The aim of this study is to investigate potential impacts of climate change on the seasonality of runoff in a mountainous watershed, located in the Austrian Alps. In order to consider the full range of possible climate variation, hypothetical climate change scenarios were used to force a hydrological model to simulate runoff time series for potential future climate conditions. The variation of runoff seasonality is illustrated with a three-dimensional representation of daily discharge data, directional statistics of annual flood peaks and the analysis of seasonal occurrence of runoff peaks. The results show that changes in temperature and precipitation patterns could have considerable effects on seasonal runoff variability in the investigated watershed. Generally, a possible increase in temperature may cause an increase in seasonal variability of runoff. Further, annual flood peaks are projected to occur throughout the entire year in the investigated Alpine watershed, whereas moderate high flows may increase in winter (December–February).

ABBREVIATIONS

     
  • AMS

    Annual maximum series

  •  
  • DJF

    December, January and February

  •  
  • GCM

    General circulation model

  •  
  • JJA

    June, July and August

  •  
  • MAM

    March, April and May

  •  
  • m a.s.l.

    metres above sea level

  •  
  • MMF

    Modal month of flood

  •  
  • R²

    Coefficient of determination

  •  
  • RCM

    Regional climate model

  •  
  • SON

    September, October and November

INTRODUCTION

Climate change is expected to trigger changes in both frequency and intensity of floods in many regions around the world (e.g., Solomon et al. 2007; Dankers & Feyen 2008; Hirabayashi et al. 2008; IPCC 2008,, 2012; EASAC 2013). In snow-dominated mountain watersheds, climate change may substantially affect melt characteristics (e.g., Immerzeel et al. 2010) and thus may lead to changes in runoff seasonality. This may have considerable effects on aspects of human–environmental systems and ecosystems (e.g., Hall et al. 2006; IPCC 2014), such as flood risk management, hydro power generation and the biosphere of aquatic species, respectively.

Usually, investigations of climate change impacts are based on a ‘one-way’ modelling chain consisting of general circulation models (GCMs), downscaling techniques and an impact model, such as a hydrological model (e.g., Bronstert 2004; Graham et al. 2007; Blöschl & Montanari 2010). Recent studies have shown that projections based on such modelling chains are associated with large uncertainties, which arise from different model components (e.g., Wilby & Harris 2006; Chen et al. 2011; Dobler et al. 2012b). Many studies, for example Kay et al. (2009), have already explored the significant uncertainties related to GCMs. Further studies have pointed out that the uncertainties in the projections even increase when the focus rests on the investigation of runoff extremes, such as droughts and floods (e.g., Blöschl & Montanari 2010; Bosshard et al. 2014). Thus, it is very difficult to investigate possible impacts of climate change on runoff extremes, especially in mountainous regions (e.g., Dobler et al. 2012a). Recently, a discussion has been initiated whether the current generation of climate models allow reliable projections, for example, about future runoff peaks (e.g., Blöschl & Montanari 2010; Beven 2011; Brown & Wilby 2012). Furthermore, the application of such modelling chains in complex mountain regions is questionable, as topography is still only poorly resolved in regional climate models (Allamano et al. 2009). Thus, only few attempts have been carried out so far to assess possible impacts of climate change on floods in complex mountain regions, and to the authors' knowledge, the most recent relevant studies dealing with this subject are summarised in the following paragraph.

In their study on climate change impacts on future flood hazard in Europe, Dankers & Feyen (2008) highlight both an increase in the 100-year return period of river discharge in spring and a decrease in summer for the Alps. Allamano et al. (2009) showed that large floods in mountain basins have recently become more frequent and may become even more frequent under global warming. In their assessment on possible impacts of climate change on floods in the Alpine Lech watershed, Dobler et al. (2012a) on the one hand obtained no clear signals for extreme flood events (i.e., those with a return period above 10 years), while on the other hand, they found a significant extension of the potential flood period. Köplin et al. (2013) investigated changes in seasonality and magnitude of floods in Switzerland under future climate conditions. They found pronounced changes in flood seasonality in catchments where snow accumulation and melting is important. Furthermore, they showed that the magnitude of mean annual floods and maximum floods may increase in the future.

The aim of this study is to contribute to a better understanding of the complex relationship between climate change and the seasonality of runoff in an Alpine basin. Instead of using a scenario-led strategy based on climate models and downscaling techniques (Wilby & Dessai 2010), an alternative approach is applied. Simple sensitivity analyses of basin responses to a range of hypothetical climate change scenarios are carried out (Prudhomme et al. 2010). Although such scenario-neutral approaches have been increasingly applied during the last years (e.g., Jain et al. 2010; Prudhomme et al. 2010; Brown & Wilby 2012), they are still relatively rare compared to scenario-led approaches.

The Lech watershed, located in the western part of the Austrian Alps, was selected as the investigation area. This watershed is particularly suitable for climate change impact studies as it is a typical Alpine watershed with a relative high amount of annual precipitation (i.e., due to topographic effects), the river Lech is practically free from anthropogenic influence (e.g., no substantial influence due to hydropower operation) and the density of weather station is high enough to conduct a climate change impact study. In such Alpine catchments, flood events typically occur during spring and summer months, which indicate the existence of snow and glacier melt floods (Merz et al. 1999; Merz & Blöschl 2003; Parajka et al. 2009). Although Merz & Blöschl (2003) identified long-rain as the dominant process for the generation of annual maximum peak flows in Austria, snow-related events (combination of rain-on-snow and snowmelt) are important in the Lech watershed, too. According to Merz & Blöschl (2003), 33% of the annual maximum series (AMS) at the gauge Lechaschau were caused by these processes during the last years. Recent studies (e.g., Barnett et al. 2005; Horton et al. 2006; Köplin et al. 2013) indicate that under climate change conditions in snow-dominant mountain regions and mid- to higher latitudes, the hydrological cycle will experience pronounced changes, with shifts in the runoff seasonality. Thus, the Lech watershed is especially appropriate to investigate climate change impacts on the seasonality of runoff.

