Réf. Birsan & al. 2005 - A

Référence bibliographique complète

BIRSAN, M. V., MOLNAR, P., BURLANDO, P., PFAUNDLER, M. 2005. Streamflow trends in Switzerland. Journal of Hydrology, 314, 312-329, doi:10.1016/j.jhydrol.2005.06.008

Abstract: Mean daily streamflow records from 48 watersheds in Switzerland with an undisturbed runoff regime are analysed for trends with the Mann–Kendall nonparametric test in three study periods (1931–2000, 1961–2000, 1971–2000). The statistical significance of trends is tested for each station on an annual and seasonal basis and for different streamflow quantiles. The field significance of trends is tested by a bootstrap procedure. Identified trends in streamflow are examined together with changes in precipitation and air temperature, and correlated with watershed attributes. Complex changes in the streamflow regime in Switzerland especially in the more recent periods are demonstrated. The main identified trends are an increase in annual runoff due to increases in the winter, spring and autumn season runoff, an increase in winter maximum streamflow (at more than 60% of the stations) and an increase in spring and autumn moderate and low flows. The behaviour in the summer period is different, with both upward and downward trends present in moderate and low flow quantiles. Many of the trends are field significant. Changes in precipitation are not sufficient to explain the observed trends in streamflow. Air temperature, most notably a substantial increase in the number of days with minimum daily temperature above 0 °C, may explain some of the observed increases in winter and spring season runoff. Correlation analyses reveal a strong relationship between streamflow trends and mean basin elevation, glacier and rock coverage (positive), and basin mean soil depth (negative). These relationships suggest that the most vulnerable environments from the point of view of streamflow change are mountain basins.

Mots-clés
Streamflow; Trend analysis; Watershed properties; Switzerland

Organismes / Contact

• Institute of Hydromechanics and Water Resources Management, ETH Zurich, Switzerland - birsan@ihw.baug.ethz.ch ; molnar@ihw.baug.ethz.ch ; burlando@ihw.baug.ethz. ch
• Federal Office for Water and Geology, Bern-Ittigen, Switzerland - martin.pfaundler@bwg.admin.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, Precipitation River (runoff regime)    

Pays / Zone
Massif / Secteur
Site(s) d'étude
Exposition
Altitude
Période(s) d'observation
Switzerland   48 watersheds     1931-2000

(1) - Modifications des paramètres atmosphériques
Reconstitutions
 
Observations

In the Alpine region of Switzerland there is general agreement that precipitation (especially wintertime precipitation) and air temperature have increased in the last century (e.g., Beniston et al., 1994; Widmann and Schär, 1997; Frei and Schär, 2001). It is however unclear how these changes may have affected streamflow.

Precipitation and air temperature trends:
Trends in the number of wet days are generally upward and field significant for annual, winter and spring records for the period 1931–2000. Only autumn maintains significant trends in the more recent periods. Changes in the precipitation amount are also upward and field significant only for 1931–2000 for the annual and winter series.

There are few significant trends in the precipitation amount in the recent winter seasons, when most significant increasing trends in streamflow are observed. The authors also looked at the behaviour of annual and seasonal precipitation maxima, and found a substantial number of statistically significant trends in winter maximum precipitation only for the 1931–2000 period, not for the more recent ones. Trends in daily precipitation in Switzerland were analysed by Widmann and Schär (1997); Frei and Schär (2001) from the same network of gauges, but for a longer record period (1901–1994). For intense precipitation (average return period 30 days), these studies found significant increasing trends for the winter and autumn seasons, especially in the northern part of Switzerland. It was argued that the observed trends were not due to changing frequencies of weather types, but rather an increase in precipitation activity within a particular weather type (Widmann and Schär, 1997). However, for extreme precipitation (with an average return period of 365 days) observed trends were generally not strong (Frei and Schär, 2001).

In the analysis the authors focussed on trends in annual and seasonal temperature extremes extracted from daily minimum tmin, maximum tmax data, and the diurnal range Δt = tmax – tmin; and the number of days with tmin > 0 °C which is assumed to be a surrogate variable for the probability of precipitation falling as rain. Generally, the results showed an increase in tmin for all seasons; this was accompanied by a decrease in tmax in all seasons except winter. The results varied slightly for the different study periods, but were consistent. As a result, the temperature range Δt decreased significantly at most stations in all study periods, despite a general warming in the region. Similar observations have been made by Beniston et al. (1994) for four stations in Switzerland; and others in different regions of the world (e.g., Michaels et al., 1998; Ventura et al., 2002).

