EGU Journal Logos (RGB) Open Access Open Access Open Access Advances in Annales Nonlinear Processes Geosciences Geophysicae in Geophysics Open Access Open Access Nat. Hazards Earth Syst. Sci., 13, 263–277, 2013 Natural Hazards Natural Hazards www.nat-hazards-earth-syst-sci.net/13/263/2013/ doi:10.5194/nhess-13-263-2013 and Earth System and Earth System © Author(s) 2013. CC Attribution 3.0 License. Sciences Sciences Discussions Open Access Open Access Atmospheric Atmospheric Chemistry Chemistry and Physics and Physics Simulating future precipitation extremes in a complex Discussions Open Access Open Access Alpine catchment Atmospheric Atmospheric Measurement Measurement C. Dobler1,2, G. Burger¨ 3,4, and J. Stotter¨ 1,2 Techniques Techniques 1 Institute of Geography, University of Innsbruck, Innrain 52, Innsbruck, Austria Discussions 2 Open Access alpS – Centre for Climate Change Adaptation Technologies, Grabenweg 68, Innsbruck, Austria Open Access 3Pacific Climate Impact Consortium, University of Victoria, 2489 Sinclair Road, Victoria, Canada 4 Institute of Earth and Environmental Science, University of Potsdam, Karl-Liebknecht-Str.Biogeosciences 24-25, Potsdam-Golm, Germany Biogeosciences Discussions Correspondence to: C. Dobler ([email protected]) Received: 27 October 2011 – Published in Nat. Hazards Earth Syst. Sci. Discuss.: – Open Access Open Access Revised: 30 November 2012 – Accepted: 10 December 2012 – Published: 8 February 2013 Climate Climate of the Past of the Past Abstract. The objectives of the present investigation are 1 Introduction Discussions (i) to study the effects of climate change on precipitation ex- tremes and (ii) to assess the uncertainty in the climate pro- Open Access Open Access jections. The investigation is performed on the Lech catch- Global warming may cause an increase of the atmospheric Earth System ment, located in the Northern Limestone Alps. In order to water vapor content andEarth an intensification System of the global hydrological cycle (Solomon et al., 2007; O’Gorman and estimate the uncertainty in the climate projections, two statis- Dynamics Dynamics tical downscaling models as well as a number of global and Schneider, 2009). Precipitation extremes may increase in fre- Discussions regional climate models were considered. The downscaling quency and intensity over many areas of the globe (Sun et al., models applied are the Expanded Downscaling (XDS) tech- 2007; Allan and Soden, 2008), with substantial consequences Open Access nique and the Long Ashton Research Station Weather Gen- for a variety of socio-economicGeoscientific systems (e.g. Easterling et al., Geoscientific Open Access erator (LARS-WG). The XDS model, which is driven by an- 2000; Diffenbaugh et al.,Instrumentation 2005). Instrumentation In the Alps, precipitation is among the major controlling alyzed or simulated large-scale synoptic fields, has been cal- Methods and Methods and ibrated using ECMWF-interim reanalysis data and local sta- meteorological variables for human–environment systems. tion data. LARS-WG is controlled through stochastic param- Through its triggering effect,Data precipitation Systems may be seen as Data Systems eters representing local precipitation variability, which are the key variable for specific natural hazard processes, i.e. for Discussions Open Access flash floods (e.g. Frei and Schar,¨ 1998; Beniston,Open Access 2007), de- calibrated from station data only. Changes in precipitation Geoscientific mean and variability as simulated by climate models were bris flow (e.g. Chiarle et al.,Geoscientific 2007; Szymczak et al., 2010), then used to perturb the parameters of LARS-WG in order to landslides (e.g. Raetzo et al., 2002; Crosta et al., 2004), Model Development hail (e.g. Vinet,Model 2001) andDevelopment avalanches (e.g. Martin et al., generate climate change scenarios. In our study we use cli- Discussions mate simulations based on the A1B emission scenario. The 2001). In the period from 1982 to 2005, natural hazard pro- results show that both downscaling models perform well in cesses caused economic losses in the range of C57 billion Open Access Open Access reproducing observed precipitation extremes. In general, the in the Alps (Agrawala,Hydrology 2007). Potential and future changes in Hydrology and results demonstrate that the projections are highly variable. the frequency and magnitude of precipitation extremes may The choice of both the GCM and the downscaling method are have serious impacts on ecological,Earth System economic and sociolog- Earth System found to be essential sources of uncertainty. For spring and ical systems. Consequently, studyingSciences the effects of climate Sciences autumn, a slight tendency toward an increase in the intensity change on precipitation