In recent years, a number of studies have been carried out in the Lech watershed, focussing on climate change impacts on the runoff regime (Dobler et al. 2010), flood hazard potential (Dobler et al. 2012a) and precipitation extremes (Dobler et al. 2013). However, due to the deficiencies of climate models and downscaling techniques, the results obtained from these studies are limited and subject to large uncertainties. The application of a scenario-neutral approach may help to better understand the complex relationship between climate change and the seasonality of runoff regimes in general and of floods in particular in the investigated Alpine watershed. Furthermore, this approach may help to identify critical threshold values of temperature and precipitation changes, which may lead to substantial changes in runoff behaviour.

First, the paper introduces the study area, second, gives an overview of the applied methods, followed by results and discussion, and finally draws conclusions.

STUDY AREA

Situated in the northern limestone Alps in Austria, the investigated watershed is drained by the River Lech, a tributary of the River Danube (see Figure 1). The watershed upstream of the gauge Lechaschau covers around 1,000 km² with elevations ranging from around 800 metres above sea level (m a.s.l.) at the outlet to around 3,000 m a.s.l. The average elevation is around 1,700 m a.s.l. More details of the study area can be found in Dobler et al. (2010).

Figure 1

Overview of the study area.

Figure 1

Overview of the study area.

 Figure 2 gives an overview of temperature and precipitation characteristics of the Lech basin, based on data of the weather station Holzgau (see Figure 1; located at an elevation of 1,080 m a.s.l.) for the period from 1971 to 2005. The mean annual temperature is +6.1 °C, the monthly minimum temperature of −3.5 °C occurs in January, whereas the monthly maximum temperature of +15.2 °C is observed in July. The mean annual area-averaged precipitation is ∼1,500 mm, the maximum monthly area-averaged rainfall is ∼190 mm in July and the minimum monthly area-averaged precipitation of ∼90 mm is observed in April.

Figure 2

Mean monthly temperature (weather station Holzgau – located at an elevation of 1,080 m a.s.l., and mean monthly area-averaged precipitation, based on the period from 1971 to 2005.

Figure 2

Mean monthly temperature (weather station Holzgau – located at an elevation of 1,080 m a.s.l., and mean monthly area-averaged precipitation, based on the period from 1971 to 2005.

METHODS

Hydrological model HQsim

The hydrological model HQsim (Kleindienst 1996) was selected for this study. HQsim can be categorised as a conceptual, semi-distributive hydrological model. The model has already been applied in a number of Alpine watersheds and is currently used for flood forecasting of the River Inn in Tyrol, Austria (Achleitner et al. 2012). The simulation of snowmelt, which is a key process during winter and spring, is based on a modified degree day approach. The current vegetation cover, a seasonal correction parameter and a radiation factor are used to calculate the degree day factor. Dobler & Pappenberger (2012) report that the snowmelt parameters reveal high importance in the investigated watershed and confirm that snowmelt is simulated well by the model. Further details of the model are given by Dobler & Pappenberger (2012).

The HQsim model has been calibrated and validated based on discharge measurements of the period from 1971 to 2005. The time period from 1981 to 2000 was selected for calibrating the model, while the time periods from 1971 to 1980 and 2001 to 2005 were used to validate the model. The model was run in daily time steps, with daily temperature and precipitation input data of the stations shown in Figure 1. In this study, a model setup is used as presented by Dobler et al. (2012b).

Hypothetical climate change scenarios

Although there have been great advances in downscaling climate data from GCMs over the past decades, uncertainties are still very large when generating high-resolution climate scenarios in mountain regions. Hence, instead of following traditional approaches based on downscaling climate model data, simple hypothetical scenarios were used to explore the link between climate changes and changes of floods seasonality in an Alpine watershed. Similar approaches have been applied by Prudhomme et al. (2010) and Islam et al. (2012).

A range of hypothetical climate scenarios is considered with temperature changes varying from +0.5 to +7 °C with an increment of 0.5 °C and precipitation changes varying from −25% to +25% with an increment of 5%. The temperature and precipitation scenarios cover a reasonable range of possibilities and are based on the output of different GCMs for the study area (Girvetz et al. 2009). In total, 154 hypothetical climate scenarios were considered in this investigation.

In a next step, observed temperature and precipitation data of the period from 1971 to 2005 were perturbed by adding temperature data to the historic time series and by multiplying them with the precipitation factors. Note that the delta change approach does not alter the wet- and dry-day frequency in the climate scenarios and that future precipitation extremes are scaled from observed extremes. This may be insufficient when concentrating on extreme flood events (e.g., Dobler et al. 2012b). Being aware of these limitations of the delta change approach, this study has a focus on moderate flood events (i.e., mean annual floods) only.

Statistical analysis

In order to analyse changes in the seasonality of flows, three methods are used.

Directional statistics, Burn vector

The directional (circular) statistic is a useful approach to analyse the seasonality of runoff data (Magilligan & Graber 1996; Mardia & Jupp 1999). Therefore, each event in a flood series (e.g., AMS) is translated into a location on a unit circle. The mathematical convention stipulates that the start of the year is plotted on the easternmost point of the circle and the months proceed in a counter-clockwise sense (Mardia & Jupp 1999). In several studies (e.g., Bayliss & Jones 1993; Merz et al. 1999; Koutroulis et al. 2010), 1 January is used as the easternmost point of the circle; however, other authors recommend the use of the first day of the hydrological year. In the second case, the results are difficult to compare because of the different convention of how to split the hydrological year depending on the field of investigation (e.g., glaciology, hydrology). Therefore, 1 January is used as the starting point.

