deutscher
Wetterdienst, Meteorologische Analyse und Modellierung, Offenbach, Germany.
Ekström M, Jones PD, Fowler HJ, Lenderink G, Buishand TA, Conway D. 2006. Regional climate model data used within the SWURVE
project 1: projected changes in seasonal patterns and estimation of PET. Hydrology and Earth System Sciences accepted for publication.
Etchevers P, Golaz C, Habets F, Noilhan J. 2002. Impact of a climate change on the Rhone river catchment hydrology. Journal of Geophysical
Research: Atmospheres 107(D16): 4293, DOI: 10.1029/2001JD000490.
Frei C, Christensen JH, Deque M, Jacob D, Jones RG, Vidale PL. 2003. Daily precipitation statistics in regional climate models: evaluation
and intercomparison for the European Alps. Journal of Geophysical Research: Atmospheres 108(D3): 4124, DOI: 10.1029/2002JD002287.
Gibelin AL, Déqué M. 2003. Anthropogenic climate change over the Mediterranean region simulated by a global variable resolution model.
Climate Dynamics 20(4): 327–339.
Giorgi F, Marinucci MR, Bates GT. 1993a. Development of a second generation regional climate model (REGCM2). Part I: boundary layer
and radiative transfer processes. Monthly Weather Review 121: 2794– 2813.
Giorgi F, Marinucci MR, Bates GT, DeCanio G. 1993b. Development of a second generation regional climate model (REGCM2). Part II:
cumulus cloud and assimilation of lateral boundary conditions. Monthly Weather Review 121: 2814– 2832.
Gordon C, Cooper C, Senior CA, Banks H, Gregory JM, Johns TC, Mitchell JFB, Wood RA. 2000. The simulation of SST, sea ice extents
and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dynamics 16(2–3): 147– 168.
Gross G, Kerschner H, Patzelt G. 1976. Methodische Untersuchungen über die Schneegrenze in alpinen Gletschergbieten. Zeitschrift für
Gletscherkunde und Glazialgeologie 12: 223–251.
Hingray B, Mezghani A, Buishand TA. 2006a. Development of probability distributions for regional climate change from uncertain global
mean warming and an uncertain scaling relationships. Hydrology and Earth System Sciences accepted for publication.
Hingray B, Mouhous N, Mezghani A, Bogner K, Schaefli B, Musy A. 2006b. Accounting for global warming and scaling uncertainties in
climate change impact studies: application to a regulated lakes system. Hydrology and Earth System Sciences accepted for publication.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)
CLIMATE-CHANGE IMPACTS ON ALPINE DISCHARGE REGIME
2109
Horton P, Schaefli B, Mezghani A, Hingray B, Musy A. 2005. Prediction of climate change impacts on alpine discharge regimes under A2
and B2 SRES emission scenarios for two future time periods (2020–2049, 2070–2099). HYDRAM Technical Report, Swiss Federal Office
for Energy. Bern (available from HYDRAM, EPFL, CH-1015 Lausanne).
Hulme M, Jenkins GJ, Lu X, Turnpenny JR, Mitchell TD, Jones RG, Lowe J, Murphy JM, Hassell D, Boorman P, McDonald R, Hill S.
2002. Climate change scenarios for the United Kingdom: the UKCIP02 scientific report . Tyndall Centre for Climate Change Research,
School of Environmental Sciences, University of East Anglia, Norwich, UK.
Jacob D. 2001. A note to the simulation of the annual and inter-annual variability of the water budget over the Baltic Sea drainage basin.
Meteorology and Atmospheric Physics 77(1– 4): 61–73.
Jasper K, Calanca P, Gyalistras D, Fuhrer J. 2004. Differential impacts of climate change on the hydrology of two alpine river basins.
Climate Research 26(2): 113– 129.
Jenkins G, Lowe J. 2003. Handling uncertainties in the UKCIP02 scenarios of climate change. Technical note 44, Hadley Centre, Exeter,
UK.
Kirchhofer W, Sevruk B. 1991. Mean annual corrected precipitation depths 1951– 1980. In Atlas Hydrologique de la Suisse. Service
Hydrologique et Géologique National: Bern.
Klok EJ, Jasper K, Roelofsma KP, Gurtz J, Badoux A. 2001. Distributed hydrological modelling of a heavily glaciated alpine river basin.
Hydrological Sciences Journal 46(4): 553– 570.
Laternser M, Schneebeli M. 2002. Spatial grouping of snow stations, snow and avalanche climatology of Switzerland . ETH Zürich.
Meier MF, Post AS. 1962. Recent variations in mass net budgets of glaciers in western North America. In Symposium of Obergurgl, 10–18
September 1962, Variations of the Regime of Existing Glaciers. IAHS Publication No. 58. IAHS Press: Wallingford; 63–77.
Nakićenović N. 2000. Greenhouse gas emissions scenarios. Technological Forecasting and Social Change 65: 149–166.