Most striking were the results of trend analyses for the number of days with tmin > 0 °C. Statistically significant increasing trends were dominant on an annual basis and in all seasons for all study periods (the summer season results are to be taken with caution because only very few days have tmin < 0 °C). The trend results were especially consistent for the winter and spring seasons. Trends in the number of days with tmin > 0 °C are field significant for all three study periods for the annual and spring data series. Field significant results were equally found in winter for 1931–2000 and 1961–2000, and in autumn for the 1931–2000 period. The results suggest that generally more precipitation may have been falling as rain in the winter and spring seasons.

Modélisations
 
Hypothèses
 

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

Changes in precipitation and air temperature have been the primary focus of the climate research community. It has been found that trends in observed daily precipitation are generally a complex function of the climatic environment, precipitation intensity and season (e.g., Karl and Knight, 1998; Osborn et al., 2000; Brunetti et al., 2001; Ventura et al., 2002).

Changes in precipitation and air temperature presumably should be able to explain some of the observed trends in streamflow. Precipitation is an intermittent process with a compound distribution, i.e. a nonzero probability of no precipitation on any given day and a continuous distribution on days with measured precipitation (wet days). For this reason the authors looked not only at trends in different quantiles of the precipitation distribution, but also at changes in the number of wet days (defined here as days having at least 0.1 mm of measured precipitation) and the total amount of precipitation. To support direct comparisons with streamflow, the analyses were conducted for the same three study periods [see below]. Basin mean annual precipitation and 1-day maximum precipitation were interpolated from data provided by MeteoSwiss. The number of statistically significant trends in annual and seasonal number of days with precipitation and the precipitation amount are shown.


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

Streamflow trends:
The results show that increasing trends in annual streamflow in Switzerland are dominant for all study periods. It is evident that changes in runoff are not distributed evenly between seasons. The observed increases in annual streamflow come primarily from winter and spring season runoff (also autumn runoff for the period 1961–2000). Interestingly, the summer season shows a different behaviour, with both significant upward and downward trends present in the data, especially in the 1961–2000 period.

The runoff contributions of the different seasons to the annual total vary with region. For regions with an Alpine influence spring and summer are the most important seasons, because they provide most of the runoff volume due to snow and ice melt and summer rainfall. Winter runoff is an important contributor in the Midland-Jura region in the North of Switzerland, while a secondary autumn runoff peak can be observed in the Southern Alpine region influenced by the Mediterranean climate (e.g., Aschwanden and Weingartner, 1985).

Most insightful are the results of trend analyses in streamflow quantiles on a seasonal basis, which highlight the striking difference in trend behaviour between seasons. Winter runoff shows generally increasing trends in all quantiles, most importantly in winter maxima (at least 63% of the analysed stations show statistically significant trends in winter maxima for all study periods). Spring and autumn runoff also shows general increasing trends since 1961, but mostly for moderate and low flow quantiles. However, the behaviour of summer runoff differs markedly. There is no consistent trend direction in summer, although decreasing trends dominate slightly. Trends appear to affect moderate and low flow quantiles, rather than summer maxima. Interestingly, in contrast to the other seasons, in summer the period 1931–2000 shows the most consistent change towards a decrease in streamflow (although one has to keep in mind that only 12 stations are analysed in this period).

No coherent regional effects have been found in trend behaviour throughout Switzerland despite the different hydrological regimes to the North and South of the Alpine range. For example, for the period 1971–2000 and for all stations, positive trends in winter streamflow maxima are spread throughout Switzerland and are found in all runoff regime regions. Results for other streamflow quantiles were similar. The field significance of the trend results determined by the bootstrap procedure showed higher pcrit for the average flow quantiles and lower for the extremes (both minimum and maximum). Field significant trends were found for winter maxima and upper flow quantiles in general throughout the three study periods. Field significant trends in spring, summer and autumn were generally identified in the low flow quantiles and not in maxima. These results point to a substantial country-wide change in some aspects of the natural flow regime in Switzerland.

Precipitation and air temperature trends:
Precipitation quantile analyses did not reveal any clear connections with observed trends in streamflow when basins were compared with their closest precipitation gauges. The authors expect this to be due to the large spatial variability in precipitation, the inadequate representativeness of point measurements, and the inherent nonlinear nature of the transformation of precipitation into runoff.