Following Bayliss & Jones (1993) and Burn (1997), the dates of the events in a flood series with n entries have to be converted into Julian date (D) where 1 January is day 1 and 31 December is day 365. The Julian date of each event i can be plotted in polar coordinates with angle (in radians) as follows: 
formula
1
Each event can be interpreted as a vector starting from the centre with a direction given by and a unit magnitude. The mean direction, (mean day of flood occurrence), is calculated as the addition of unit vectors as follows: 
formula
2
where 
formula
3
The angle of the mean flood occurrence can be translated into a mean day value by multiplying by 365/360 and can be viewed as a measure of central tendency based on all data points. In addition to the angle, the length of the mean vector can be used as a dimensionless measure for dispersion (variability) similar to the standard deviation in non-circular statistics (Cunderlik et al. 2004). The variability measure is calculated as 
formula
4
where each value close to 0 indicates a high variability of floods and 1 means that all events occur on the same day. The variability measure can be used to classify the seasonality. Following the study of Merz & Blöschl (2003), we defined weak , medium and strong seasonality with , and , respectively.

The pair and defines the Burn vector/index (De Michele & Rosso 2002; Parajka et al. 2009). The mean direction  is supposed to be linked to catchment characteristics such as catchment size, geographical location and flood generating mechanisms (long- and short-rainfall, rain-on-snow and snowmelt events) (De Michele & Rosso 2002). The variability measure also provides information about the presence of a dominant process (e.g., glacier- and snowmelt) which results in high values for and in contrast weak seasonality which indicates the presence of several different processes causing a flow regime (Merz & Blöschl 2003).

Flow regime classification

Generally, hydrological regimes are defined as average seasonal conditions over many years (Gottschalk 2006). Similarly, the flow regime describes the average behaviour of river runoff and reflects climatic and physiographic conditions in a river basin (Krasovskaia & Gottschalk 2002). Over the decades, several authors have proposed classifications of hydrological regimes according to the mean monthly streamflow (e.g., Pardé 1933; Gottschalk et al. 1979; Aschwanden & Weingartner 1985). A general prerequisite for the definition of flow regime is that the average seasonal flow conditions are stable over time and depict the same general pattern from year to year. Snowmelt-driven catchments in cold climate are stable and the flow regime patterns during individual years are less chaotic (Gottschalk 2006). Under climate change conditions, the flow regime might change and turn into another regime type (Krasovskaia & Gottschalk 2002). Therefore, the climate conditions, which cause a variation in flow regime classification, could be an indicator to show the climatic setting where changes in the runoff system occur.

Aschwanden & Weingartner (1985) proposed a classification of Swiss watersheds into 16 regime types ranging from ‘glacial’ to ‘pluvial’. Their classification is used in this study because of the spatial proximity and the similarity of climatic conditions in many parts of Switzerland and the Lech watershed. The flow regimes according to Aschwanden & Weingartner (1985) are determined on the basis of the dimensionless Pardé coefficient (Pardé 1933). For each month i, the Pardé coefficient is defined by 
formula
5
where is the mean monthly discharge in month i and year j, and n is the length of flow record. The maximum Pardé coefficient of the 12 months and the month in which occurs are important parameters to determine the flow regime. For instance, a ‘nival alpine’ regime is a snow-dominated flow regime with  between 2.1 and 2.3, is June and a mean elevation ≥1,550 m a.s.l. For full details about the flow regime classification, see the original publication of Aschwanden & Weingartner (1985).

Monthly and seasonal occurrence of annual flood peaks

The monthly frequency of AMS is represented by rose diagrams (Mardia & Jupp 1999), whereas the month with the largest number of events is the modal month of flood (MMF) (Bayliss & Jones 1993). In addition, the seasonal occurrence was investigated with seasons defined as: winter (December–February), spring (April–May), summer (June–August) and autumn (September–November) and the number of AMS events in each season was counted and displayed. The division of the year into seasons is a good compromise between detailed information and the less specific information provided by the Burn vector.

In recent studies, AMS were used for the investigation of seasonality of floods (e.g., Burn 1997; Castellarin et al. 2001; Parajka et al. 2010; Schneeberger et al. 2012; Köplin et al. 2013). In order to obtain comparable results, the analysis of both seasonality and intensity of high flows is also based on AMS in this study.

Intensity of high flows

In order to analyse the magnitude of moderate flood events, the 25%, 50%, and 75% quantiles of the AMS are used. The visualisation is similar to box plots and provides a concise graphical summary of the data characteristics.

RESULTS AND DISCUSSION

Since results of climate impact studies are associated with a wide range of uncertainties, several authors (e.g., Böhm 2008; Blöschl & Montanari 2010) proposed a differentiation of the results into ‘soft’ and ‘hard facts’ in order to describe the robustness of the results. Accordingly, the following results (excluding calibration and validation of the model) are categorised.

Calibration and validation of the hydrological model

First, the performance of the hydrological model HQsim in simulating flood characteristics is evaluated. Figure 3 illustrates observed and simulated AMS during the period from 1971 to 2005. As can be seen, the model is able to reproduce observed floods fairly well with a coefficient of determination (R²) of approx. 0.87. The deviations between observation and simulation are within reasonable bounds, indicating that the model performs well in the investigated Alpine catchment. Figure 4 illustrates box plots of observed and simulated annual and seasonal maximum series. These results show that the model is able to reproduce observed flood events in this complex Alpine catchment.

Figure 3

Observed and simulated AMS, based on the period from 1971 to 2005.

Figure 3

Observed and simulated AMS, based on the period from 1971 to 2005.

