Réf. Bavay & al. 2009 - A

Référence bibliographique complète

BAVAY, M., LEHNING, M., JONAS, T., LÖWE, H. 2009. Simulations of future snow cover and discharge in Alpine headwater catchments, Hydrological Processes, 22, DOI: 10.1002/hyp.7195.

Abstract: The snow cover in the Alps is heavily affected by climate change. Recent data show that at altitudes below 1200 m a.s.l. a time-continuous winter snow cover is becoming an exception rather than the rule. This would also change the timing and characteristics of river discharge in Alpine catchments. We present an assessment of future snow and runoff in two Alpine catchments, the larger Inn catchment (1945 km²) and the smaller Dischma catchment (43 km²), based on two common climate change scenario (IPCC A2 and B2 (IPCC, 2007)). The changes in snow cover and discharge are predicted using Alpine3D, a model for the high-resolution simulation of Alpine surface processes, in particular snow, soil and vegetation processes. The predicted changes in snow and discharge are extreme. While the current climate still supports permanent snow and ice on the highest peaks at altitudes above 3000 m a.s.l., this zone would disappear under the future climate scenarios. The changes in snow cover could be summarized by approximately shifting the elevation zones down by 900 m. The corresponding changes in discharge are also severe: while the current climate scenario shows a significant contribution from snow melt until middle to late summer, the future climate scenarios would feature a much narrower snow melt discharge peak in spring. A further observation is that heavy precipitation events in the fall would change from mainly snow to mainly rain and would have a higher probability of producing flooding. Future work is needed to quantify the effect of model uncertainties on such predictions.

Mots-clés
Climate change; Catchment hydrology; Water resources; Flow regime; Flooding; Snow melt

Organismes / Contact

WSL Institute for Snow and Avalanche Research, SLF Davos, Flüelastr. 11, 7260 Davos Dorf, Switzerland - bavay@slf.ch (emails: bavay, lehning, jonas, loewe @slf.ch)


(1) - Paramètre(s) atmosphérique(s) modifié(s)
(2) - Elément(s) du milieu impacté(s)
(3) - Type(s) d'aléa impacté(s)
(3) - Sous-type(s) d'aléa
Temperature Snow cover, River (discharge) Floods  

Pays / Zone
Massif / Secteur
Site(s) d'étude
Exposition
Altitude
Période(s) d'observation
Switzerland eastern Swiss Alps Inn and Dischma catchments Dischma: mainly S–N
Inn: mainly W–E
Dischma: 1677-3130m a.s.l.
Inn: 1045-3868m a.s.l.
 

(1) - Modifications des paramètres atmosphériques
Reconstitutions
 
Observations
 
Modélisations
 
Hypothèses
 

Informations complémentaires (données utilisées, méthode, scénarios, etc.)
 

(2) - Effets du changement climatique sur le milieu naturel
Reconstitutions
 
Observations

Massive changes of the mountain snow cover are expected and have already been detected (Marty, 2008). Trend analyses of changes in mountain snow covers have been carried out by Hantel et al. (2000); Laternser and Schneebeli (2003) and Scherrer et al. (2004) for the Alps [...]. All studies agree that the most pronounced changes are detected for lower altitudes. This finding is consistent with the explanation that temperature trends are most visible at those elevations for which a small change in temperature leads to increased rain versus snow precipitation (change in snow line).

Modélisations

A number of model-based assessments have been made of the hydrological response of mountain catchments to climate change (e.g. Singh and Kumar, 1997; Horton et al., 2006; Stahl et al., 2008). These first quantitative results point in particular to the influence of shrinking ice masses. They show that pronounced effects are predicted for glaciated catchments, where high runoff levels shift from summer to spring because melt rates of the ice sheet decrease after a transitional phase of increased runoff (Stahl et al., 2008). In a glaciated catchment, a major effect, i.e. the disappearance of permanent snow cover for this particular catchment, has been found for the accumulation zone at higher altitudes (Magnusson et al., 2008).

Results of the present study

Snow cover changes:
Under the current climate, the transition between marginal and dominant snow cover appears to be around 2000 m a.s.l. Above this altitude, snow is present for 9–10 months per year.

In simulations of the average snow distribution for scenario runs A2 and B2, the average snow water equivalent for each altitude band as a function of the time of the year changes dramatically in both scenarios and for both catchments (Dischma and Inn). At all elevation ranges the maximum snow water equivalent and the duration of a significant snow cover would be reduced. The reduction in snow would be even more severe under the A2 scenario.

