The climate of daily precipitation in the Alps
Transcription
The climate of daily precipitation in the Alps
INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. (2013) Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/joc.3794 The climate of daily precipitation in the Alps: development and analysis of a high-resolution grid dataset from pan-Alpine rain-gauge data Francesco A. Isotta,a* Christoph Frei,a Viktor Weilguni,b Melita Perčec Tadić,c Pierre Lassègues,d Bruno Rudolf,e Valentina Pavan,f Carlo Cacciamani,f Gabriele Antolini,f Sara M. Ratto,g Michela Munari,h Stefano Micheletti,i Veronica Bonati,j Cristian Lussana,k Christian Ronchi,l Elvio Panettieri,m Gianni Marigo,n and Gregor Vertačnik,o a b Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland Bundesministerium für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft, Abteilung VII/3—Wasserhaushalt, Vienna, Austria c Meteorological and Hydrological Service of Croatia, Zagreb, Croatia d Direction de la Climatologie Météo-France, Toulouse, France e German Weather Service, Hydrometeorology, Offenbach am Main, Germany f ARPA Emilia-Romagna, Bologna, Italy g Regione Autonoma Valle d’Aosta, Aosta, Italy h Provincia Autonoma di Bolzano, Ufficio Idrografico, Bolzano, Italy i OSMER Regional Meteorological Observatory of ARPA Friuli Venezia Giulia, Visco, Italy j ARPA Liguria, Genoa, Italy k ARPA Lombardia, Milan, Italy l ARPA Piemonte, Turin, Italy m Provincia Autonoma di Trento, Trento, Italy n ARPA Veneto, Arabba, Italy o Slovenian Environment Agency, Meteorology Office, Ljubljana, Slovenia ABSTRACT: In the region of the European Alps, national and regional meteorological services operate rain-gauge networks, which together, constitute one of the densest in situ observation systems in a large-scale high-mountain region. Data from these networks are consistently analyzed, in this study, to develop a pan-Alpine grid dataset and to describe the region’s mesoscale precipitation climate, including the occurrence of heavy precipitation and long dry periods. The analyses are based on a collation of high-resolution rain-gauge data from seven Alpine countries, with 5500 measurements per day on average, spanning the period 1971–2008. The dataset is an update of an earlier version with improved data density and more thorough quality control. The grid dataset has a grid spacing of 5 km, daily time resolution, and was constructed with a distance-angular weighting scheme that integrates climatological precipitation–topography relationships. Scales effectively resolved in the dataset are coarser than the grid spacing and vary in time and space, depending on station density. We quantify the uncertainty of the dataset by cross-validation and in relation to topographic complexity, data density and season. Results indicate that grid point estimates are systematically underestimated (overestimated) at large (small) precipitation intensities, when they are interpreted as point estimates. Our climatological analyses highlight interesting variations in indicators of daily precipitation that deviate from the pattern and course of mean precipitation and illustrate the complex role of topography. The daily Alpine precipitation grid dataset was developed as part of the EU funded EURO4M project and is freely available for scientific use. KEY WORDS Alpine region; mountain climate; spatial analysis; extreme events; Alpine climatology Received 4 April 2013; Revised 27 June 2013; Accepted 2 July 2013 1. Introduction The Alps are one of the major mountain ranges in Europe. Their significance for the regional water cycle is commonly expressed in the notion ‘water tower of Europe’, which hints to their climate with abundant * Correspondence to: F. A. Isotta, Federal Office of Meteorology and Climatology, Krähbühlstrasse 58, Postfach 514, Zurich, Switzerland. E-mail: francesco.isotta@meteoswiss.ch 2013 Royal Meteorological Society precipitation and to the role of snow and ice in storing freshwater and moderating variations in river runoff (e.g. EEA, 2009). Indeed, the Alps are the source region of four major European rivers (the Danube, Rhine, Po and Rhone), to which the head-waters contribute a significant share of the runoff. Near the mouth of river Rhine, for example, almost 50% of the mean discharge originates from the Alpine region, an area only about 20% of the total catchment (Viviroli et al., 2003). Precipitation in the Alpine region contributes freshwater and hydropower, F. A. ISOTTA et al. within and beyond the region; it enables transportation on rivers and shapes the distribution and diversity of ecosystems. But the vigorous precipitation climate of the Alps can also be a threat to life and civil infrastructure. Water-related natural hazards (river floods, flash floods, landslides, debris flows and avalanches) cause loss of life and damages more frequently, and this requires regional planning and prevention (e.g. Greminger, 2003). Accurate knowledge of the precipitation climate, including the occurrence of heavy precipitation and dry periods, is fundamental to planning and management tasks concerned with water resources, water use and the protection against natural hazards. These tasks are increasingly supported by quantitative models, such as river runoff models in hydrology, crop models in agriculture and glacier mass balance models in glaciology (e.g. Viviroli et al., 2009; Holzkämper et al., 2012; Machguth et al., 2009). Concerns over impacts from future climate change and the potential needs for adaptation have particularly fostered such model applications over the past few years. An important pre-requisite, however, are reliable meteorological input data, ideally provided on a regular grid. In the Alpine region, spatial analyses of precipitation that resolve the mesoscale distribution in complex terrain are particularly relevant to modelling the components of the hydrosphere. In its report on climate change adaptation in the Alps, the European Environment Agency has identified a need for monitoring and data collection to expand the knowledge base and widen the scope for analysis of long data series (EEA, 2009). There exist several sources of objective spatial precipitation analyses for the Alps. Firstly, the Alpine region is covered in gridded climate datasets that extend over global or continental domains (e.g. Adler et al., 2003; Hijmans et al., 2005; Haylock et al., 2008). While much improvement was made in these datasets with higher station density, finer grid resolution and advanced interpolation methods (e.g. Klok and Klein Tank, 2008; Hofstra et al., 2008), their reliability in the Alpine region is still handicapped by the heterogeneity of available measurements and the large spatial variation of precipitation (Hofstra et al., 2009, 2010). Secondly, gridded precipitation datasets have been developed in many Alpine countries but confined to the national territories or subregions (e.g. Ceschia et al., 1991; Gyalistras, 2003; Perčec Tadić, 2010; Brunetti et al., 2012; Rauthe et al., 2013). The focus of recent developments in this category was towards high spatial and sub-daily temporal resolutions. This has fostered techniques that incorporate, in addition to conventional rain-gauge measurements, also data from radar and from analyses with numerical weather prediction models (e.g. Paulat et al., 2008; Lussana et al., 2009; Vidal et al., 2010; Wüest et al., 2010; Haiden et al., 2011; Erdin et al., 2012; Mounier et al., 2012). The segmentation into national sub-domains and methodological differences in these datasets still complicate a consistent climatological overview for the region as a whole and hamper applications with a trans-national area of interest. 2013 Royal Meteorological Society A third category of precipitation analyses aims at filling the gap between large-scale and national analyses by providing climate information consistently for the region as a geographic entity. Building on ideas from earlier trans-national climatologies (Fliri, 1974; Fliri and Schüepp, 1983), Frei and Schär (1998) have developed a mesoscale Alpine-wide grid dataset, based on a collation of rain-gauge data from national climate networks. The dense spatial coverage (more than 6000 stations), daily resolution and multi-decadal extent (1971–1990) of this dataset have contributed to a high-resolution climatology (Schwarb et al., 2001), analyses of daily statistics (Frei and Schmidli, 2006) and a description of long-term precipitation changes (Schmidli et al., 2002). In the meantime, the daily trans-Alpine grid dataset was used in numerous applications, including the validation of regional climate models, the study of extreme events, river runoff modelling and the evaluation of climate datasets (e.g. Suklitsch et al., 2008; Smiatek et al., 2009; Bougeault et al., 2001; Martius et al., 2006; Kleinn et al., 2005; Hofstra et al., 2009; Rubel and Rudolf, 2001). A view on the greater Alpine region was later also pursued by Efthymiadis et al. (2006), who developed a gridded precipitation dataset, at monthly resolution and covering a period of more than 200 years (1800–2003). This dataset is based on 192 high-quality climate records (Auer