Figure 4

Box plots of observed runoff (left box plots of each group) and reference simulation (right box plots) of AMS and the seasonal maximum series.

Figure 4

Box plots of observed runoff (left box plots of each group) and reference simulation (right box plots) of AMS and the seasonal maximum series.

Three-dimensional representation of river runoff and classification of flow regime

Figure 5 shows a three-dimensional representation (based on Montanari 2012) of the discharge time series at the gauging station Lechaschau (see Figure 1) for the present (Figure 5(a)) and one hypothetical climate scenario (Figure 5(b)). The mean daily discharge divided by the mean annual discharge is plotted on the vertical axis as a function of the year (inter-annual) and the calendar day (intra-annual).

Figure 5

Three-dimensional representation (based on Montanari (2012)) of runoff for present (1971–2005) (a) and potential future (temperature +5 °C) climate conditions (b). The resulting surface is based on daily discharge time series by applying moving average in both horizontal directions and a Gaussian filter in intra-annual direction. The width of the moving average window in the intra-annual direction was of seven values (days) and in inter-annual direction five values.

Figure 5

Three-dimensional representation (based on Montanari (2012)) of runoff for present (1971–2005) (a) and potential future (temperature +5 °C) climate conditions (b). The resulting surface is based on daily discharge time series by applying moving average in both horizontal directions and a Gaussian filter in intra-annual direction. The width of the moving average window in the intra-annual direction was of seven values (days) and in inter-annual direction five values.

The runoff series based on present (1971–2005) climate conditions (Figure 5(a)) show clear seasonal patterns with high runoff occurring between May and September and low runoff during winter months. This reflects a typical hydrological behaviour of an Alpine watershed, which is dominated by snowfall in winter and snowmelt in spring and summer. According to Dobler et al. (2010), the ratio of the area above the snowline (estimated by the 0 °C isotherm) to the entire catchment area is high (above 90%) from November to February, resulting in snow accumulation during winter and melt water runoff during spring and summer. Under warmer temperature conditions, the snowline is expected to rise by about 150 m/°C (Beniston 2003). As a consequence, the proportion of the area above the snowline decreases and more liquid precipitation leads to higher winter runoff. Thus, the seasonal variability of runoff under a warming scenario (e.g., +5 °C, illustrated in Figure 5(b)) is higher and runoff higher than the mean yearly discharge likely occurs throughout the entire year. This result is similar to the findings of Zierl & Bugmann (2005), who investigated five Alpine catchments and showed that the intra-annual amplitude between high and low flows decreases by the end of the 21st century.

In a next step, a classical characterisation of the flow regime by means of the Pardé coefficient (Aschwanden & Weingartner 1985) is used to identify possible changes in the runoff regime in more detail. In the reference period, the flow regime can be viewed as ‘nival alpine’ . With increasing temperature (+2 °C: Pk = 1.69, imax: May), the ‘nival’ character is less pronounced, although the pattern of high runoff during snowmelt in spring is observed. However, the peak occurs early within the year and loses intensity in terms of Pardé coefficient. The flow regimes of hypothetical scenarios with a temperature increase of more than +3.5 °C (Pk = 1.33, imax: July) cannot be regarded as ‘nival’ anymore. It changes the character into a complex runoff system of first order with two peaks: one small peak caused by snow-related processes in spring and a second peak caused by precipitation in late summer/early autumn. Horton et al. (2006) and Farinotti et al. (2012) found an analogous tendency for high Alpine Swiss catchments, i.e., an evolution from ‘glacial’ and ‘glacio-nival’ regime types to ‘nival’ and from ‘nival’ to ‘pluvial’ regime types under future climate conditions.

Generally, the three-dimensional representation of the daily mean runoff shows that under warmer climate conditions, the seasonal variability of the Lech watershed is expected to increase. This can be considered to be a ‘hard fact’, because temperature-driven processes are relatively robust (Blöschl & Montanari 2010). For the analysis of the consequences of these seasonal shifts on the timing of floods, the variation of the mean occurrence date, the monthly frequency, the seasonal occurrence of the annual maximum peak and the intensity of high flows are studied.

Directional statistics, Burn vector

Figure 6(a) shows the two components (mean occurrence date and variability measure ) of the Burn vector (see also schematic drawing in Figure 6(b)) under reference conditions and 154 hypothetical climate change scenarios. Under present climate conditions (reference), the mean occurrence date is at the end of June with a strong seasonality of around 0.8. This indicates the presence of a dominant flood generating process (i.e., snow-related processes). Under moderate temperature scenarios (up to +2 °C), the variability measure can be viewed as ‘strong’, and the mean occurrence remains in a time window of about 0.5 month. Furthermore, up to a temperature increase of around +2 °C, the majority of annual maximum peaks occurs during the summer months. Temperature scenarios above +2 °C show almost no change in the occurrence date; however, the variability increases, indicating that there is no single dominant flood generating process anymore (Merz & Blöschl 2003). Similar to linear statistics, the mean occurrence date as a measure of central tendency loses meaningfulness with increasing variability (decreasing ). Thus, the interpretation of  above temperature scenarios of 5 °C is difficult.

Figure 6

Burn vector: (a) variability measure against the mean occurrence date of annual peaks for climate scenarios; dots with crosses indicate temperature change and no precipitation change; small dots show different precipitation scenarios. (b) Schematic drawing of Burn vector for the temperature scenario +2.5 °C and no precipitation change.

Figure 6

Burn vector: (a) variability measure against the mean occurrence date of annual peaks for climate scenarios; dots with crosses indicate temperature change and no precipitation change; small dots show different precipitation scenarios. (b) Schematic drawing of Burn vector for the temperature scenario +2.5 °C and no precipitation change.