extremes is of high societal and eco- nomic relevance. Discussions Open Access of future precipitation extremes is obtained, as a number of Open Access simulations show statistically significant increases in the in- Despite the high significance of this topic for the Alps, in- tensity of 90th and 99th percentiles of precipitation on wet vestigations on climate change impacts on precipitation ex- Ocean Science tremes have been veryOcean limited soScience far. Beniston (2006) re- days as well as the 5- and 20-yr return values. Discussions ported a considerable increase in the frequency of heavy Open Access Published by Copernicus Publications on behalf of the European Geosciences Union. Open Access Solid Earth Solid Earth Discussions Open Access Open Access The Cryosphere The Cryosphere Discussions 264 C. Dobler et al.: Simulating future precipitation extremes in a complex Alpine catchment precipitation events in parts of Switzerland during autumn local climate processes (Engen-Skaugen, 2007). Statistical and winter, while Smiatek et al. (2009) found an increase downscaling is then the only way to generate higher- in the frequency of high precipitation amounts for all over resolution climate change scenarios and is thus, particularly the Alps in winter only. According to Frei et al. (2006) and important for the Alps. Schmidli et al. (2007), precipitation extremes are projected Despite the fact that a number of different statistical down- to increase north of about 45◦ N in winter, whereas there is scaling approaches exist, only a few techniques are reported an insignificant change or a decrease south of it. However, to downscale extreme events reliably (e.g. Fowler et al., most of these studies have only used one General Circula- 2007; Tryhorn and DeGaetano, 2011). Modeling extreme tion Model (GCM) and thus, the results only cover a small events is known to be a difficult challenge, as these phenom- range of possible changes. Furthermore, most of these inves- ena lie at the margins of the distribution functions and are tigations were performed on a scale compatible with the grid often beyond the range of calibration data sets (Harpham and spacing of Regional Climate Models (RCMs), and thus, the Wilby, 2005; Tolika et al., 2008; Benestad, 2010). So far only obtained results can only give an incomplete picture of pos- a few attempts have been carried out to compare different sible changes at local scales. But, as changes in precipitation statistical downscaling techniques with a focus on extreme are expected to vary significantly on small horizontal scales events. within complex regions like the Alps (Solomon et al., 2007), Burger¨ and Chen (2005) compared three regression-based investigations on more detailed scales are very relevant. statistical downscaling techniques: randomization, inflation GCMs are the only physically based tools to assess and Expanded Downscaling (XDS). The obtained results changes in climate resulting from increasing atmospheric were quite diverse, highlighting that the choice of the down- greenhouse gases in the atmosphere. The models perform scaling approach is a considerable source of uncertainty. well in reproducing the climate on a global to continental Burger¨ et al. (2012) compared five statistical downscaling scale. However, the horizontal resolution of GCMs is too methods in simulating climate extremes. The methods con- coarse for investigating processes on regional or even local sidered are: automated regression-based statistical down- scales. In recent years, a variety of different techniques have scaling, bias correction spatial disaggregation, quantile re- been developed to bridge this scaling gap. The methods can gression neural networks, a weather generator (TreeGen) be divided into (i) dynamical downscaling and (ii) statistical and XDS. The XDS method was found to perform best, downscaling (Fowler et al., 2007). A comprehensive review followed by the bias correction and spatial disaggregation is given by Maraun et al. (2010). and quantile regression neural networks methods. Liu et Dynamical downscaling is based on highly resolved al. (2011) compared the nonhomogeneous hidden Markov numerical computer models (Regional Climate Models – model and the statistical downscaling model SDSM in terms RCMs), nested into a GCM over a limited region of interest. of downscaling precipitation. Both models performed simi- The higher horizontal resolution of RCMs, typically 25 km lar in simulating dry- and wet-spell length, while the non- or 50 km, captures regional climate processes much better. homogeneous hidden Markov model showed better skill in Statistical downscaling, instead, establishes an empirical re- modeling the wet-day precipitation amount. Hundecha and lationship between large-scale