Most notable is that there are few significant trends in the precipitation amount in the recent winter seasons, when most significant increasing trends in streamflow are observed. The results of the present analysis show that trends in precipitation and streamflow in Switzerland in recent periods are not conclusively related, and that changes in winter precipitation measured at sparsely located rain gauges can only partly explain the observed trends in streamflow since 1961.

Air temperature plays a crucial role in the water cycle in Switzerland, because of the impacts it has on the occurrence of snowfall and snowmelt in this mountainous country. Because some of the most substantial changes observed in streamflow occurred in the winter and spring seasons, it is likely that temperature changes have played a role (e.g., Laternser and Schneebeli, 2003; Scherrer et al., 2004). Generally, the analysis showed an increase in tmin for all seasons; this was accompanied by a decrease in tmax in all seasons except winter. The results varied slightly for the different study periods, but were consistent. As a result, the temperature range Δt decreased significantly at most stations in all study periods, despite a general warming in the region.

Most striking were the results of trend analyses for the number of days with tmin > 0 °C. Statistically significant increasing trends were dominant on an annual basis and in all seasons for all study periods (the summer season results are to be taken with caution because only very few days have tmin < 0 °C). The trend results were especially consistent for the winter and spring seasons. Trends in the number of days with tmin > 0 °C are field significant for all three study periods for the annual and spring data series. Field significant results were equally found in winter for 1931–2000 and 1961–2000, and in autumn for the 1931–2000 period. The results suggest that generally more precipitation may have been falling as rain in the winter and spring seasons, which together with more snowmelt due to higher temperatures in spring may explain the observed increases in streamflow. In fact, Scherrer et al. (2004) have shown that recent decreases in low altitude snow cover in Switzerland can be attributed mainly to an increase in air temperature.

Correlation of streamflow trends with basin attributes:
A classification of the basin attributes into three distinct groups resulted from principal component analysis (PCA). The first component contained highest loadings on basin mean altitude, slope, rock coverage, CN, and soil depth. The second component contained highest loadings on basin mean annual precipitation, 1-day maximum precipitation and river density. The third component loading was basin glacier coverage. The PCA reproduced the correlation matrix between the basin attributes perfectly. These results indicate that the primary discriminatory basin properties are related to altitude, then to climatic conditions and glacier coverage. It is important to note that glacier coverage is also highly correlated with basin altitude (r = 0.65), however the compound distribution of this variable warrants a separate classification (note that 57% of the analysed basins have no glacier).

Simple and multiple (stepwise) regression analyses were subsequently conducted with basin attributes as independent variables and streamflow trend results (the MK trend test statistic Z for different quantiles and seasons) as dependent variables. The attributes presenting significant correlations with streamflow trends vary with season and study period. For the periods since 1961 (the 30-year and the 40-year periods) the correlation results are quite similar and significant, for the 70- year period since 1931 there are fewer significant correlations. Most significant individual correlations were found between streamflow trends and altituderelated attributes: basin mean altitude, slope, soil depth, and glacier and rock coverage. Practically no correlation was found with basin area and shape index.

The most interesting individual correlations between streamflow trends and selected basin attributes are shown for the annual, winter, spring and summer seasons and all analysed quantiles for the 1971–2000 period. The results suggest that streamflow trends are generally positively correlated with basin altitude and glacier coverage and negatively correlated with mean soil depth. This means that with increasing basin altitude and glacier coverage, trends in streamflow also generally increased. Conversely, with increasing basin mean soil depth, i.e. with increasing soil water storage capacity of the basins, trends in streamflow generally decreased.

Another important result at the annual timescale is that basin attributes correlate well with streamflow trends for moderate flow quantiles (about 0.3 < q < 0.8). The correlations decrease for both extreme low and high flow conditions, when it appears that factors other than general basin properties play a major role in runoff production.

Mountain basins:
The positive correlation between streamflow trends and basin attributes related to altitude points to the higher vulnerability of mountain basins to changes in precipitation and air temperature. This seems to be supported by the fact that most consistent statistically significant correlations are observed in winter and spring (especially higher streamflow quantiles in spring), when we expect the influence of precipitation and temperature on runoff production in mountain basins to be strongest. It has been shown that temperature changes are particularly significant for basins below 1000 m a.s.l. in Switzerland because the winter mean temperature at this altitude is around 0 °C and even small variations in temperature may determine whether precipitation falls as rain or snow (Scherrer et al., 2004).