Generally, Figure 6(a) shows that the Burn vector responds sensitively to changes in temperature, whereas possible changes in precipitation patterns have almost no influence on the Burn vector. These findings can be considered to be a ‘hard fact’ and are in consensus with several other studies (e.g., Leung & Wigmosta 1999; Merz & Blöschl 2003; Horton et al. 2006), which point out that seasonal variations of floods and changes of the runoff regime in Alpine catchments are mainly due to changes in temperature.

Monthly frequency of annual flood peaks

Figure 7 provides rose diagrams of the reference simulation (grey shaded) as well as selected temperature and precipitation scenarios (bold lines). Under reference conditions, the annual flood peaks occur from May until August mainly while June is the MMF. A comparison with the modest scenario (+0.5 °C, no precipitation change) shows that the events occur exactly in the same month and that the difference for instance in May and June is only one event per month (one event is approximately equivalent to 3% in the polar histograms). Under a warming scenario of +2.5 °C, the variability of the occurrence date of annual flood peaks increases. Fewer events occur in May and June and more in the cooler/less warm seasons. Scenarios with higher temperature (e.g., +5 °C) show the same tendency, whereas more annual flood peaks occur between December and March. This is a typical response of an Alpine watershed to warmer temperatures and can be explained by earlier snowmelt and an increase of rainfall in winter due to a higher snowline (e.g., Barnett et al. 2005; Dobler et al. 2010). Similar results were obtained by Farinotti et al. (2012), who investigated nine high Alpine Swiss catchments and showed that the fraction of liquid precipitation will increase up to 35% by the end of the century. Moreover, the seasonal variability of runoff increases, i.e., a higher percentage of annual runoff will occur earlier and later within one year.

Figure 7

Rose diagram of the AMS considering the reference simulation (grey shaded) and selected climate scenarios (bold lines). The rows represent precipitation scenarios (1st row: − 25%; 2nd row: no prec. change; 3rd row: +25%) and the column represents the temperature change (1st column: +0.5 °C; 2nd column: +2.5 °C; 3rd column: +5 °C).

Figure 7

Rose diagram of the AMS considering the reference simulation (grey shaded) and selected climate scenarios (bold lines). The rows represent precipitation scenarios (1st row: − 25%; 2nd row: no prec. change; 3rd row: +25%) and the column represents the temperature change (1st column: +0.5 °C; 2nd column: +2.5 °C; 3rd column: +5 °C).

Furthermore, Figure 7 indicates that under warmer temperature conditions, the MMF shifts from June to August and is less pronounced. The analysis of the rose diagrams focussing on extreme precipitation scenarios shows the small effect of precipitation changes on the variability of the monthly occurrence of AMS. The observed tendency toward a more equal distribution of annual flood peaks under warmer temperature condition tends to be a ‘hard fact’.

Seasonal occurrence of annual flood peaks

The results obtained so far indicate that warmer temperatures may lead to substantial changes in the seasonality of annual flood peaks in the Alpine Lech watershed, whereas the impact of precipitation change itself has only minor consequences on the seasonality. Hereinafter, the effect of temperature change on the seasonal occurrence of AMS will be analysed in more detail. Therefore, the occurrence date of the annual flood peaks stratified by seasons was analysed under temperature scenarios (all scenarios from +0.5 to +7 °C were considered) and displayed in Figure 8.

Figure 8

Seasonal occurrence of annual flood peaks under temperature scenarios.

Figure 8

Seasonal occurrence of annual flood peaks under temperature scenarios.

In winter, no annual flood peaks are projected under moderate temperature change (up to +2 °C). When the temperature increase is above +2 °C, annual flood peaks also occur in winter and their number increases further with rising temperature. A contrary development can be observed in summer, when the number of events decreases under scenarios of a temperature increase above +2.5 °C. In spring, the number of annual flood peaks decreases under moderate temperature conditions and remains more or less stable under warmer temperature conditions. In autumn, again an opposite tendency can be observed, as the number of events increases under moderate conditions. The change in the seasonal occurrence of annual flood peaks can be once more explained by the variation of the snowmelt characteristics and the increase of rainfall in winter. For example, in years in which no high flood peaks occur in summer and spring under reference conditions, changing climatic conditions could cause annual flood peaks in winter. In summary, the variability of the occurrence of annual flood peaks increases under warmer temperature conditions. This general trend can be considered to be a ‘hard fact’, whereas the mentioned temperatures that indicate changes in the seasonal occurrence can be seen as ‘soft facts’.

Intensity of high flows

Figure 9 shows the intensity of the annual and seasonal maximum series under climate change conditions. The median (depicted by the horizontal line) and the 25% and 75% quantiles (lower and upper end of the vertical line) of each series are displayed for the reference simulation and selected temperature scenarios (from +1 to +7 °C). Here, impacts of temperature change alone are analysed as well as combined effects of temperature and precipitation changes.

Figure 9

Flood quantiles (horizontal line: 50%; lower end of the vertical line: 25%; and upper end: 75%) of the AMS and the seasonal maximum series for reference simulation and temperature scenarios (+1 to +7 °C). The bold lines (blue) of each temperature interval indicate no precipitation change and the lines on the left (cyan) and right (magenta) side show changes of −25 and +25%. Please refer to the online version of this paper to see this figure in colour: http://www.iwaponline.com/jwc/toc.htm.

Figure 9

Flood quantiles (horizontal line: 50%; lower end of the vertical line: 25%; and upper end: 75%) of the AMS and the seasonal maximum series for reference simulation and temperature scenarios (+1 to +7 °C). The bold lines (blue) of each temperature interval indicate no precipitation change and the lines on the left (cyan) and right (magenta) side show changes of −25 and +25%. Please refer to the online version of this paper to see this figure in colour: http://www.iwaponline.com/jwc/toc.htm.