et al., 2005). This study introduces an update and enhancement of the trans-Alpine precipitation dataset of Frei and Schär (1998, hereafter referred to as FS98). Hence, we present a new daily grid dataset and consistent analysis of the precipitation climate across seven Alpine countries (Austria, Croatia, France, Germany, Italy, Slovenia and Switzerland). The enhancements of FS98 are threefold: firstly, the collated rain-gauge database is extended for the recent years (after 1990) and newly expands over a 38-year period (1971–2008). Our data collection efforts have also resulted in a significant improvement of data density in previously under-sampled regions and led to a more homogenous data coverage. This was achieved by the inclusion of newly digitized data and by assembling datasets from additional institutions. In fact, the major parts of the rain-gauge dataset of FS98 were newly assembled. Secondly, substantial efforts have been made to ensure the consistency and quality of the different data contributions. Part of these efforts was the development and application of a data quality control procedure designed to detect typical problems of data coding in rain-gauge datasets. Thirdly, the spatial interpolation procedure of FS98 was enhanced to include climatological precipitation–topography relationships in order to improve the reliability of the resulting grid dataset. The daily spatial analyses are newly provided at a grid spacing of 5 km, as compared to 25 km in FS98. Apart from the technical information we also present and discuss, in this publication, a selection of analyses utilizing the new grid dataset. These highlight some of the main features of the Alpine precipitation climate, including the occurrence of heavy precipitation and long dry periods. Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS Paris Stuttgart Vosges Region 3 GERMANY lb nA Black bia a Forest w S Bohemian Forest Wien München AUSTRIA Morves SWITZERLAND Bern J u r a Zürich FRANCE Region 2 Valais Lyon Alto-Adige Dolomiti Ticino Trentino Ljubljana Valle d‘ Aosta Lombardy Milano Piedmont Po Valley ITALY Liguria Marseille Venezia Ap pe Zagreb CROATIA ps Al Torino Julian Alps S L O V E N I A ric Massif Central Glarner Alps L S P Innsbruck na Di A Allgäuer Alps nn Gulf of Genoa in o Adriatic Sea Region 1 Firenze Split 100 km 4200 3900 3600 3300 3000 2700 2400 2100 1800 1500 1200 900 600 300 0 Figure 1. Geographic and topographic (grey colours, altitude in metres) map of the domain. The cuboids are the regions used in Chapter 4. This study is part of the EU project EURO4M (European Reanalysis and Observations for Monitoring) that aims at preparing and analyzing datasets for monitoring European climate variations from in situ and satellite observations and from model-based regional reanalyses (International Innovation, 2011). The gridded Alpine precipitation dataset is publicly available for scientific use. This article is organized as follows: in Section 2 we introduce the updated Alpine rain-gauge dataset and in Section 3 the procedures for data quality control. The method of spatial analysis is described in Section 4, together with an analysis of interpolation errors and aids for the professional interpretation of the gridded analyses. Section 5 illustrates example analyses of heavy precipitation events. Several statistics of the daily precipitation climate, as derived from the new grid dataset, are discussed in Section 6. The conclusions of this study are provided in the final section. 2. Alpine precipitation dataset 2.1. Study domain Our efforts for collating high-resolution rain-gauge data have focused onto a rectangular longitude–latitude domain from 2–17.5◦ E to 43–49◦ N (Figure 1). The domain extends over approximately 1200 km from Central France across Switzerland to eastern Austria and over about 700 km from northern Italy to southern Germany. The territory of Slovenia and large parts of Croatia are also included in the study area. The main topographic feature of the domain is the arc-shaped mountain ridge of the European Alps, which extends over about 1000 km from the French and Italian Mediterranean coasts across Switzerland, northern Italy 2013 Royal Meteorological Society and Austria into the central parts of the continent. The highest Alpine mountain peaks reach elevations of more than 4000 mMSL and major cross-ridge passes are at 1500–2000 m above the adjacent flatland. The ridge is intersected with deep valleys, some of which are more than 100 km long and divide the ridge into major mountain massifs. To the north and northwest, the Alpine mountain range descends into a wide area of flat and hilly orography with a base elevation at around 300 mMSL. There are several small-scale hill ranges in this part of the Alpine foreland, including the Vosges and Jura mountains, the Black Forest and the Bohemian Forest, all with peak elevations below 1500 mMSL. The Massif Central in France is a wide region of plateaus and mountain ranges. With peak elevations up to 1900 mMSL it constitutes a prominent orographic feature upstream the south-western sector of the Alpine arc. To the south, the Alps run out into the Po Valley of northern Italy, a vast basin below 200 mMSL. It is delimited also by the Ligurian Alps and the northern Apennines. Finally, in the southeast of the domain, the Dinaric Alps stretch southward from the Carnic and Julian Alps along the east coast of the Adriatic Sea. The Alpine region is influenced by and delineates between several large-scale European climate regimes (e.g. Schär et al., 1997; Auer et al., 2005). The wide belt of adjacent mountain foreland included in our study domain allows to better distinguish between topographically influenced precipitation patterns and those of the ambient climate regimes. 2.2. Dataset The first version of the Alpine rain-gauge dataset of FS98 was covering the period 1971–1990 and included records Int. J. Climatol. (2013) F. A. ISOTTA et al. Table 1. Providers (institutions by nation) that have contributed data to the renewed Alpine rain-gauge dataset. The right column gives the number of stations with valid data available for the national territory averaged over all days of the period 1971–2008. Country Provider Austria Federal Ministry of Agriculture, Forestry, Environment and Water (BMLFUW), Division Water Balance, Vienna (data status of December 2010) Meteorological and Hydrological Service (DHMZ), Zagreb MétéoFrance, Toulouse German Weather Service (DWD), Offenbach Climatological Archive for Northern Italy (ARCIS, www.arcis.it) Bolzano Alto Adige: Servizio Meteorologico and Ufficio Idrografico Valle d’Aosta: Centro Funzionale Regionale Emilia-Romagna: Servizio Idro-Meteo-Clima ARPA Emilia-Romagna Friuli Venezia-Giulia: Osservatorio Meteorologico Regionale, ARPA Liguria: ARPA Liguria Lombardy: Servizio Meteorologico ARPA Lombardia Piedmont: ARPA Piemonte Trentino: Centro Funzionale di Protezione Civile, Meteotrentino Veneto: ARPA Veneto and Unità Idrografica regionale ARPA is the Italian abbreviation for ‘Regional agency for the protection of the environment’ National Meteorological Service of Slovenia, Ljubljana Federal Office of Meteorology and Climatology MeteoSwiss, Zurich Croatia France Germany Italy Slovenia Switzerland of daily precipitation totals from about 6600 stations. Several updates of this dataset were made in the following years for some of the national territories. Since 2009 a major renewal of the FS98 dataset was pursued as part of the EURO4M project. The renewal aimed at an extension to more recent years (up to 2008) and an improvement of spatial and temporal coverage where the earlier version was deficient. An extension of the dataset back in time (pre-1971) could not be accomplished because, for some Alpine countries, earlier high-resolution data is not available in digital form. With the present renewal, several national components were entirely recompiled from the original data provider. This ensures a more upto-date status of the collated dataset, taking advantage of the data owners’ efforts into data quality and data rescue since the first collection more than 15 years ago. The new version of the dataset encompasses more than 8500 time series in total over the 38-year period 1971–2008. It integrates available measurements from the high-resolution rain-gauge networks operated in seven Alpine countries. Table 1 lists the data providers that have contributed to the renewed dataset. Except for Italy, the providers are national meteorological and hydrological services, each of which has contributed a large number of time series for the respective national territories. In Italy, the operation of high-resolution measurement networks is under the responsibility of regional meteorological and environmental services. Therefore a central point of contact was missing. A major contribution of data from Italy was obtained through project ARCIS (Archivio Climatologico per l’Italia Settentrionale, climatological archive for northern Italy). ARCIS is a coordination effort of the regional services in northern Italy promoting the exchange and common analysis of long-term climate data (Pavan et al., 2013). Our contacts 