Also striking is the strong correlation between streamflow trends and glacier coverage for low and moderate streamflow in the summer, when other basin attributes (even basin mean altitude) show no (or very little) correlation. We expect that this is indicative of summer snow and ice melt on the glaciers, which contributes to steady but low streamflow (e.g., Collins, 1987; Chen and Funk, 1990). Almost all basins that are more than 10% glaciated also exhibit statistically significant increasing trends in summer streamflow, while the basins with less than 10% glacier coverage present downward or no significant trends.

One of the suggestions from this work is that the observed increase in winter, spring and summer streamflow in glaciated basins is due to warmer temperatures, more precipitation, as well as more snow and ice melt. A closer examination was conducted on data from the Rhone Glacier in Southern Switzerland, where the annual mass balance of the glacier was reconstructed by Chen and Funk (1990) for the period 1883–1987, and where we have found statistically significant increases in winter and spring streamflow. On an annual basis, there was a statistically significant correlation between changes in the mass balance of the glacier (in terms of snow water equivalent) and low magnitude streamflow for the period 1961–1987. These low flows are due to winter and spring runoff. The observed gradual decrease in the mass of the Rhone Glacier coincides with a gradual increase in winter and spring low flows. In this case the retreat of the glacier has been related to increased temperatures in the last century (Chen and Funk, 1990).

Altogether the present results suggest that mountain basins are the most vulnerable environments from the point of view of climate change, because of their watershed properties which promote fast runoff and because of their fundamental sensitivity to temperature changes which affect rainfall, snowfall and snowmelt.

Modélisations
 
Hypothèses

Correlation analyses reveal a strong relationship between streamflow trends and mean basin elevation, glacier and rock coverage (positive), and basin mean soil depth (negative). These relationships suggest that the most vulnerable environments from the point of view of streamflow change are mountain basins.


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

Streamflow integrates the influence of atmospheric variables over a watershed. Presumably, if consistent changes are observed in point measurements of precipitation and air temperature, these should also be reflected to some degree in streamflow at a watershed scale. As a spatially integrated variable streamflow is more appealing for detecting regional trends than point measurements of precipitation which is highly variable in space and time. Comprehensive trend analyses of streamflow conducted in the United States and Canada [see references in the study] have shown that, like precipitation, streamflow records show complex behaviour in which trend significance depends on flow magnitude and season. Shifts in the distributions of daily streamflow were observed, with high and low frequency components exhibiting different behaviour in different seasons. The results of these and other studies have shown that analysing trends in annual or monthly streamflow totals can only give a very rough picture of runoff behaviour. In most cases, only detailed examination of high resolution streamflow data can identify the many complex changes that may have occurred in the instrumental records.

There is however one obvious complication in interpreting trends in streamflow data which is the influence of the watershed itself. Land surface and subsurface properties (such as vegetation, drainage density, soil cover, geology, etc.) impact the precipitation-runoff transformation process. The impact of intrinsic watershed properties depends on streamflow magnitude, and may result in the amplification or dampening of any consistent trend in precipitation over time. Furthermore, some properties, such as land use, forest cover, urbanised area, etc., may change over time, adding additional uncertainty to the precipitation-runoff relationship. From this point of view, establishing the connection between watershed properties and observed trends in streamflow is an important task because it provides a picture of the vulnerability of basins to climate change.

It is clear that the study period has an impact on trend identification. It has been noted that runoff records may contain large scale periodic behaviour, and that trend analyses should always be conducted on periods that span one ormultiple full cycles of this process if it exists (e.g., Pekarova et al., 2003). It is clear from the standardised annual streamflow from their stations that all three study periods analysed contain both low and high flow phases. The authors conclude that the trends reported here for even the shortest study period are not due to a low-frequency large-scale behaviour in the data and are representative of changes in the runoff regime.

An attempt was made to investigate possible causes of observed trends in streamflow in an environment which is highly variable in terms of land surface and atmospheric conditions, and where snow and ice melt play an important role in the natural hydrological regime. The authors report the results of data analyses from 48 basins in Switzerland. The goals were: (1) to identify significant trends in observed streamflow data and their occurrence in space and time in the Swiss Alpine region; (2) to analyse the connection between observed changes in streamflow, precipitation and air temperature; and (3) to investigate the correlation between streamflow trends and watershed properties. This study complements previous analyses of mesoscale precipitation variability (Schmidli et al., 2002), trends in observed heavy precipitation (Frei and Schär, 2001), in snow depth and snow cover (Laternser and Schneebeli, 2003; Scherrer et al., 2004), and trends in long term water balance components (Schädler, 1987) in Switzerland.