When looking at timing and magnitude of the peak snow water equivalent, the authors arbitrarily introduce two classes: above and below 2000 m a.s.l. For the lower elevation of the Dischma catchment, the peak in snow water equivalent shifts by 36 days from 11 April to 6 March in the A2 scenario, and by 28 days from 11 April to 14 March in the B2 scenario. The maximum snow water equivalent is reduced by 47 and 36% in the A2 and B2 scenarios, respectively. For the higher elevation, the change would be similar (44 and 29% for the two scenarios) but the time shift would be less pronounced (16 and 7 days respectively). It is interesting to note that the changes in the A2 scenario would roughly correspond to shifting the reference simulation down by 900 m for both catchments. The numbers for the Inn catchment are similar for the maximum snow water equivalent reduction, but the trends on the time shift are opposite, seeing a slightly increased shift for the higher altitudes compared to the lower altitudes (15 days above 2000 m compared to 7 days below 2000 m for both scenarios). This is suspected to be the result of the uncertainties of the approach and may be related to the lack of input data representative of the highest elevations for the Inn catchment.

The time of the year at which the last snow disappears in a given altitude band has also been extracted from the simulations and averaged over the simulation period. In order to do so, a 20 mm snow water equivalent threshold has been arbitrarily defined to be ’no snow left’. The beginning of a new snow season for a given altitude band is defined as the first day from which the average snow water equivalent of the entire altitude band does not goes below the threshold anymore. The end of the snow season is defined as the first day that sees the average snow water equivalent of the entire altitude band dropping below the threshold. Practically, this means that snow accumulated in a shaded area could significantly contribute to alter the beginning and end dates of the snow season for its elevation band. Nevertheless, this definition makes sense from the hydrological point of view as it represents the contributions of a given elevation band to the global discharge of the catchment. In the snow cover duration curve, the reference simulation as well as the climate scenarios have been plotted for comparison.

The snow cover duration curve shows that the beginning of the snow season differs between the Inn and the Dischma catchments. It might come from the relatively dry climate of the Inn catchment. This is very visible for the Inn catchment where, for a lot of altitude bands, the snow season starts during a specific precipitation event at the end of November. The melt phase however, is not significantly affected by precipitation, therefore, the two catchments show very similar behaviour. Note that the continuous snow cover criteria can never be fulfilled in an average year below 1000 m a.s.l. (for the reference run), therefore, no data is produced for these low elevations.

For the Dischma catchment, the two climate scenarios would differ approximately by a constant offset. However, the snow season duration is more affected at higher altitudes. This result on future snow cover temporal patterns differs from the observations on past snow cover patterns made by Marty (2008), who found that altitudes below about 1800 m have already shown a stronger reaction on observed warming than higher altitudes. Since the Inn simulation does not show such an elevation dependence, the result might again be a consequence of the altitude—exposition distribution in the Dischma catchment. A further reason might be that the strong scenario shifts the altitude band of maximum sensitivity upwards. This is further discussed in the conclusions.

It appears with the predicted warming that permanent snow or ice would no longer be possible in the Dischma or Inn catchments, which is in agreement with the prediction that the few smaller glaciers at these lower altitudes would completely disappear in the Alps (Zemp et al., 2006). The elevations which have the highest area fractions would experience a complete melt approximately 40 days earlier in the A2 scenario and 35 days earlier in the B2 scenario. The highest elevations would experience a 60- or 50-day shift earlier for scenarios A2 or B2, respectively. For the Inn catchment, most altitudes would experience a complete melt approximately 40 days earlier in the A2 scenario.

Comparing the average snow depths at three stations (Stillberg, Teufi and Davos) between the reference run and scenarios A2 and B2, the snow season would be shortened as described above, but the snow height would be more affected at lower elevations: the A2 scenario would experience a 40% reduction of snow height for Stillberg, a 49% reduction at Teufi and a 54% reduction at the SLF, Davos, station. In the B2 scenario, these would become 30, 33 and 42%, respectively.

Discharge:
The changes in snow cover discussed above are consistently represented in discharge changes.