2013 Royal Meteorological Society Mean number of stations 950 140 2420 800 760 200 480 with the ARCIS consortium were through ARPA Emilia-Romagna. ARCIS contributed 627 records, which provided a continuous coverage at climatological quality standards for all of northern Italy, yet with limited spatial density. To further improve data density, we have contacted the regional agencies individually (see Table 1) and could obtain many more time series. In general these are, however, more interrupted and extending over shorter periods only. The renewed Alpine rain-gauge dataset comprises typically 5500 observations on any day of the period 1971–2008. Figures 2 and 3 show, respectively, the distribution of stations in the domain and the evolution of station density over time. For Austria, France, Germany, and Switzerland, the renewed dataset carries on the high data density of the FS98 dataset, essentially by extending previously available station records in time. In these regions, data density varies between 7 and 14 stations per 1000 km2 (one station per 80–150 km2 ) and a large proportion of the stations dispose of long records (Figure 2). However, there is a tendency towards coarser coverage in more recent years (Figure 3), which reflects the reduction of climatological networks implemented in these countries. For Germany, the reduction is particularly pronounced after 2003, which is to be understood as a compensation for improved coverage by radar. In Italy, the ARCIS dataset and the contributions from regional services, together, yield a fairly stable data density of six stations per 1000 km2 (one station per 180 km2 , Figure 3). However, this number is an average over the entire Italian sector and the spatial coverage is inhomogeneous (Figure 2). Comparatively higher density could be achieved over the north-eastern part of the country, along the Ligurian coast and the Apennines (Emilia-Romagna), while the central and western parts Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS 0.8 0.6 0.4 0.2 Figure 2. Distribution of stations from which records of daily precipitation are integrated in the renewed Alpine rain-gauge dataset. Shading represents the fraction of the full period (1971–2008) covered by the respective record. 14 Germany # stations/1000 km2 12 Switzerland 10 Slovenia Austria 8 France 6 Italy Italy old 4 Slovenia old 2 0 1970 Croatia 1975 1980 1985 1990 Year 1995 2000 2005 2010 Figure 3. Evolution of station density (number of stations per 1000 km2 ) over time (1971–2008) in the renewed Alpine rain-gauge dataset. Density is defined as the number of available observations over the national territory divided by area. For comparison, the station density is also shown for Italy and Slovenia as represented in the dataset before renewal (labelled ‘old’). (Lombardy and Piedmont) have only coarse coverage. Note that the dense data for Lombardy is from a network that was established starting in year 2000 only, and hence these records cover only a small portion of the full period (Figure 2). Clearly, over northern Italy the Alpine raingauge dataset has a more heterogeneous and a more interrupted data coverage compared to most other parts of the domain. But the renewal of the dataset in this region was also more fundamental and this brought a marked improvement over previous versions of the FS98 dataset. Note, for example, that the latter had very limited data coverage over Italy after 1986, which could be improved substantially with this renewal (Figure 3). For the south-eastern parts of the domain, new datasets for Slovenia and Croatia have been integrated in the Alpine rain-gauge dataset. In the case of Slovenia, advantage could be made of digitization efforts since the last delivery, so that data density is increased by a factor of about four (Figure 3). With this contribution the region 2013 Royal Meteorological Society of the Carnic and Julian Alps, a region of particularly heavy precipitation in the Alps, is now covered fully at high-resolution. As for Croatia, the station coverage is less dense than elsewhere (about one station per 300 km2 ). Nevertheless, this contribution is valuable to follow the transition of the Alpine precipitation climate into the south-eastern flatlands. The national and regional services of the Alpine region operate their rain-gauge networks with different instruments and observing practices. As for the instruments, the differences are not fundamental. The classical Hellmanntype collector and variants thereof are the most widely used measurement devices. Since around 1980 (depending on service), subsets of the networks are operated with automatic devices, typically tipping bucket instruments. In most contributions, the proportion of automatic gauges is small (less than 15%) and they are mostly at lower elevations. Although the systematic measurement error of a rain-gauge varies between instrument type and height of Int. J. Climatol. (2013) F. A. ISOTTA et al. deployment, the exposition to wind and hence the environment of the rain-gauge is the most important factor for measurement bias (e.g. Sevruk and Zahlavova, 1994; Sevruk, 2005). Therefore in our analysis we do not distinguish between differences of instrument type. As for the observing practice, the daily reading times vary slightly between the different services. For Switzerland they are 06:30 UTC, for Italy 08:00 UTC and 06:00 UTC for all others. These differences are ignored in our analysis. During the collation of the 16 different data contributions special attention was required to adequately account for differences in archiving conventions. Labelling of missing values, specification of station coordinates, the assignment of dates to the data, for example, were different and not always specified in sufficient detail. Hence, apart from the time needed to arrive at a formal agreement on the exchange of data, additional efforts were needed to resolve technical issues. Numerous comparisons and checks have been conducted to test and ensure the internal consistency of the resulting collated database. 3. Data quality control There are numerous sources for gross errors in the collated Alpine rain-gauge dataset, ranging from occasional instrument failures, to transmission, digitization and storage errors. Gross errors can significantly affect the reliability of statistical analyses. For example, erroneous coding of data gaps by zero precipitation will result in underestimates in the frequency of wet days. Indeed, deficiencies in data quality were spotted in several early analyses, and were manifest particularly as isolated wet or dry reports. All institutions contributing to the Alpine rain-gauge dataset have applied their native quality control procedures before providing the data. The testing employed is however very variable. In some cases it included laborious manual validation and versatile automatic spatial checking. In others, just single-site tests for physical plausibility. To remedy frequent problems of data quality as evident during the climatological analyses, we devised an elementary testing procedure that could be applied systematically across the entire region. Its purpose was the detection and flagging of gross errors. Other issues of data quality, such as the systematic measurement bias from wind-induced undercatch, wetting and evaporation losses (e.g. Sevruk, 2005, Groisman and Legates, 1994) as well as temporal inhomogeneities due to station relocations and instrument changes (e.g. Wijngaard et al., 2003, Begert et al., 2005) were not addressed with this procedure. Detectability of gross errors strongly depends on station density. Hence, its large variation across the Alps poses a particular challenge to the design of a suitable procedure. The philosophy of our approach is conservative in the sense that flagging is only applied when there is strong evidence for implausibility. As a result, the Alpine rain-gauge dataset contains gross errors even after 2013 Royal Meteorological Society checking and their rate is likely larger in areas with coarse station coverage, where the testing is less powerful. In the following we describe sequentially the three steps of the applied quality checking procedure. These encompass (a) the scanning of time series for coding problems, (b) a fully automatic spatial consistency check, and (c) the identification of over all suspicious time series. 3.1. Checks for coding problems The initial testing step aims at identifying well-known and recurring coding errors in the time series. Firstly, the internal consistency of the time coding (reference date for the non-calendaric 24-h accumulation period) between the different data contributions was verified. To this end nearby time series from different providers were checked for suspicious lags. Secondly, the time series were scanned for suspicious duplicate values, as well as for negative and non-physical large values. In the latter case values exceeding 450 mm per day were manually checked. Apart from an extreme flash flood event in France (more than 500 mm per day on 8 September 2002, see Anquetin et al., 2005), all exceedances turned out to be implausible, likely the result of arithmetic shifts or multi-day accumulations, and they were flagged. Finally, all instances were manually checked when a time series reported zero precipitation throughout an entire calendar month. Real dry periods over more than 30 days are rare in the Alps, except along the Mediterranean coast. Many of these occurrences in the dataset turned out to come from erroneous coding of long data gaps. These episodes were therefore flagged unless simultaneous long dry periods were found at nearby stations. 