Analysis methods used are standard. Mean daily streamflow and precipitation data, and maximum and minimum daily air temperature data were analysed for trends with the Mann–Kendall nonparametric trend test. To identify shifts in the distribution of daily streamflow, a range of quantiles on annual and seasonal bases were studied for three different observation periods. In order to discriminate trends from stochastic fluctuations and the influence of serial correlation in the time series, the series presenting positive lag-1 serial correlation after detrending were prewhitened by applying a first order autoregressive filter to the data prior to trend analysis. Observed streamflow trends are presented; then selected precipitation and temperature trends are shown and linked to changes in streamflow and a correlation analysis between watershed attributes and observed streamflow trends is presented with a focus on mountain basins.

Streamflow data used in this study were high quality records of mean daily discharge from 48 gauging stations distributed across Switzerland. The three main criteria for station selection (Pfaundler, 2001) were: (a) no substantial influence by water withdrawals for hydropower or other water-use purposes; (b) spatial independence between station records; and (c) at least 30 years of continuous and complete observations. Other analysed hydroclimatic data included daily precipitation, and daily minimum and maximum air temperature measured at climate stations of the MeteoSwiss network. Basin area, mean altitude, mean slope and a basin shape index were derived from a digital elevation model (DEM) with a 25 m resolution provided by the Federal Office for Topography. River network density was determined for each basin from 1:25,000 topographic maps of Switzerland and not derived automatically from the DEM. Basin mean soil depth and percentage of rock and glacier coverage were derived from the digital soil and geotechnical maps of Switzerland provided by the Federal Statistical Office (Geostat Data). The potential maximum retention of the soils in each basin was translated into a SCS Curve Number (CN) which incorporates landuse and soil characteristics (Pfaundler, 2001). Basin mean annual precipitation and 1-day maximum precipitation were interpolated from data provided by MeteoSwiss. Trend analyses in this study were conducted by the nonparametric Mann–Kendall (MK) test (Helsel and Hirsch, 1992). The causal aspects of identified trends in streamflow were investigated by correlation analyses with precipitation, air temperature and various basin attributes. The basin attributes were first analysed by principal component analysis to identify structure and redundancy in the variables. The nonparametric rankbased Spearman correlation coefficient was used to report the results of correlation analyses. [see details in the study]


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

Altogether the present results suggest that mountain basins are the most vulnerable environments from the point of view of climate change, because of their watershed properties which promote fast runoff and because of their fundamental sensitivity to temperature changes which affect rainfall, snowfall and snowmelt.

Modélisations
 
Hypothèses

Correlation analyses reveal a strong relationship between streamflow trends and mean basin elevation, glacier and rock coverage (positive), and basin mean soil depth (negative). Altogether the present results suggest that mountain basins are the most vulnerable environments from the point of view of climate change, because of their watershed properties which promote fast runoff and because of their fundamental sensitivity to temperature changes which affect rainfall, snowfall and snowmelt. Climate change in mountain basins could therefore result in substantial changes in the runoff regime with subsequent impacts on erosion and flood risk, ecology, water supply and scarcity, etc., and with profound consequences for downstream populations.


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

Open questions:
The results of this study point to several fundamental open questions. Perhaps the most important are whether the trends observed in streamflow data point to a substantial change in the natural streamflow regime in Switzerland, and whether the observed trends in streamflow can be conclusively connected to changes or variability in the climate and/or watershed properties.

The first question arises from the observation that although many trends in streamflow were found to be statistically significant, it is unclear whether they are significant from the water resources point of view and whether they could have been caused by anthropogenic rather than climatic influences. The authors have taken great care to exclude stations where significant anthropogenic influences on the streamflow regime, especially water withdrawal for hydropower purposes, have occurred in the past. However, it is not possible to exclude anthropogenic influences altogether. Furthermore, gradual land use and land cover changes, glacier retreat, etc., have occurred in some of the analysed basins in the past and may have influenced the streamflow regime. For example, Schädler (1987) has argued that an increase in evaporation has occurred in some regions of Switzerland due to human factors involving intensive agriculture and land use. The impacts of changes in the glaciated area of a basin on runoff from snow and ice melt are conditioned on the finite size of the glaciers. These and other impacts have been impossible to account for in our analysis. From the water management and flood control perspective any persisting trends in streamflow are important, especially those affecting extreme events. In this light, the observed increase in winter extreme events is most important. However, whether this trend (and others) will continue into the future is dependent on their causes, which are not yet fully understood.