The major change that is visible when comparing the reference simulation discharge and the climate change scenarios discharges is a strong peak in early May in the climate change scenarios not visible in the reference. This is the beginning of snow melt in the catchment, which occurs a month earlier in both scenarios than in the reference case. There are two main observations to be made: (i) the snow melt discharge peak is not only earlier but also more pronounced for the Dischma catchment; (ii) the higher snow melt peak is more concentrated in late spring/early summer, while it is spread over a longer period under the current climate. This means that snow contributes to the discharge over a much shorter time, but at the same time, the probability of spring flooding might increase. The two scenarios are qualitatively similar, yet showing differences in the magnitude of snow melt discharge. The two scenarios merge again in June, indicating the time at which the influence of snow melt on discharge becomes negligible in the catchment for both scenarios. Over the summer, discharge is then determined by rainfall. In contrast, the reference run shows snow influence on discharge until early August—discharge from snow, therefore, being significant for a large part of the year.

For the Inn catchment, the spring discharge peak is not as pronounced as for the Dischma. The Inn catchment being much dryer than the Dischma catchment, particularly in winter, the spring discharge depends on snow built up during late fall. Since in the A2 scenario, these precipitation events occur as liquid precipitation for a much larger fraction of the altitudes, there is less snow available to melt in spring. Therefore, despite an earlier and shorter melt period for the Inn catchment, there is no increase in the amplitude of the spring discharge. Moreover, the lowest elevations are not able to build up a snow pack anymore, and thus, cannot contribute to the discharge in the spring. This is in contrast to the Dischma simulation which contains a much narrower range of elevations, hence the fraction of snow-free altitude bands is insignificant. Finally, the more consistent distribution of exposure in the Inn simulation means that shading effects do not have a strong impact on the discharge (a given altitude band melts more uniformly) compared to the Dischma catchment.

Another feature of the data is local discharge peaks in late autumn: in the reference simulation for both catchments, the precipitation events at that time of the year produce snow down to 1600 m a.s.l., not leading to a significant discharge increase. In the climate change scenario, no snow is seen below 2200 m a.s.l., and significant discharge peaks are created. Again, since the Inn catchment contains a significant proportion of low elevations, this effect is amplified and might increase the risk of flooding.

Hypothèses
 

Sensibilité du milieu à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)
 

The authors present a comparative analysis of catchments of two different sizes. They use the advanced modelling system, the Alpine3D, for their model runs, which provides a detailed process representation (Lehning et al., 2006) especially for snow (Lehning et al., 2002) and appears, therefore, particularly suited for change studies.

The rectangular simulation domain encompassing the Dischma catchment (43.3 km²) covers a wider area of 12.8 × 15.4 km (197 km²). The domain is meshed by 100 × 100-m-square cells. For the Inn catchment (4771 km²), which contains the Dischma catchment (which is, however, not part of the Inn catchment), the domain is meshed by 250 × 250-m-square cells.

The Alpine3D model:
The Alpine3D is a model system of Alpine surface processes with an emphasis on snow processes. The model is based on a fully distributed application of SNOWPACK (Lehning et al., 1999; Lehning et al., 2002), which is an advanced model of snow cover development. SNOWPACK has been used for studying snow dynamics in the context of avalanche warning (e.g. Nishimura et al., 2005), glacier mass balance and hydrology (e.g. Michlmayr et al., 2008) and already for climate change scenarios in high latitudes (Rasmus et al., 2004), and for Alpine permafrost (Luetschg et al., 2008). The single point snow column description provided by SNOWPACK includes a parameterized module for vegetation. Snow and soil dynamics are numerically represented by a large arbitrary number of layers. The high resolution of the surface snow or soil layers allows a more accurate surface energy balance to be provided as well as allowing the vertical water transport in snow and soil to be described with a simple bucket scheme (Bartelt and Lehning, 2002).

In Alpine3D, atmospheric forcing is allowed to vary continuously in space such that the one-dimensional vegetation—snow—soil columns at the grid points will also vary as a result. Runoff is calculated by a simple conceptual model, which is fed from the vegetation—snow—soil columns (Lehning et al., 2006). It is important to note that for climate change scenarios, one should also consider that vegetation and even soil would change. This could not be implemented for this study since predictions of these changes were not available. It is, however, assumed that the effects of changes in vegetation and soil will remain small compared to other model errors such as the interpolation of the meteorological forcing. By contrast, the changing climate is unlikely to change much of the deeper sub-surface flow paths such that the assumption of stationarity appears to be justified for the conceptual runoff model.