3.2. Spatial consistency check A fully automatic spatial consistency test was applied by comparing all individual daily reports to simultaneous reports at stations in the neighbourhood. The procedure is a modified version of that described in Scherrer et al. (2011) (see also Behrendt, 1992). It distinguishes between three situations, namely, that of an isolated wet report, that of an isolated dry report and that of an out-of-range report. The situation of an isolated wet report occurs if a non-zero test value is surrounded by dry conditions (all stations within a 50 km radius report less than 0.3 mm per day). The test value is considered implausible and is flagged if it exceeds a certain threshold. The threshold is defined to vary with distance to the closest neighbour station and ranges from 0.3 mm per day for a co-located nearest neighbour to 3 mm per day for a nearest neighbour as far as 15 km. If the nearest neighbour is further than 15 km no constraints are employed and the test value is retained anyway. To account for the short-scale variability during the convective season, the tolerance with far nearest neighbours is increased to 3.5 mm per day and the maximum distance is set to 20 km between May and September. The definition of these thresholds is based on supervised tests in different regions of the Alps. Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS The procedure for an isolated dry report is employed, if the test value is less than 0.3 mm per day but all stations in a 50 km radius are wet. In this case and if the closest neighbour station is at less than 15 km distance, an estimate of precipitation is calculated by spatial interpolation using the procedure described in Section 4 but without the test station itself. The dry test value is considered implausible, if this cross-validation estimate exceeds a threshold. The test threshold is defined similarly to the case of isolated wet reports, i.e. depending on distance of closest neighbour and time of the year, but with slightly larger values (larger by 0.5 mm per day). The checking for out-of-range reports is adopted to all situations not considered with the above cases. The procedure compares the test value against a cross-validation estimate, again using the interpolation procedure of Section 4. The tolerance (critical difference between test value and cross-validation estimate) is formulated for square-root transformed values and in dependence of the value range of reports in the surrounding. The latter ensures that the procedure is more permissive in case of larger spatial variations in the neighbourhood. To account for the limited detectability of gross errors with coarse network density, the out-of-range test is only effective if at least three stations are within a 20-km distance around the test station and if at least one has an elevation difference of less than 300 m. Like with the other automatic tests the chosen threshold dependence is the result of detailed case studies and of experiments with different settings and in different regions of the Alps. 3.3. Identification of suspicious time series The spatial consistency test as just described revealed several data records with a particularly frequent rejection of reports. To check whether there were more fundamental quality problems with these stations, the corresponding records were inspected manually and compared to nearby stations. As a result, some of the time series (longer episodes of the whole record) had to be declared as highly implausible and these were systematically flagged. In other cases, the inspection revealed that records for isolated stations were lagged in time by 1 d, in which case these records were shifted. If the manual inspection could not clearly identify systematic data quality problems, the non-rejected reports of these potentially suspicious time series were retained. 3.4. Rejection statistics The rate with which original reports have been objected by the adopted quality control procedure varies considerably between the different providers (countries). This reflects the variable level of native quality control, the complications for reliable measurement in complex terrain, and the reduced power for gross error detection in coarse networks. In total the rejection rates for the spatial consistency checks range from 0.2% to 0.8% between the different providers. For the subset of stations above 1500 mMSL, the rejection increases to 0.3–1.1%. Note 2013 Royal Meteorological Society that the out-of-range test was not effective for many of the high-elevation stations (lack of nearby stations at comparable altitude) and the higher rate there likely reflects technical difficulties (wind, snow and ice). When stratified between the various components of the spatial consistency test, the largest rejection rates are found for isolated dry reports (1.4–4.5‰) and for isolated wet reports (0.5–3.2‰). Objections of out-of-range reports were less frequent (0.1–0.7‰). The number of deleted entries due to coding problems, especially when a whole month has no precipitation at a certain station, is moderate for most providers. Altogether, the employed control procedure has improved data quality and consistency in the collated Alpine rain-gauge dataset. This is discernible, for example, in far less frequent occurrence of artefacts in the daily gridded analyses and in more coherent spatial patterns of climatological summary statistics, such as the frequency of wet/dry days and quantiles. 4. Spatial analysis We have used the collated Alpine rain-gauge data to establish a dataset of pan-Alpine daily precipitation fields on a regular grid. This gridded dataset supersedes several earlier versions that were based on the FS98 station database (e.g. Frei, 2006, Schmidli et al., 2007). Apart from the more extended temporal and spatial coverage of the underlying station data (see Section 2), the most significant change in this new grid dataset is the adoption of a finer grid spacing. The spatial analyses are now provided on a regular 5 × 5 km grid in the ETRS89LAEA coordinate system (Lambert Azimuthal Equal Area Coordinate Reference System). Note that ETRS coordinates have become a mapping standard in Europe (Annoni et al., 2003). The finer resolution is motivated by the improved station density and demands for new applications, for which the previous datasets at 50 and 20 km grid spacing were not satisfactory. The spatial analysis (interpolation) employed for the new grid dataset is not substantially altered from that for the earlier datasets. Essentially, the changes involve new settings of the method’s parameters that allow better reproduction of fine-scale variations in regions with dense station coverage. In the following sub-sections we recapitulate on the principles of the method, discuss issues relevant for the application of the grid dataset and describe the structure of interpolation errors. The grid dataset provides the basis for our analysis of daily precipitation statistics in the Alps (Sections 5 and 6). 4.1. Method The adopted method of spatial analysis relies on the widely used anomaly concept where separate analyses are calculated for some reference condition, typically a long-term mean, and for the relative anomaly from that reference on the day under consideration (e.g. Widmann and Bretherton, 2000; Haylock et al., 2008). Int. J. Climatol. (2013) F. A. ISOTTA et al. Multiplication of the anomaly and reference grids finally yields the daily precipitation analysis. In our case, the reference condition for a day is the long-term mean precipitation (period 1971–1990) of the pertinent calendar month. The construction of the climatological reference fields is based on the PRISM method of Daly et al. (1994, 2002). For the estimation at a grid point PRISM uses a linear precipitation–elevation relationship that is estimated from stations in the neighbourhood of the target point by linear regression. A key feature of PRISM is to ensure that stations are representative for the physiographic conditions at the target location (see also Daly et al., 2008). For this purpose, PRISM attributes larger weights in the regression to stations with an exposition, slope orientation, etc. similar to the target location. Schwarb (2000) has adapted PRISM for the Alpine region and applied it with validated station normals from the Alpine rain-gauge dataset and a highresolution digital elevation model to derive mean monthly precipitation fields at a 2-km resolution (see also Schwarb et al., 2001). The aggregation of these fields onto the 5km ETRS grid serves as a climatological reference for the daily interpolation employed in this study. Note, that our