The second question is motivated by large scale connections between streamflow and climate variability. A relationship has been observed between streamflow trends and the North Atlantic Oscillation (NAO), which is an index of a large scale climate anomaly in the northern Atlantic affecting the strength of westerly flow and weather patterns in Europe in particular in winter (e.g., Hurrell, 1995; Rodwell et al., 1999; Wanner et al., 2001). The authors of the present study found high correlations between summer streamflow and the NAO index of the previous winter season. Schmidli et al. (2001) observed strong correlations between the NAO and precipitation for the region also only for the winter season. As the NAO index is a fairly reliable predictor of large scale moisture and energy flow into central Europe we can speculate that winter precipitation and its storage in the snow cover affect streamflow in the subsequent spring and summer melting season. However, large scale climate anomalies and their impacts on water resources in the Alpine region remain an open question and an important avenue for future research (e.g., Beniston et al., 1994).

The present results show some connections between observed streamflow trends and changes in atmospheric variables, but the authors do not claim that these completely explain the variability in streamflow. What remains unexplained are trends in streamflow due to continuous changes in other watershed properties, such as land cover and land use, natural and anthropogenic changes to the fluvial system, and others. Although they selected watershed to be as much undisturbed as possible, and they expect that most of the (natural) watershed changes occur on much longer timescales than those studied here, they recognise that they do contribute to hydrological variability.


(5) - Syntèses et préconisations

Conclusions:
This study presents a statistical analysis of trends in mean daily streamflow records from 48 watersheds in Switzerland with an undisturbed runoff regime for three study periods (1931–2000, 1961–2000 and 1971–2000). Statistically significant trends were identified for each station on an annual and seasonal basis and for different streamflow quantiles. The field significance of trends was analysed by a Monte Carlo bootstrap procedure. Identified trends in streamflow were related to observed changes in precipitation and air temperature, and correlated with watershed attributes. The main conclusions are as follows.

(1) A general increase in annual streamflow has been observed, mostly due to increases in winter, spring and autumn runoff. Most dominant changes have occurred in the winter season. Winter streamflow has increased over the whole distribution, but especially markedly for maximum flows: at least 63% of the basins show a statistically significant increase in winter maxima in all periods. On the other hand, increases in spring and autumn runoff are concentrated mostly in moderate and low flow quantile ranges. Trend behaviour in the summer season is different. Both decreasing and increasing trends are found concentrated in moderate and low flow quantiles rather than summer maxima. Summer season behaviour is important because summer runoff provides most water on an annual basis in basins with an Alpine influence. Most of the trends in winter high flows, and spring, summer and autumn moderate and low flows were field significant. However, significant trends were spread across the country and found in all of the different runoff regime regions.

(2) Trends in daily precipitation in the same study periods were generally not as significant as those in streamflow. Streamflow changes could not be explained on the basis of changes in precipitation alone, especially for the period after 1961. Trends in air temperature point towards a general increase in minimum daily temperatures and a decrease in the maximum daily temperatures, leading to a significant decrease in the temperature range. Perhaps most importantly, strong increasing trends appear in the number of days with minimum daily temperature greater than 0 °C (up to 50% of all stations in some cases), which are concentrated in the winter and spring seasons. The authors speculate that the observed increases in winter runoff are due to a shift of snowfall into rainfall. Similarly, low and moderate flow increases in the spring could be explained in part by increased and earlier snow melt due to the observed air temperature rise.

(3) Correlation analyses of streamflow trends with basin attributes show statistically significant relationships between streamflow trends and mean basin elevation, glacier and rock coverage (positive), and basin mean soil depth (negative). The correlations are generally strongest for the moderate flow ranges, and decrease for extreme flows. For extreme flows it appears that factors other than general basin properties play a key role in runoff production. The relationships between streamflow and watershed attributes suggest that the most vulnerable environments are, not surprisingly, mountain basins. For example, a strong correlation was found between streamflow trends and glacier coverage for low and moderate streamflow in the spring and summer. The authors expect that this is indicative of snow and ice melt in mountain basins which contributes to steady but low streamflow.