Reference meteorological data:
Hourly data from 10 weather stations in the Dischma and 18 stations in the Inn catchment, respectively, were used as model input for establishing a reference run. The stations represent an altitude range from 1560 to 2725 m a.s.l. in the Dischma and 1078–3315 m a.s.l. in the Inn catchment. Temperature, humidity, wind speed and precipitation input data were distributed across the grid using WINMET (Zappa et al., 2003), a tool for geo-statistical interpolation of meteorological data. For both catchments, we used hourly measurements of incoming longwave radiation from the Weissfluhjoch station (2537 m a.s.l.) above Davos, 4 km north of the Dischma catchment outlet. For the Dischma catchment, shortwave radiation was also taken from Weissfluhjoch, whereas the Samedan meteorological station (1707 m a.s.l.) served as shortwave radiation input for the Inn catchment.

RCM predictions:
To generate model input data representative of future climate, observed data were altered according to expected changes as predicted by a set of RCMs (Regional Climate Models) which participated in the PRUDENCE project (Prediction of Regional scenarios and Uncertainties for Defining European Climate Change Risks and Effects) (Christensen et al., 2007). A sub-set of 12 runs was selected to provide simulations for the changes in temperature, precipitation and longwave radiation between a reference period, 1961–1990, and a future period, 2071–2100, for the two IPCC greenhouse gas emission standard scenarios, SRES A2 and B2. The following five RCMs were used for the simulations: HIRHAM (used by the Danish and Norwegian Meteorological Institutes), PROMES (Universidad Complutense de Madrid), RCAO (Swedish Meteorological and Hydrological Institute), HadRM3P (Hadley Centre for Climate Prediction and Research) and RegCM (Abdus Salam Intl. Centre for Theoretical Physics). For the present study, the authors used daily minimum and maximum temperatures, daily accumulated precipitation and daily averages of incoming longwave radiation. These parameters were expected to have the greatest influence on snowpack characteristics. Potential changes in relative humidity, wind speed and shortwave radiation were assumed to be more uncertain and were, therefore, not taken into account. For each catchment, we used the data of the nearest grid cell from the RCM output (spatial resolution ~50 km).

Stochastic generation of input data for A2/B2 runs:
We modified the observed data series of precipitation, longwave radiation and temperature to represent climatic conditions towards the end of the twenty-first century by largely following the methods described in López-Moreno et al. (2008) [see details in the study].

Presentation of the results:
Since this study focuses on general trends, it has been decided that most of the results would be presented as yearly averages, resulting in more condensed figures. The modelling has been performed for each year of the period of interest, and the outputs have been averaged to compute the general trends of an average year according to the chosen scenario (from August 2000 to August 2006 for the Dischma catchment, and from August 2001 to August 2006 for the Inn catchment). Therefore, year-toyear variability is smoothed in the results presented.

Verification:
In order to explore the reliability of the method used in this study, a reference Alpine3D run was defined using unmodified meteorological input. The outputs of this reference run were compared to measurements in order to check the accuracy of the simulation. Two parameters were examined: the snow cover and the catchment discharge. The comparison between the measured and the simulated data shows the model to be in reasonable agreement with the snow cover measurements. The overall nature of the discharge is satisfactorily captured by Alpine3D. The authors judge the discharge simulation to be sufficiently accurate to proceed with climate change simulations.


(3) - Effets du changement climatique sur l'aléa
Reconstitutions
 
Observations
 
Modélisations

The major change that is visible when comparing the reference simulation discharge and the climate change scenarios discharges is a strong peak in early May in the climate change scenarios not visible in the reference. This is the beginning of snow melt in the catchment, which occurs a month earlier in both scenarios than in the reference case. There are two main observations to be made: (i) the snow melt discharge peak is not only earlier but also more pronounced for the Dischma catchment; (ii) the higher snow melt peak is more concentrated in late spring/early summer, while it is spread over a longer period under the current climate. This means that snow contributes to the discharge over a much shorter time, but at the same time, the probability of spring flooding might increase. The two scenarios are qualitatively similar, yet showing differences in the magnitude of snow melt discharge. The two scenarios merge again in June, indicating the time at which the influence of snow melt on discharge becomes negligible in the catchment for both scenarios. Over the summer, discharge is then determined by rainfall. In contrast, the reference run shows snow influence on discharge until early August—discharge from snow, therefore, being significant for a large part of the year.