climatological reference is valid for the 20-year average from 1971 to 1990. This choice proved to yield averages for many more stations than if the full 38-year period would have been chosen as reference (see also Figure 3). The interpolation of relative anomalies for a particular day is performed with a weighting scheme that emphasizes measurements from stations closer to and/or with stronger directional isolation relative to the analysis grid point. To this end a modified version of the SYMAP algorithm by Shepard (1984) is employed. One of the modifications is the adoption of a smoother radial dependence in the weighting function (see Equation (1) in FS98), which aims at estimating an area-mean value directly, rather than via preliminary analysis on a primary grid (see e.g. Legates and Willmott, 1990). Moreover, Shepard’s original ‘gradient correction’ is omitted in our application for the same reason. In extension of the original procedure we apply a spatially and temporally variable search neighbourhood, similarly to FS98. In the present application (5-km ETRS grid) the search radius is successively increased from 15 to 60 km (in steps of 5 km) until at least three valid station measurements are contained. A grid point is left at ‘missing’ if less than three measurements are available within 60 km. The variation of the search radius allows to better account for the large spatial variations in station density, so that fine-scale information can be exploited where the network is dense. The main purpose of the anomaly concept in the daily interpolation is to reduce the risk of systematic errors due to the non-representative distribution of stations, especially the prevalence of valley and lowland stations over high-elevation stations (see FS98, Konzelmann et al., 2007). Indeed, comparisons of daily interpolations with and without the anomaly approach have shown differences in long-term mean precipitation of several 2013 Royal Meteorological Society 10% (Widmann and Bretherton, 2000; Frei et al., 2003). Interpolation with a reference and anomalies typically yields larger mean values compared to interpolation of daily precipitation directly. The comparisons suggest that interpolation without reference is more prone to systematic precipitation underestimates in high-mountain regions due to the prevalence of stations in comparatively dry valley conditions. These biases would give rise to substantial inconsistencies in the long-term water balance as is suggested by Schädler and Weingartner (2002) who compared the PRISM-derived long-term climatology against water balance based precipitation estimates in Switzerland (see also Weingartner et al., 2007). 4.2. Interpretation It is important to point out some general issues of the construction of the grid dataset that may need attention in its application and the interpretation of results. Firstly, the variations in the daily precipitation analyses at scales near the grid spacing are mostly imprints of the climatological reference fields. Because these patterns reflect long-term mean conditions, they may be of limited representativeness for individual days. Care should therefore be exercised in interpreting these small-scale variations. They represent the temporally stable component of small-scale variability and are more realistic for precipitation sums over longer periods. Their primary function is to ensure consistency with the finescale pattern of the long-term climatology and, hence, to reduce biases from unrepresentative station distributions (see previous subsection). Secondly, the procedure of spatial interpolation by distance and angular weighting has a smoothing effect. The daily analyses do not replicate peak values at individual stations even if these were located at grid points. It is therefore more appropriate to interpret grid point values as estimates of area-mean precipitation. However, there is ambiguity about the spatial scales effectively resolved by the analysis. At the daily timescale the station spacing (10–15 km in high density areas) can be considered as a lower bound for the effective resolution. For longer-term averages, it may be finer, depending on the strength of systematic topography imprints. In conclusion, the fine spacing of the underlying grid does not mean that daily precipitation patterns are effectively resolved at these scales. The true resolution varies across the domain and with time, depending on the density of stations. Thirdly, the grid dataset is affected by biases in raingauge measurements (Neff, 1977, Yang et al., 1999). The ‘gauge undercatch’ is comparatively larger during episodes with strong wind and during weather with low rainfall intensity or with snowfall. Hence the related errors in the grid dataset are quite variable. Sevruk (1985) and Richter (1995) have estimated the magnitude of systematic measurement errors on seasonal mean precipitation in Switzerland and Germany. Their estimates range from 7% (5%) over the flatland regions in winter (summer) to 30% (10%) above 1500 mMSL in winter Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS (summer). Sevruk’s estimate of a season independent bias of 4% for southern Switzerland may be characteristic for the Po valley. Finally, our collection of data has not made prerequisites regarding the long-term consistency (temporal homogeneity) of the time series. Artefacts from changing measurement conditions must therefore be expected in the rain-gauge data series. These and the variations of network density over time are compromising the long-term consistency of the grid dataset (see e.g. Hofstra et al., 2009). Therefore, the present dataset is not suitable for applications where long-term consistency is crucial, such as for trend analysis. 4.3. Interpolation errors To allow for well-informed applications of the present grid dataset, it is desirable to quantify errors of the spatial interpolation. The primary source of reference data currently available for an evaluation of the daily grid point estimates is station data. This is far from ideal when grid point estimates more likely represent area-mean conditions (see previous subsection). The scale mismatch can inflate interpolation errors by sampling errors of the reference (see e.g. Villarini et al., 2008). Despite this we proceed here with a comparison of grid point values against point measurements, but will consider the statistics as being representative for a user who interprets grid point values as point estimates. The statistics will provide an upper bound of the errors for area means. Our quantification of interpolation errors is based on a leave-one-out cross-validation (similar to jackknifing) where station observations are left out in turn and estimates are calculated by spatial interpolation from the surrounding stations. The cross-validation runs over all wet observations (≥1 mm per day, period 1971–2008) of all stations in three subregions of the Alps. The regions represent variable conditions in station density and terrain complexity (see domains in Figure 1): Region 1 is over Croatia, an area with coarse station coverage (one station per 240 km2 ), Region 2 for high-Alpine topography with a dense network (one station per 66 km2 ), and Region 3 over a flatland area of France with a dense network (one station per 86 km2 ). At the daily time-scale, interpolation errors must be expected to vary considerably with precipitation intensity. Absolute (relative) errors are larger for large (small) daily precipitation. For this reason, the cross-validation errors are stratified with respect to precipitation intensity. Figure 4 depicts boxplots of the distribution of relative interpolation errors for each region, separately for winter and summer (each of the subsamples comprises several thousand values). The x -axis distinguishes between classes of precipitation intensity. These classes are defined in terms of quantiles of daily precipitation (station dependent, wet days only), so that each box describes error statistics in a section of the distribution function. For all regions and seasons, there is a strong sensitivity of the error distribution on precipitation 2013 Royal Meteorological Society intensity. Relative errors are largest at low-intensity and decrease continuously with higher intensity. Obviously this is because even small absolute errors in the interpolation can mean substantial relative errors when daily totals are small (the sensitivity for absolute errors is reverse). Interestingly, the error distributions are slightly biased with a tendency for overestimates at low intensities, particularly in summer, and a clear shift towards underestimates at high intensities. This reflects the smoothing effect of the interpolation, which tends to smear out localized rainfall peaks into low-intensity regions and to damp high-intensity peaks. The systematic errors (intensity dependent biases) are more substantial in summer when rainfall patterns are more convective, with space scales closer to the station spacing and, hence, more detrimental smoothing. Quantitatively, there are large differences in both, the systematic and random errors (offset and spread of boxes) between