The present results show that changes in the streamflow regime in Switzerland since 1931 were quite complex and need to be analysed in terms of seasonal distributions to be better understood. The results also point to the fact that mountain basins are the most vulnerable environments from the point of view of climate change, because of their watershed properties that promote fast runoff and because of their fundamental vulnerability to temperature changes which affect rainfall, snowfall, and snow and ice melt. Future research should be directed at the causal aspects of streamflow change in these environments.

Références citées :

Aschwanden, H., Weingartner, R., 1985. Runoff regimes in Switzerland. Pub. Gewaesserkunde nr. 65, Geographical Institute, University Bern. 238pp. (in German).

Beniston, M., Rebetez, M., Giorgi, F., Marinucci, M.R., 1994. An analysis of regional climate-change in Switzerland. Theor. Appl. Climatol. 49 (3), 135–159.

Brunetti, M., Colacino, M., Maugeri, M., Nanni, T., 2001. Trends in the daily intensity of precipitation in Italy from 1951 to 1996. Int. J. Climatol. 21, 299–316.

Chen, J., Funk, M., 1990. Mass balance of Rhonegletscher during 1882/83-1986/87. J. Glaciol. 36 (123), 199–209.

Collins D.N., 1987. Climatic fluctuations and runoff from glacierised Alpine basins. IAHS Publ. no. 168, pp. 77–89.

Frei, C., Schär, C., 2001. Detection probability of trends in rare events: Theory and application to heavy precipitation in the Alpine region. J. Climate 14, 1568–1584. [Fiche biblio]

Helsel, D.R., Hirsch, R.M., 1992. Statistical Methods in Water Resources. Elsevier, Amsterdam (522 pp.).

Hurrell, J.W., 1995. Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science 269, 676–679.

Karl, T.R., Knight, R.W., 1998. Secular trends of precipitation amount, frequency and intensity in the United States. Bull. Am. Meteor. Soc. 79, 231–241.

Laternser, M., Schneebeli, M., 2003. Long-term snow climate trends of the Swiss Alps (1931-99). Int. J. Climatol. 23, 733–750. [Fiche biblio]

Michaels, P.J., Balling, R.C., Vose, R.S., Knappenberger, P.C., 1998. Analysis of trends in the variability of daily and monthly historical temperature measurements. Climate Res. 10, 27–33.

Osborn, T.J., Hulme, M., Jones, P.D., Basnett, T.A., 2000. Observed trends in the daily intensity of United Kingdom precipitation. Int. J. Climatol. 20, 347–364.

Pekarova, P., Miklanek, P., Pekar, J., 2003. Spatial and temporal runoff oscillation analysis of the main rivers of the world during the 19th–20th centuries. J. Hydrol. 274, 62–79.

Pfaundler, M., 2001. Adapting, analysing and evaluating a flexible index flood regionalisation approach for heterogeneous geographical environments. Institute of Hydromechanics and Water Resources Management, IHW Publ. 008, ETH Zurich (189 pp.).

Rodwell, M.J., Rowell, D.P., Folland, C.K., 1999. Oceanic forcing of the wintertime North Atlantic Oscillation and European climate. Nature 398, 320–323.

Schädler, B., 1987. Long water balance time series in the upper basins of four important rivers in Europe-indicators for climatic changes? IAHS Publ. no. 168, pp. 209–219.

Scherrer, S.C., Appenzeller, C., Laternser, M., 2004. Trends in swiss alpine snow days: the role of local-and large-scale climate variability. Geophys. Res. Lett. 31, L13215. doi:10.1029/ 2004GL020255.

Schmidli, J., Schmutz, C., Frei, C., Wanner, H., Schär, C., 2002. Mesoscale precipitation variability in the region of the European Alps during the 20th century. Int. J. Climatol. 22, 1049–1074. [Fiche biblio]

Ventura, F., 2002. Temperature and precipitation trends in Bologna (Italy) from 1952 to 1999. Atmos. Res. 61, 203–214.

Wanner, H., Brönnimann, S., Casty, C., Gyalistras, D., Luterbacher, J., Schmutz, C., Stephenson, D.B., Xoplaki, E., 2001. North atlantic oscillation—concepts and studies. Surv. Geophys. 22, 321–382.

Widmann, M., Schär, C., 1997. A principal component and longterm trend analysis of daily precipitation in Switzerland. Int. J. Climatol. 17, 1333–1356.