For the Inn catchment, the spring discharge peak is not as pronounced as for the Dischma. The Inn catchment being much dryer than the Dischma catchment, particularly in winter, the spring discharge depends on snow built up during late fall. Since in the A2 scenario, these precipitation events occur as liquid precipitation for a much larger fraction of the altitudes, there is less snow available to melt in spring. Therefore, despite an earlier and shorter melt period for the Inn catchment, there is no increase in the amplitude of the spring discharge. Moreover, the lowest elevations are not able to build up a snow pack anymore, and thus, cannot contribute to the discharge in the spring. This is in contrast to the Dischma simulation which contains a much narrower range of elevations, hence the fraction of snow-free altitude bands is insignificant. Finally, the more consistent distribution of exposure in the Inn simulation means that shading effects do not have a strong impact on the discharge (a given altitude band melts more uniformly) compared to the Dischma catchment.

Another feature of the data is local discharge peaks in late autumn: in the reference simulation for both catchments, the precipitation events at that time of the year produce snow down to 1600 m a.s.l., not leading to a significant discharge increase. In the climate change scenario, no snow is seen below 2200 m a.s.l., and significant discharge peaks are created. Again, since the Inn catchment contains a significant proportion of low elevations, this effect is amplified and might increase the risk of flooding.

Hypothèses
 

Paramètre de l'aléa
Sensibilité des paramètres de l'aléa à des paramètres climatiques
Informations complémentaires (données utilisées, méthode, scénarios, etc.)
 

[See above]


(4) - Remarques générales

For the current study, the authors do not attempt to discuss the different reasons for a change in snow cover such as, change in snowline or change in melt in detail, but instead, concentrate on changes in the spatial patterns of snow cover distribution and temporal patterns of catchment discharge.

To assess the overall effect of climate change on the seasonal snow cover, snow models have been used, but investigations continue to be focused on individual point measurement stations only (Rasmus et al., 2004; López-Moreno et al., 2008). There are, thus, two major reasons for additional model studies on the combined behaviour of snow and discharge under climate change scenarios. First, a variety of models with different complexity and physics must be used to build confidence in the model projections. Second, the critical behaviour of the snow cover needs to be assessed with respect to runoff formation by applying detailed snow models at a distribution of points across the catchment. The novel aspects of the present approach are that, to the author's knowledge, no combined and detailed investigation of snow and discharge response to climate change scenarios for Alpine headwater catchments is available. [...] Note that change studies require extrapolation from current conditions, and therefore, conceptual models which require a high degree of calibration, may not always be the best choice for change prediction.

Limitations

Climate models:
The meteorological input data for the Alpine3D future snow cover simulations was derived from observed data outside the RCMs’ reference period. Better practice would have been to have a model run driven by observed data within the RCMs’ reference period. This would have been possible but at the expense of losing the spatial resolution of input data, i.e. losing 9 out of 10 stations for the Dischma run. This is because the monitoring network, IMIS (Lehning et al., 1999) which was founded in 1998, only achieved a reasonable station density around 2002. Hence, for the reference run, we chose a period (2002–2006) slightly after the RCM reference period (1961–1990) assuming this would still be reasonably close to the reference period if compared to the 2071–2100 period. Applying shifts in meteo variables stemming from simulations of two periods 110 years apart to describe snow cover changes over a 90-year period might lead to a slight overestimation of predicted changes. However, this problem should be partly mitigated if, as expected, climate change will become more pronounced towards the second half of the century.

The authors also chose to combine the climate change predictions from six RCMs to stipulate at least four numerical experiments (two scenarios, two catchments) as it would not have been possible to run 24 numerical experiments due to computational constraints. Therefore, we cannot discuss the influence of uncertainties in the input data on variation in the output results. However, using two scenarios (SRES A2 & B2) allows at least some basic consideration of uncertainty in the driving data.