the three regions. In Region 1 (low station density) relative errors are substantial throughout the intensity spectrum. At moderate intensity (near the median of wet-day totals), in winter, the interpolation under- or overestimates point observations by a factor of 1.5 or more (relative error outside 0.6–1.5) in more than half the cases. For Regions 2 (Alps) and 3 (Flatland), the same values of typical relative error are 1.25 and 1.2, respectively. In winter, high-intensity precipitation events are systematically underestimated by about 20, 12 and 8% in the three regions. All these variations are plausible in view of the different sampling conditions and climate characteristics in the three regions. Evidently, the quality of the grid dataset is depending critically on the density of the underlying rain-gauge network. Between summer and winter, there is an interesting change in the error ranking between Region 2 and Region 3: despite its flat terrain, Region 3 exhibits larger errors at high-intensity than Region 2 (Alps). This may reflect differences in the nature of heavy precipitation. Over the northern flatlands, heavy precipitation in summer is almost exclusively from convection, whereas, in the Alps, heavy events can also occur in connection with stationary flow and more wide-spread precipitation (e.g. MeteoSwiss, 2006; Schmutz et al., 2008). The measures of uncertainty in Figure 4 are representative when grid point values are interpreted as point estimates. Clearly, we do not promote such a practice for applications of the grid dataset over the Alps. The dilemma is posed by the lack of a suitable observational reference, necessary for verifying the interpolation at the grid-pixel scale or at some ‘effectively resolved’ scale, compliant with more suitable interpretations by users. As alternative, statistical models of the spatial precipitation variability could also serve to estimate uncertainties in area means. For example the framework of geostatistics has been used for quantifying interpolation uncertainties and their dependence on spatial scale (Ahrens and Jaun, 2007; Frei et al., 2008; Vogel, 2013). Clearly, further developments will be required for a systematic application of such methods over a large area such as the Alps. Int. J. Climatol. (2013) Q10–Q90 Q25–Q75 Q50 DJF Region 1 Region 2 Region 3 Q10–Q90 Q25–Q75 Q50 JJA 0.80 1.25 Region 1 Region 2 Region 3 1.25 0.80 0.99–0.999 0.98–0.99 Category (wet-day quantile prob) 0.95–0.98 0.9–0.95 0.8–0.9 0.6–0.8 0.4–0.6 0.2–0.4 0.1–0.2 0.33 0.50 Error (factor) 2.00 3.00 0.33 0.50 Error (factor) 2.00 3.00 F. A. ISOTTA et al. Figure 4. Boxplot of interpolation errors determined from a systematic leave-one-out cross-validation in three characteristic regions (see domains in Figure 1). Region 1: area with a coarse station network, Region 2: mountainous area with a dense network, Region 3: flatland area with a dense network. Errors are expressed as the ratio between the interpolation (at the location of the station) and the observation at the station. The errors are stratified into bins of precipitation intensity, which are defined in terms of quantiles (wet days only) with low (high) intensities on the left (right). Bins are labelled by probabilities. For example, the second bin from the left, includes error measures from all days when the observed daily precipitation fell between the 20% and 40% quantiles of all wet days at the same station. The upper plot is for winter (DJF), the lower for summer (JJA). The boxplots represent the median (bold line), the inter-quartile range (box) and the 10–90% quantile range (whisker) of the error distribution. 5. Example cases The high-resolution grid dataset is a comprehensive resource of the day-by-day course of precipitation in the Alps. It lends itself to study episodes of specific interest, such as heavy precipitation events, in a wider spatial context than is usually available in case studies at the national level. To illustrate this potential Figure 5 displays two well-known examples of the past decades. In both cases heavy precipitation fell over large areas and several national territories. The case in Figure 5(a) is for the 3 d from 23 to 25 August 1987. The synoptic situation on these days was characterized by a strong cold front advancing from the northwest and a cyclone near the Gulf of Genoa advecting maritime air masses towards the southern Alps (LHG, 1991). The combination of orographic, frontal and large-scale uplift as well as the continuous triggering of convection in this dynamic environment resulted in 2013 Royal Meteorological Society a richly structured rainfall pattern (Steinacker, 1988). Regions particularly affected were the coastal regions of north-western Italy and the southern flanks of the Alps, where rainfall regionally exceeded 300 mm. But large areal amounts were also recorded in inner-Alpine regions and along the northern pre-Alps. The event caused serious flooding and landslides in many valleys of the Italian, Swiss and Austrian Alps. In Switzerland, it was one of the most damaging flooding events of the past four decades (e.g. Hilker et al., 2009). The event of 20–22 August 2005 (Figure 5(b)) was associated with a cyclone of type Vc in the cyclone track classification of van Bebber (1891) that developed over the Gulf of Genua and moved eastward over the Adriatic Sea and the Balkans. The sustained transport of moist air masses around the eastern Alps led to longlasting and intense rainfall along the northern rim of the Alps. Rainfall totals reached well beyond 200 mm Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS 300 250 210 180 150 120 90 70 50 35 20 10 5 2 0.3 Figure 5. Precipitation sum in mm for the 3-d periods of 23–25 August 1987 (upper panel) and 20–22 August 2005 (lower panel). The thick line represents the 800 mMSL topographic contour. over a large area from central Switzerland to western Austria and southern Germany. Heavy rainfall was also recorded in the eastern Alps of Austria and Slovenia, likely due to the more south-easterly and hence upslope flow direction there. At many stations in Switzerland, eastern Austria and Bavaria the rainfall reached the local 100-year return level (Frei, 2006). The floods caused damages of nearly three billion Euros in these three countries (UVEK, 2008; Amt der Vorarlberger Landesregierung, 2005; Bayerisches Staatsministerium für Umwelt, Gesundheit und Verbraucherschutz, 2005). 6. The Alpine precipitation climate In this section we analyse the Alpine dataset to work out and discuss some key characteristics of the precipitation climate in the Alps. While previous pan-Alpine analyses have focused mostly on long-term mean precipitation (e.g. Fliri, 1974; Baumgartner et al., 1983, FS98), the present compilation emphasizes aspects of the daily statistics, including the occurrence of intense precipitation and long dry periods. For this purpose our analysis considers a subset of the descriptive indices that 2013 Royal Meteorological Society have been proposed by the CCl/CLIVAR/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI, Klein Tank et al., 2009). In the course of this study a comprehensive list of these indices was investigated. The subset presented aims at spanning a range of daily statistics and portraying the most important regional variations. The following sub-sections discuss the distribution of mean annual precipitation (6.1), key indices of daily precipitation statistics (6.2), and long dry and wet spells (6.3). An illustration of the annual cycle is given for the example of heavy precipitation in subsection (6.4). All the results were derived by calculating the indices directly from the gridded daily precipitation analyses of the 38 years from 1971 to 2008 (Section 4). 6.1. Mean annual precipitation Figure 6 depicts the distribution of mean annual precipitation (1971–2008). There is a wide range of values from approximately 400 to more than 3000 mm per year, with occasionally pronounced contrasts over short distances. Topographic effects manifest in patterns at variable space scales. For example, wet conditions are found in an elongated zone along the northern rim of the Alps. Embedded Int. J. Climatol. (2013) F. A. ISOTTA et al. in this zone are smaller scale variations that go along with prominent mountain massifs (e.g. the Glarneralpen in eastern Switzerland and the Allgäuer Alps at the western border between Austria and Germany). There is a similar wet band extending along the southern rim of the Alps (towards the Po valley), but with a more pronounced sectioning into two wet zones at major embayments of the main ridge. The first is centred over southern Switzerland (Ticino) and the Italian regions of Lombardia and Piemonte. This anomaly is connected to the wet anomaly along the northern rim across the section of St. Gottard, a particularly narrow cross-section of the ridge. The second wet region along the southern rim extends from the Dolomiti massif in north-eastern Italy eastward into the Julian and Carnic Alps (boarder between Italy and Slovenia) where particularly large annual mean values are observed. Compared to the two ridge-scale anomalies, regions in the inner Alps