Figure 9(a) shows that possible increases in temperature with no changes in precipitation may only lead to minor shifts in the intensity of mean floods. Further, the medians of the scenarios with precipitation change are compared with the median of corresponding temperature scenario without precipitation change, and thereby, the −25% and +25% precipitation scenarios cause approximately 60% and 140% runoff, respectively. The variation of approximately 40% (increase and decrease) can be observed for all temperature ranges of the AMS (Figure 9(a)) and the four seasonal series (Figure 9(b)(e)). The comparison of the −25% and +25% precipitation scenarios with the scenarios without precipitation change of the investigated time series and of all temperature ranges shows a maximum variation of 52% and 67% and 132% and 155%, respectively. Since extreme precipitation scenarios of the five investigated time series have similar effect on the variation of runoff, only the impact of temperature change is further discussed.

Figure 9(b) demonstrates that the flood quantiles of winter maximum series show pronounced increasing trends under warmer climate conditions, which can be explained due to earlier snowmelt and an increase in liquid precipitation (similar results were obtained by Farinotti et al. (2012)). In contrast, in summer (Figure 9(d)), decreasing trends of flood quantiles are observed. This can be partly explained by the effect of more evapotranspiration under warmer conditions and therefore drier antecedent conditions in the soil (Wetterhall et al. 2011). Figure 9(c) shows a slight decrease of the spring flood quantiles for warmer temperatures. The monthly changes of the flood quantiles (not shown here) were also analysed and they show that the flood quantiles in March slightly increase, remain constant in April and decrease in May. As displayed in Figure 9(e), the flood quantiles in autumn remain constant under warmer conditions. The change of the intensity of high flows tends to be a ‘soft fact’.

Similar results were achieved by Middelkoop et al. (2001) and Bosshard et al. (2013) (the Rhine basin), Zierl & Bugmann (2005) (five Alpine catchments) and Blaschke et al. (2011) (for the Alpine western part of Austria) who showed that the summer runoff decreases and the winter runoff increases.

CONCLUSION

This study analysed the impact of climate change on runoff conditions in the Lech watershed. Hypothetical climate scenarios with temperature changes varying between +0.5 and +7 °C and precipitation changes varying from −25% to +25% were generated and used to force a hydrological model to simulate future runoff conditions. Accordingly, 154 runoff series were gathered and analysed with particular emphasis on the variation of the seasonality of runoff. The results were categorised into ‘soft’ and ‘hard facts’ to describe their reliability and robustness. These categories refer to the general process behaviour under climate change conditions and not to the specified temperature ranges.

The main conclusions of this study are as follows:

  • Possible changes in temperature could have strong impacts on the seasonality of runoff in the investigated watershed. In contrast, possible changes in precipitation alone may only have low impacts on the seasonality (hard fact).

  • Under warmer temperature conditions, the mean occurrence date of floods and the MMF is expected to move backwards in time (hard fact), i.e., from June under reference condition to end of July /August (MMF) when temperature increase is +2.5 °C and higher.

  • The seasonal variability of AMS increases under warmer climate conditions and the variability measure (dispersion part of the Burn vector) declines (hard fact). Above a temperature increase of +2 °C, annual flood peaks occur throughout the entire year.

  • The intensity of winter floods may increase, while the intensity of summer floods may decrease under warmer climate conditions (soft fact).

  • The evolution of the runoff system under rising temperature conditions suggested that above +2.5 °C, the flow regime changes considerably. This result can be found by analysing the variation of the Burn vector (Figure 6(a)) and the seasonal occurrence of AMS (Figure 8). The first component of the Burn vector (mean occurrence date ) changes up to +2.5 °C and remains almost constant beyond that. The analysis of seasonal occurrence of AMS shows that the number of events in summer and winter season remains almost constant up to +2.5 °C and changes beyond that. In contrast, in spring and autumn, changes take place up to this temperature range and remain roughly constant above (soft fact).

The runoff regime of the River Lech can be categorised as ‘nival alpine’ under current conditions, and consequently, snow accumulation and the release of meltwater are important processes in the watershed. Thus, snow-dominated watersheds tend to be sensitive to temperature change (e.g., Horton et al. 2006; Barnett et al. 2005) under warmer conditions, the seasonal variability of runoff is expected to increase and the flood characteristics change. Under future climate conditions, likely more different flood generating processes (i.e., snow-related processes are less dominant) are responsible for the hydrological response of Alpine watersheds.

ACKNOWLEDGEMENTS

This work results from a research project funded by the Austrian Research Promotion Agency (FFG) within the scope of the programme COMET, the Tyrolean Insurance Company (Tiroler Versicherung) and the Vorarlberger Insurance Company (Vorarlberger Landesversicherung). The required basis data were provided by the Tyrolean State Government. The authors would like to thank the anonymous reviewer for the thorough revision and the valuable comments.