Alpine3D limitations:
This study uses Alpine3D to predict snow cover and associated changes in discharge based on climate change scenarios. In addition to the uncertainties coming from the climate change calculations, this introduces uncertainties related to the model, which are discussed here. (i) A major weakness of any such study is that the climate change scenario data are a priori inconsistent with the result of the physical downscaling. The example of the snow cover is very illustrative in this context. While most climate models use a very simplified snow description and are not able to resolve smaller-scale topography in the mountains because of their coarse resolution, the downscaling produces a much more detailed snow representation, which must therefore be locally inconsistent with the climate model. In our study, we assume that the stochastic method of applying changes in the forcing parameters to locally measured time series is a reasonable way to deal with this inconsistency. (ii) A second source of uncertainty is the interpolation of meteorological fields necessary to drive Alpine3D. While the input fields are based on local measurements as described above, the resolution of the stations is still insufficient to represent the smallscale variation of weather in such mountainous terrain. Therefore, the predictions of the local snow cover dynamics represent the error from the interpolation of weather input. While the absolute error arising from this may be quite large, the uncertainty should be limited as far as the assessment of change is concerned. Nonetheless, this is considered to be a significant yet hard to quantify source of uncertainty. (iii) For the scale investigated here, lateral transport of snow through wind and avalanches has been neglected. This error is also thought to have more influence on the absolute amount of snow present at a grid point than for a prediction of change. (iv) The onedimensional energy and mass balance of SNOWPACK is thought to be very accurate and a minor source of error. This has been validated for many climates worldwide and the change simulations should not, therefore, have a major uncertainty derived from this. (v) The conceptual runoff part is only responsible for the sub-soil water movement and should, therefore, also be robust for our climate change simulations.


(5) - Syntèses et préconisations

Conclusions and implications:
In this paper, the authors presented model simulations of climate change impact (scenarios A2 and B2) on snow and runoff in two Alpine catchments. The two catchments, Dischma and Inn, have different sizes (43 km² and 1945 km², respectively) and different characteristics. Both represent rather dry inner-Alpine conditions with the Inn catchment being very dry and the Dischma catchment being climatologically at the transition between the wetter north and the dryer Inn valley.

It has been shown that even with very different catchment sizes, the response of the runoff and snow cover to the climate change scenarios is qualitatively similar. On the other hand, the distribution of altitudes in a given catchment has a strong quantitative influence on the runoff: catchments whose elevation range is mostly centred around the current snow line are experiencing the most changes in these scenarios. In this context, the authors also looked at the sensitivity of snow cover duration to an increase in temperature, similar to what Hantel and Hirtl-Wielke (2007) define as sensitivity of the Alpine snow cover (not shown). For the two catchments investigated no clear trend (or a clear local maximum) of this sensitivity could be found, probably because of two reasons: (i) the joint distribution of altitude and exposition will mask such a trend for the limited size of catchment investigated here; and (ii) the strong changes in the two scenarios will lead to a shift of the zone of maximum sensitivity as described by Hantel and Hirtl-Wielke (2007) over several altitude bands. This sensitivity will however be further investigated in future.

The most prominent finding is that the mountains in the southeast of Switzerland would no longer support a permanent snow cover and that glaciers would therefore disappear if the model assumptions made by the authors are justified. The changes in snow cover would generally be significant and comparable to a shift in altitude of up to 900 m. The runoff would also be heavily affected. The current regime of having a significant part of the annual runoff shaped by snow melt would change to a regime where snow melt would occur during a short time in late spring, producing a large but short runoff peak. This new regime may be problematic regarding water resource management in the dry inner-Alpine valleys and may also increase the risk of flooding from rain on snow events.

In the future, the authors plan to eliminate known weaknesses of Alpine3D, such as the errors introduced by spatial interpolation of the meteorological fields and the bias towards too much snow on steep slopes in high altitudes. We will also use Alpine3D to investigate individual components of the hydrological balance as a function of the climate change scenarios. We, therefore, plan to be more rigorous about the stochastical climate generation for the local catchments, and to look at individual drivers of climate change such as changes in snowline, changes in melt and changes in precipitation. Additionally, we will aim to understand the changes in soil moisture and evaporation. Not much work has been done on evaporation from snow surfaces, and the changes in snow cover will also result in changes of evaporation patterns.

Références citées :

Bartelt PB, Lehning M. 2002. A physical SNOWPACK model for avalanche warning services. Part I: numerical model. Cold Regions Science And Technology 35/3: 123–145.

Christensen JH, Carter TR, Rummukainen M, Amanatidis G. 2007. Evaluating the performance and utility of regional climate models: the PRUDENCE project. Climatic Change 81: 1–6.

Hantel M, Ehrendorfer M, Haslinger A. 2000. Climate sensitivity of snow cover duration in Austria. International Journal of Climatology 20(6): 615–640.

Hantel M, Hirtl-Wielke LM. 2007. Sensitivity of Alpine snow cover to European temperature. International Journal of Climatology 27(10): 1265–1275.