Nakićenović N, Swart R (eds). 2000. Special Report on Emissions Scenarios (Intergovernmental Panel on Climate Change). Cambridge
University Press: Cambridge, UK.
Nash JE, Sutcliffe JV. 1970. River flow forecasting through conceptual models. Part I, a discussion of principles. Journal of Hydrology
10(3): 282– 290.
New M, Hulme M, Jones P. 2000. Representing twentieth-century space–time climate variability. Part II: development of 1901– 96 monthly
grids of terrestrial surface climate. Journal of Climate 13(13): 2217– 2238.
Oerlemans J, Anderson B, Hubbard A, Huybrechts P, Johannesson T, Knap WH, Schmeits M, Stroeven AP, van de Wal SW, Wallinga J,
Zuo Z. 1998. Modelling the response of glaciers to climate warming, Climate Dynamics 14(4): 267– 274.
Pardé M. 1933. Le régime des cours d’eau de l’Europe orientale (en collaboration avec St. Kolupaila). Revue Geographie Alpine T.XXI(IV):
651– 748.
Paterson WSB. 1994. The Physics of Glaciers. Pergamon: Oxford.
Pope VD, Gallani ML, Rowntree PR, Stratton RA. 2000. The impact of new physical parametrizations in the Hadley Centre climate
model—HadAM3. Climate Dynamics 16: 123–146.
Porter SC. 1975. Equilibrium-line altitudes of late Quaternary glaciers in Southern Alps, New Zealand. Quaternary Research 5(1): 27–47.
Räisänen J, Hansson U, Ullerstig A, Döscher R, Graham LP, Jones C, Meier HEM, Samuelsson P, Willén U. 2004. European climate in
the late 21st century: regional simulations with two driving global models and two forcing scenarios. Climate Dynamics 22(1): 13–31.
Rango A, Martinec J. 1995. Revisiting the degree-day method for snowmelt computations. Water Resources Bulletin 31(4): 657– 669.
Roeckner E, Bengtsson L, Feichter J, Lelieveld J, Rodhe H. 1999. Transient climate change simulations with a coupled atmosphere–ocean
GCM including the tropospheric sulfur cycle. Journal of Climate 12(10): 3004– 3032.
Royer JF, Cariolle D, Chauvin F, Déqué M, Douville H, Planton S, Rascol A, Ricard JL, Salas y Melia D, Sevault F, Simon P, Somot S,
Tyteca S, Terray L, Valcke S. 2002. Simulation of climate changes during the 21st century including stratospheric ozone. Comptes Rendus
Geoscience 334: 147–154.
Schädler B, Bigler R. 2002. Components of the natural water balance 1961– 1990. In Atlas Hydrologique de la Suisse. Service Hydrologique
et Géologique National: Bern.
Schaefli B. 2005. Quantification of modelling uncertainties in climate change impact studies on water resources: application to a glacier-fed
hydropower production system in the Swiss Alps. Doctoral thesis, Ecole Polytechnique Fédérale de Lausanne.
Schaefli B, Hingray B, Niggli M, Musy A. 2005. A conceptual glacio-hydrological model for high mountainous catchments. Hydrology and
Earth System Sciences 9: 95–109.
Shabalova MV, Van Deursen WPA, Buishand TA. 2003. Assessing future discharge of the River Rhine using regional climate model
integrations and a hydrological model. Climate Research 23: 223–246.
SwissTopo. 1995. Digital height model of Switzerland—DHM25 . Swiss Federal Office of Topography, Wabern, Switzerland.
SwissTopo. 1997. Digital national maps of Switzerland—PM25 . Swiss Federal Office of Topography, Wabern, Switzerland.
Vidale PL, Lüthi D, Frei C, Seneviratne S, Schär C. 2003. Predictability and uncertainty in a regional climate model. Journal of Geophysical
Research: Atmospheres 108(D18): 4586. DOI: 10.1029/2002JD002810.
Wallinga J, van de Wal RSW. 1998. Sensitivity of Rhonegletscher, Switzerland, to climate change: experiments with a one-dimensional
flowline model. Journal of Glaciology 44(147): 383– 393.
Weber RO, Talkner P, Auer I, Bohm R, GajicCapka M, Zaninovic K, Brazdil R, Fasko P. 1997. 20th-century changes of temperature in the
mountain regions of central Europe. Climatic Change 36(3–4): 327– 344.
Wilby RL, Wigley TML. 1997. Downscaling general circulation model output: a review of methods and limitations. Progress in Physical
Geography 21(4): 530– 548.
Xu CY. 1999. From GCMs to river flow: a review of downscaling methods and hydrologic modelling approaches. Progress in Physical
Geography 23(2): 229– 249.
Zierl B, Bugmann H. 2005. Global change impacts on hydrological processes in Alpine catchments. Water Resources Research 41(2):
W02028. DOI: 10.1029/2004WR003447.
Copyright  2006 John Wiley & Sons, Ltd.
Hydrol. Process. 20, 2091– 2109 (2006)