are comparatively dryer with mean values comparable to those over the flatland regions adjacent to the Alps. Particularly dry conditions are found in inner-Alpine valleys that are shadowed against directions with prevailing moisture bearing winds. Noteworthy examples are the Valais, the Aosta valley (north-western Italy), the upper Inn valley and the Vinschgau (northern Italy, near border to Austria). In addition to the pattern of the main ridge, several smaller scale hill ranges in the region show wet anomalies too. Examples are the Jura mountains, the Vosges mountains, Black Forest and Bohemian Forest. Mean values there are comparable to those along the northern Alpine rim, but the precipitation signal is mostly centred at the highest elevations. An exception to this is found in the Massif Central, which exhibits two distinct wet anomalies, one at the north-western and one at the south-eastern flanks of the Plateau. The comparatively dryer conditions of the southern Alps (south-western France), suggests that the Massif Central may have a rain-shadow effect on the latter. The wet anomaly of the Apennine is linked to that of the southern Alps by a narrow wet anomaly along the mountains of the Ligurian coast. Particularly dry conditions in the Alpine region are found at the mouth of river Rhone (French Mediterranean coast) and in eastern Austria where mean annual precipitation is less than 600 mm per year. The general distribution of mean annual precipitation found in the present analysis is very similar to that derived from earlier versions of the Alpine rain-gauge dataset (FS98). The higher resolution of the present analysis (5 km vs 25 km grid spacing), however, reveals previously unresolved and interesting patterns that can be related to individual mountain massifs and provide a better delineation of the topographic enhancement signals from dry valley and flatland conditions. (Compare Figure 6 with Figure 9 in FS98.) 6.2. Daily precipitation statistics Figure 7 depicts four selected indices of daily precipitation in the Alps representing basic characteristics of the daily rainfall frequency distribution. The indices 2013 Royal Meteorological Society considered are frequency of wet days, defined as occurrences when the daily total is 1 mm or larger (Figure 7(a)), the mean precipitation on wet days (Figure 7(b)), the mean of annual maximum daily totals (averaged over all years, Figure 7(c)) and the fraction of precipitation coming from moderate to intense events, i.e. days when the precipitation is equal to or larger than the 75% quantile (wet-day quantile at the gridpoint, Figure 7(d)). Note, that our definition of the wet-day threshold refers to the recommendation of the ETCCDI (Klein Tank et al., 2009), which is larger than the lowest measurable precipitation. We have investigated several other popular indices but found that these would not add substantially to the overall picture. For example, the popular indices for the 95% wet-day quantile and the frequency of days exceeding 20 mm show distributions that are qualitatively very similar to that of mean annual precipitation (Figure 6). The four indices reveal a pronounced asymmetry of the daily precipitation frequency distribution between regions to the north and south of the Alpine main crest. Along the northern rim and in the northern and western flatland regions, wet-day frequency is large (Figure 7(a)). In these areas rain falls on every third to every second day in the annual average. Over the southern rim and forelands, however, the wet-day frequency is lower with rainfall on one of three to five days. For wet-day mean precipitation the asymmetry is reversed (Figure 7(b)): average rainfall intensities are typically much larger along the southern compared to the northern Alpine rim with particularly large values observed at the two major embayments with large annual mean precipitation. These two regions stand out even more clearly at the tail of the frequency distribution (Figure 7(c)). Annual maxima are two to three times larger there compared to the moist northern rim. The two regions Piedmont-Ticino-Lombardy and Julian/Carnic Alps seem to be the mesoscale hot spots of heavy precipitation in the Alps. Note also that more than two thirds of the total precipitation comes from the 25% most intense wet days in these two regions (Figure 7(d)). Interestingly the anomaly along the northern rim is not manifest in Figure 7(d). This indicates that the shape of the tail of the frequency distribution at the northern rim is similar to that over the nearby flatlands and hence that the larger mean precipitation there comes primarily from the higher frequency and mean intensity of wet days. This is different over the southern portion of the domain, where the fraction from moderate to high-intensity events is not only larger but also shows spatial variation. In summary, these analyses imply that precipitation is more frequent to the north of the main ridge but more vigorous to the south, particularly so along the southern rim where much of the total precipitation is from rare intense events. Apart from the hot spots along the southern rim of the main ridge, areas of particularly heavy precipitation are also found along the Ligurian coast, along the southeastern slopes of the Massif Central and in the Dinaric Alps along the Adriatic coast (Figure 7(c) and (d)). The former two of these regions are characterized by low wet-day frequency (Figure 7(a)). Note particularly Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS 2600 2400 2200 2000 1800 1600 1400 1200 1000 800 600 Figure 6. Mean annual precipitation (mm per year) for the period 1971–2008. (b) (a) 20 18.5 17 15.5 14 12.5 11 9.5 8 6.5 5 0.48 0.45 0.42 0.39 0.36 0.33 0.3 0.27 0.24 0.21 0.18 (c) (d) 180 165 150 135 120 105 90 75 60 45 30 0.75 0.73 0.71 0.69 0.67 0.65 0.63 0.61 0.59 0.57 0.55 Figure 7. Four key statistics of the daily precipitation climate (annual, reference period 1971–2008). (a) Frequency of wet days (≥1 mm, fraction), (b) mean precipitation on wet days (mm per day), (c) mean of annual maximum daily precipitation (mm per day), (d) fraction of precipitation from days with moderate to high-intensity (≥75% percentile on wet days, fraction). that the Massif Central shows a similar asymmetry in precipitation statistics like that for the Alps: there is a marked north-west to south-east gradient in wetday frequency and a much stronger prominence in the intensity indices (Figure 7(b)–(d)) of the south-eastern compared to the north-western wet anomaly. Along the French Mediterranean coast there are the smallest values of wet-day frequency across the whole domain. The smaller scale hill ranges of the northern Alpine region (Jura, Vosges and Black Forest) receive more frequent and more intense precipitation compared to the adjacent flatlands (Figure 7(a) and (b)). But these hills are lesser distinguished from their surrounding in the other two indices (Figure 7(c) and (d)), suggesting that the 2013 Royal Meteorological Society larger mean precipitation there is mostly from increased frequency of light to moderate events. Noteworthy features of the indices over the flatland regions are the very low wet-day frequency along the French Mediterranean coast and a marked decrease of the same statistic towards eastern Austria and Croatia (Figure 7(a)). The latter indicates the influence of more continental climate near the eastern border of the domain. Finally, over the flatlands of central France (north-western corner of the domain) only a very small proportion of total rainfall comes from moderate to highintensity events (Figure 7(d)). Frequent light precipitation during much of the winter half year may be responsible for this. Int. J. Climatol. (2013) F. A. ISOTTA et al. 14 13 12 11 10 9 8 7 6 5 4 38 36 34 32 30 28 26 24 22 20 18 Figure 8. Mean annual maximum consecutive wet-day (left) and dry-day (right) period (in days). Reference period 1971–2008. 6.3. Long wet and dry spells The succession and duration of spells of wet and dry days are an integral part of a region’s precipitation climate. Figure 8 depicts the length (in days) of the longest period of consecutive dry days and wet days (critical threshold 1 mm), averaged over all years of the study period (1971–2008). Extreme wet spells (Figure 8(a)) show a distribution, which is roughly mixture of those for mean precipitation and wet-day frequency (cf. Figures 6 and 7(a)). Particularly long durations occur along the northern rim of the Alps and, to some extent, also along the southern rim. But the latter anomaly is far less pronounced in accord with the understanding that large mean precipitation there is primarily related to higher intensity (cf. Figure 7). Note, that the anomalies of mean precipitation found at the Julian and Carnic Alps as well as along the southern slopes of the Massif Central are barely distinguished in the wet spell distribution. As for the flatland long wet spells show a north-south contrast with lengths around 10 days to the north and around 6 days to the south of the ridge. The longest wet spells in the region occur at the northern and western hill ranges (Vosges, Black Forest, Jura, western Massif Central), which may reflect their exposition to moisture bearing flows during persistent weather situations. Long dry spells (Figure 8(b)) exhibit a strong gradient between the Alps and the northern foreland on the one hand (18–26 days) and regions to the south on the other (28–44 days). Particularly long dry-day periods, with average extreme spells exceeding 35 days, occur along the French Mediterranean coast and in Liguria (western Po valley). The two regions seem to be well shielded from rain-bearing weather systems by the surrounding topography. There, dry spells are met both in winter and in summer, with winter-time events being more frequent in Liguria and summer-time events in southern France. It is intuitive that long wet spells go mostly together with short dry spells and vice versa (Figure 8). But this correspondence is not universal. For example the continental regions of the domain (eastern Austria, Croatia and Slovenia) are characterized by comparatively short dry and short wet spells, indicating some fundamental differences in precipitation intermittency across the domain. 