Horton P, Schaefli B, Mezghani A, Hingray B, Musy A. 2006. Assessment of climate-change impacts on alpine discharge regimes with climate model uncertainty. Hydrological Processes 20: 2091–2109. [Fiche biblio]

IPCC. 2007. Climate Change 2007 - The Physical Science Basis, Cambridge University Press, Cambridge.

Laternser M, Schneebeli M. 2003. Long-term snow climate trends of the Swiss Alps (1931–99). International Journal of Climatology 23(7): 733–750. [Fiche biblio]

Lehning M, Bartelt PB, Brown RL, Fierz C, Satyawali P. 2002. A physical SNOWPACK model for the Swiss Avalanche Warning Services. Part III: Meteorological boundary conditions, Thin layer formation and evaluation. Cold Regions Science and Technology 35(3): 169–184.

Lehning M, Bartelt P, Brown RL, Russi T, Stöckli U, Zimmerli M. 1999. Snowpack Model calculations for avalanche warning based upon a new network of weather and snow stations. Cold Regions Science and Technology 30: 145–157.

Lehning M, Löwe H, Ryser M, Raderschall N. 2008. Inhomogeneous precipitation distribution and snow transport in steep terrain. Water Resources Research 44(7): W07404, Doi:10.1029/2007WR006545.

Lehning M, Volksch I, Gustafsson D, Nguyen TA, Stahli M, Zappa M. 2006. ALPINE3D: a detailed model of mountain surface processes and its application to snow hydrology. Hydrological Processes 20: 2111–2128.

López-Moreno JI, Goyette S, Beniston M, Alvera B. 2008. Sensitivity of the snow energy balance to climate change: Implications for the evolution of snowpack in Pyrenees in the 21st century. Climate Research 36(3): 203–217.

Luetschg M, Lehning M, Haeberli W. 2008. A sensitivity study of factors influencing warm/thin permafrost in the Alps. Journal of Glaciology 54: 696–704.

Marty C. 2008. Regime shift of snow days in Switzerland. Geophysical Research Letters 35: L12501, Doi:10.1029/2008GL033998. [Fiche biblio]

Magnusson J, Jonas T, López-Moreno I, Lehning M. 2008. Potential impact of climate change on the snowpack in a partly-glaciated basin in central Switzerland. Sveinsson OGB, Gardarsson SM, Gunnlaugsdottir S (eds). Northern Hydrology and its global role. Proceedings of XXV Nordic Hydrological Conference. Nordic Association for Hydrology, Reykjavik, Iceland, 1: 67–81.

Michlmayr G, Lehning M, Holzmann H, Koboltschnig G, Mott R, Schöner W, Zappa M. 2008. Application of Alpine3D for glacier mass balance and runoff studies at Goldbergkees, Austria. Hydrological Processes 22(19): 3941–3949, Doi:10.1002/hyp.7102/.

Nishimura K, Baba E, Hirashima H, Lehning M. 2005. Application of SNOWPACK model to snow avalanche warning in Niseko, Japan. Cold Regions Science and Technology 43: 62–70.

Rasmus S, Räisänen J, Lehning M. 2004. Estimating snow conditions in Finland in the late 21st century using the SNOWPACK model with regional climate scenario data as input. Annals of Glaciology 38: 238–244.

Singh P, Kumar N. 1997. Impact assessment of climate change on the hydrological response of a snow and glacier melt runoff-dominated Himalayan river. Journal of Hydrology 193: 316–350.

Scherrer SC, Appenzeller C, Laternser M. 2004. Trends in Swiss Alpine snow days: The role of local- and large-scale climate variability. Geophysical Research Letters 31(13): L13215.

Stahl K, Moore RD, Shea JM, Hutchinson D, Cannon AJ. 2008. Coupled modeling of glacier and streamflow response to future climate scenarios. Water Resources Research 44(2): W02422, Doi:10.1029/2007WR005956.

Zappa M, Pos F, Strasser U, Warmerdam P, Gurtz J. 2003. Seasonal water balance of an Alpine catchment as evaluated by different methods for spatially distributed snowmelt modelling. Nordic Hydrology 34: 179–202.

Zemp M, Haeberli W, Hoelzle M, Paul F. 2006. Alpine glaciers to disappear within decades? Geophysical Research Letters 33: 4. [Fiche biblio]