REFERENCES

REFERENCES
Achleitner
S.
Schöber
J.
Rinderer
M.
Leonhardt
G.
Schöberl
F.
Kirnbauer
R.
Schönlaub
H.
2012
Analyzing the operational performance of the hydrological models in an alpine flood forecasting system
.
J. Hydrol.
412–413
,
90
100
.
Allamano
P.
Claps
P.
Laio
F.
2009
Global warming increases flood risk in mountainous areas
.
Geophys. Res. Lett.
36
(
24
),
L24404
.
Aschwanden
H.
Weingartner
R.
1985
The Runoff Regimes in Switzerland (Die Abflussregimes der Schweiz
).
Geogr. Inst. Univ. Bern, Publikation Gewässerkunde 65
,
Bern, Switzerland
(in German)
.
Bayliss
A. C.
Jones
R. C.
1993
Peaks-Over-Threshold Flood Database. Summary Statistics and Seasonality
.
Institute of Hydrology
,
Wallingford, UK
.
Blaschke
A. P.
Merz
R.
Parajka
J.
Salinas
J. L.
Blöschl
G.
2011
Climate impacts on surface and subsurface water resources (Auswirkungen des Klimawandels auf das Wasserdargebot von Grund- und Oberflächenwasser)
.
Österr. Wasser Abfallw
.
63
(
1–2
), 31–41
(in German)
.
Blöschl
G.
Montanari
A.
2010
Climate change impacts – throwing the dice?
Hydrol. Process.
24
(
3
),
374
381
.
Böhm
R.
2008
Hot Air: The Climate Change Controversy; Facts – Fears – Funding (Heiße Luft – Reizwort Klimawandel; Fakten – Ängste – Geschäfte)
.
Edition VaBene
.
Wien, Austria (in German)
.
Bosshard
T.
Carambia
M.
Goergen
K.
Kotlarski
S.
Krahe
P.
Zappa
M.
Schär
C.
2013
Quantifying uncertainty sources in an ensemble of hydrological climate-impact projections
.
Water Resour. Res.
49
(
3
),
1523
1536
.
Brown
C.
Wilby
R. L.
2012
An alternate approach to assessing climate risks
.
Eos Trans. AGU
93
(
41
),
401
412
.
Chen
J.
Brissette
F. P.
Poulin
A.
Leconte
R.
2011
Overall uncertainty study of the hydrological impacts of climate change for a Canadian watershed
.
Water Resour. Res.
47
(
12
),
W12509
.
Cunderlik
J. M.
Ouarda
T. B.
Bobée
B.
2004
Determination of flood seasonality from hydrological records
.
Hydrol. Sci. J.
49
(
3
),
511
526
.
De Michele
C.
Rosso
R.
2002
A multi-level approach to flood frequency regionalisation
.
Hydrol. Earth Syst. Sci.
6
(
2
),
185
194
.
Dobler
C.
Stötter
J.
Schöberl
F.
2010
Assessment of climate change impacts on the hydrology of the Lech Valley in northern Alps
.
J. Water Clim. Change
1
(
3
),
207
218
.
Dobler
C.
Hagemann
S.
Wilby
R. L.
Stötter
J.
2012b
Quantifying different sources of uncertainty in hydrological projections in an Alpine watershed
.
Hydrol. Earth Syst. Sci.
16
(
11
),
4343
4360
.
Dobler
C.
Bürger
G.
Stötter
J.
2013
Simulating future precipitation extremes in a complex Alpine catchment
.
Nat. Hazard. Earth Syst. Sci.
13
(
2
),
263
277
.
EASAC
2013
Trends in Extreme Weather Events in Europe: Implications for National and European Union Adaptation Strategies
,
EASAC. Policy Report 22
,
Halle
. .
Farinotti
D.
Usselmann
S.
Huss
M.
Bauder
A.
Funk
M.
2012
Runoff evolution in the Swiss Alps: projections for selected high-alpine catchments based on ENSEMBLES scenarios
.
Hydrol. Process.
26
(
13
),
1909
1924
.
Girvetz
E. H.
Zganjar
C.
Raber
G. T.
Maurer
E. P.
Kareiva
P.
Lawler
J. J.
Traveset
A.
2009
Applied climate-change analysis: the climate wizard tool
.
PLoS ONE
4
(
12
),
e8320
.
Gottschalk
L.
2006
Methods of analyzing variability
. In:
Encyclopedia of Hydrological Sciences
(
Anderson
M. G.
McDonnell
J. J.
, eds).
John Wiley & Sons, Ltd
,
Chichester, UK
.
Gottschalk
L.
Jensen
J. L.
Lundquist
D.
Solantie
R.
Tollan
A.
1979
Hydrologic regions in the Nordic countries
.
Nordic Hydrol.
10
(
5
),
273
286
.
Hall
J. W.
Sayers
P. B.
Walkden
M. J. A.
Panzeri
M.
2006
Impacts of climate change on coastal flood risk in England and Wales: 2030–2100
.
Philos. Trans. Roy. Soc. A
364
(
1841
),
1027
1049
.
Hirabayashi
Y.
Kanae
S.
Emori
S.
Oki
T.
Kimoto
M.
2008
Global projections of changing risks of floods and droughts in a changing climate
.
Hydrol. Sci. J.
53
(
4
),
754
772
.
Horton
P.
Schaefli
B.
Mezghani
A.
Hingray
B.
Musy
A.
2006
Assessment of climate-change impacts on alpine discharge regimes with climate model uncertainty
.
Hydrol. Process.
20
(
10
),
2091
2109
.
Immerzeel
W. W.
van Beek
L. P. H.
Bierkens
M. F. P.
2010
Climate change will affect the Asian water towers
.
Science
328
(
5984
),
1382
1385
.
IPCC
2008
Climate change and water
. In:
IPCC Technical Paper VI
(
Bates
B. C.
Kundzewicz
Z. W.
Wu
S.
Palutikof
J. P.
, eds).
IPCC Secretariat
,
Geneva, Switzerland
.
IPCC
2012
Managing the risks of extreme events and disasters to advance climate change adaption
. In:
Special Report of the Intergovernmental Panel on Climate Change
(
Field
C. B.
Barros
V.
Stocker
T. F.
Qin
D.
Dokken
D. J.
Ebi
K. L.
Mastrandrea
M. D.
Mach
K. J.
Plattner
G.-K.
Allen
S. K.
Tignor
M.
Midgley
P. M.
, eds).
Cambridge University Press
,
Cambridge, UK and New York, NY, USA
.
IPCC
2014
Summary for policymakers
. In:
Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Field
C. B.
Barros
V. R.
Dokken
D. J.
Mach
K. J.
Mastrandrea
M. D.
Bilir
T. E.
Chatterjee
M.
Ebi
K. L.
Estrada
Y. O.
Genova
R. C.
Girma
B.
Kissel
E. S.
Levy
A. N.
MacCracken
S.
Mastrandrea
P. R.
White
L. L.
, eds).
Cambridge University Press
,
Cambridge, UK and New York, NY, USA
, pp.
1
32
.
Islam
A.
Sikka
A. K.
Saha
B.
Singh
A.
2012
Streamflow response to climate change in the Brahmani River Basin, India
.
Water Resour. Manage.
26
(
6
),
1409
1424
.
Jain
S. K.
Goswami
A.
Saraf
A. K.
2010
Assessment of snowmelt runoff using remote sensing and effect of climate change on runoff
.
Water Resour. Manage.
24
(
9
),
1763
1777
.
Kay
A. L.
Davies
H. N.
Bell
V. A.
Jones
R. G.
2009
Comparison of uncertainty sources for climate change impacts: flood frequency in England
.
Clim. Change
92
(
1–2
),
41
63
.
Kleindienst
H.
1996
Enhancement and Testing of an Application-Oriented Hydrological Model to Simulate Hydrographs in Torrential Catchments (Erweiterung und Erprobung eines anwendungsorientierten hydrologischen Modells zur Gangliniensimulation in kleinen Wildbacheinzugsgebieten)
.
Institute for Geography, Ludwig Maximilian University of Munich
,
Germany
(in German)
.
Köplin
N.
Schädler
B.
Viviroli
D.
Weingartner
R.
2013
Seasonality and magnitude of floods in Switzerland under future climate change
.
Hydrol. Process.
28
(
4
),
2567
2578
.
Koutroulis
A. G.
Tsanis
I. K.
Daliakopoulos
I. N.
2010
Seasonality of floods and their hydrometeorologic characteristics in the island of Crete
.
J. Hydrol.
394
(
1–2
),
90
100
.
Krasovskaia
I.
Gottschalk
L.
2002
River flow regimes in a changing climate
.
Hydrol. Sci. J.
47
(
4
),
597
609
.
Leung
L. R.
Wigmosta
M. S.
1999
Potential climate change impacts on mountain watersheds in the Pacific Northwest
.
J. Am. Water Resour. Assoc.
35
(
6
),
1463
1471
.
Mardia
K. V.
Jupp
P. E.
1999
Statistics of Directional Data
.
Wiley
,
New York, USA
.
Merz
R.
Blöschl
G.
2003
A process typology of regional floods
.
Water Resour. Res.
39
(
12
),
1340
1359
.
Merz
R.
Piock-Ellena
U.
Blöschl
G.
Gutknecht
D.
1999
Seasonality of flood processes in Austria
. In:
Hydrological Extremes: Understanding, Predicting, Mitigating
(
Gottschalk
L.
Olivry
J.-C.
Reed
D.
Rosbjerg
D.
, eds).
IAHS Press
,
Wallingford, UK
,
IAHS Publication 255
, pp.
273
278
.
Middelkoop
H.
Daamen
K.
Gellens
D.
Grabs
W.
Kwadijk
J.
Lang
H.
Parmet
B. W. A. H.
Schadler
B.
Schulla
J.
Wilke
K.
2001
Impact of climate change on hydrological regimes and water resources management in the Rhine Basin
.
Clim. Change
49
(
1–2
),
105
128
.
Parajka
J.
Kohnová
S.
Merz
R.
Szolgay
J.
Hlavčová
K.
Blöschl
G.
2009
Comparative analysis of the seasonality of hydrological characteristics in Slovakia and Austria
.
Hydrol. Sci. J.
54
(
3
),
456
473
.
Parajka
J.
Kohnová
S.
Bálint
G.
Barbuc
M.
Borga
M.
Claps
P.
Cheval
S.
Dumitrescu
A.
Gaume
E.
Hlavcova
K.
Merz
R.
Pfaundler
M.
Stancalie
G.
Szolgay
J.
Blöschl
G.
2010
Seasonal characteristics of flood regimes across the Alpine–Carpathian range
.
J. Hydrol.
394
,
78
89
.
Pardé
M.
1933
Fleuves et Rivières
.
Colin
,
Paris, France
.
Prudhomme
C.
Wilby
R. L.
Crooks
S.
Kay
A. L.
Reynard
N.
2010
Scenario-neutral approach to climate change impact studies: application to flood risk
.
J. Hydrol.
390
(
3–4
),
198
209
.
Schneeberger
K.
Huttenlau
M.
Stötter
J.
2012
A seasonality analysis of flood peaks as basis for flood risk analyses in two alpine regions
. In
3rd STAHY International Workshop on Statistical Methods for Hydrology and Water Resources Management
,
1–2 October 2012
,
Tunis, Tunisia
.
Solomon
S.
Qin
D.
Manning
M.
2007
Technical summary
. In:
The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change
(
Solomon
S.
Qin
D.
Manning
M.
Chen
Z.
Marquis
M.
Averyt
K. B.
Tignor
M.
Miller
H. M.
, eds).
Cambridge University Press
,
Cambridge
.
Wetterhall
F.
Graham
L. P.
Andréasson
J.
Rosberg
J.
Yang
W.
2011
Using ensemble climate projections to assess probabilistic hydrological change in the Nordic region
.
Nat. Hazard. Earth Syst. Sci.
11
(
8
),
2295
2306
.
Wilby
R. L.
Dessai
S.
2010
Robust adaptation to climate change
.
Weather
65
(
7
),
180
185
.
Zierl
B.
Bugmann
H.
2005
Global change impacts on hydrological processes in Alpine catchments
.
Water Resour. Res.
41
(
2
),
W02028
.