2013 Royal Meteorological Society 6.4. Annual cycle of heavy precipitation Spatial variations in the annual cycle of mean precipitation have been discussed in detail in FS98 (see their Figure 12). The results of this earlier analysis have not altered with the new dataset. Therefore, we rather examine the annual cycle of heavy precipitation here, considering that magnitude of intense events is of particular interest. Figure 9 depicts the evolution of the largest daily precipitation total in each month of the study period (1971–2008), averaged separately for each calendar month. The results for the sub-domains were obtained by averaging over all grid points in the domain. There are remarkable differences in the annual cycle of heavy precipitation between the domains: in the flatland regions of France there are only small variations through the year with a slight maximum in summer. This turns into a gentle pattern with two maxima at the Black Forest (representative also for the Jura and Vosges hills). The one in winter is associated with dynamically active weather systems (fronts, low-pressure systems) in combination with orographic enhancement; the one in summer is related to convective activity. Further east (sub-domains Bavaria, northern Alps and eastern Austria), the annual cycle markedly increases in amplitude and turns into a 12month cycle, with values in summer (convection) being about twice as large as in winter. Values along the northern Alps are generally larger than in the other two of these domains (see also Figure 7(c)). In the southern sub-domains, monthly maximum daily precipitation shows larger values and larger variations through the year, except over the Po valley (Figure 9). The maximum of the annual cycle is attained between September and November in all these domains, but the evolution is more complex: in some of the regions there is a secondary maximum in spring (south-eastern Massif Central, Ticino, Julian and Carnic Alps) and the minimum occurs either in winter (Ticino, Po valley) or in summer (south-eastern Massif Central, Liguria). The complex and spatially variable pattern of the annual cycle in these southern domains is related to several factors with distinct annual patterns, including the occurrence of southerly flow conditions (most frequent in spring and summer), the sea surface temperature of the Mediterranean (warm in late summer and autumn) and insolation-driven Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS 60 50 40 30 20 10 0 1 3 5 7 9 11 60 50 40 30 20 10 0 1 3 5 7 9 11 Northern Alps Bavaria Black Forest Central France 60 50 40 30 20 10 0 1 3 5 7 9 11 60 50 40 30 20 10 0 1 3 5 7 9 11 Eastern Austria 60 50 40 30 20 10 0 60 50 40 30 20 10 0 S–E Massif Central 60 50 40 30 20 10 0 1 3 5 7 9 11 Liguria 60 50 40 30 20 10 0 1 3 5 7 1 3 5 7 9 11 Julian and Carnic Alps 1 3 Ticino 9 11 60 50 40 30 20 10 0 1 3 5 7 5 7 9 11 Po Valley 9 11 60 50 40 30 20 10 0 1 3 5 7 9 11 Figure 9. Annual cycle of heavy precipitation for sub-domains. Columns represent, for each calendar month, the mean (averaged over years) of the monthly maximum daily precipitation (mm per day, reference period 1971–2008). convection (more frequent in late spring and early summer). It is worth mentioning that the annual cycle of heavy precipitation has interesting differences to that for mean precipitation (cf. Figure 12 in FS98). For example in the southern regions, the relative amplitude of spring and autumn maxima is reversed, with mean (heavy) precipitation being relatively larger in spring (autumn). The larger frequency of wet days in spring (about 0.45 in May in region Ticino) compared to autumn (about 0.29 in September) compensates for the less frequent occurrence of very high daily amounts. We have verified that these variations are not just an effect of the slight difference in sub-domains between Figures 9 and 12 of FS98. 7. Conclusions The grid dataset and climatological analyses, presented in this study, offer a comprehensive trans-national information source on Alpine precipitation of the recent 2013 Royal Meteorological Society past. The underlying rain-gauge dataset encompasses typically 5500 measurements on any day of the study period (1971–2008), probably one of the densest in situ monitoring systems available from a high-mountain area of this size. The dataset builds on an earlier data compilation effort (FS98) but major improvements in data coverage could be achieved in areas and time periods with previously poor coverage, notably the regions of northern Italy and Slovenia and the years after 1990. Also, the quality of the dataset was fundamentally reviewed by upgrading to the recent status of the national databases and by adopting a specifically designed quality control procedure that addresses common data coding errors and involved several semi-automatic checks. Compared to its earlier version, the new grid dataset is assembled at a higher nominal resolution (5 km grid spacing as compared to 20 km in FS98) by employing a new analysis method, which incorporates local precipitation–topography relationships at the climatological time-scale. The procedure aims at reducing the risk of systematic underestimates at high elevations. The higher Int. J. Climatol. (2013) F. A. ISOTTA et al. resolution of the grid dataset may not be expected to resolve precipitation at the scale of the grid spacing, but it facilitates the aggregation into and improves estimation of spatial averages over complex domain shapes. This is particularly useful in hydrological applications requiring mean precipitation over catchments, or for the evaluation of regional climate models, when the observational grid dataset needs to be assembled onto the model’s native grid structure. It is important that users of the Alpine precipitation grid dataset are aware of the limitations and uncertainties involved: while there are good prospects that the grid dataset resolves scales near the 5-km grid spacing when precipitation totals over longer periods are considered, the effective resolution for daily totals is clearly coarser than the grid spacing. At the daily scale the spatial resolution is more likely in the order of 10–25 km (typical station spacing), eventually even coarser, notably in regions and during time periods where the station coverage is more limited. When grid point values are interpreted as estimates at finer scales (e.g. as point estimates) quite substantial random errors must be expected at the daily time-scale: high precipitation intensities are systematically underestimated and low intensities overestimated. The magnitude of these errors depends on season, station coverage and topographic complexity (measures of the error for the case of interpreting grid point values as point precipitation are given in Section 4.3). Mismatches between the effective spatial resolution of the present dataset and the resolution of its application (e.g. that of a hydrological model to be driven by the dataset or that of a regional climate model to be verified against it) can give rise to systematic representativeness errors (e.g. Tustison et al., 2001). Further improvement of the effective resolution of the presented grid dataset will be difficult from in situ data alone, considering that most of the available data was incorporated. Approaches using radar data in combination with station data (e.g. Krajewski, 1987; DeGaetano and Wilks, 2008; Erdin et al., 2012) or analyses with numerical models (e.g. Häggmark et al., 2000; QuintanaSegui et al., 2008; Haiden et al., 2011) are promising, but, certainly in the case of radar, hardly applicable over climate time scales. More immediately, some improvements could be achieved with a more elaborate representation of small-scale precipitation–topography relationships, for example by conditioning them on circulation types, rather than just on month of the year (e.g. Hewitson and Crane, 2005). We are currently exploring the added value of such an approach. A further major limitation of the presented grid dataset is that we have to make reservations with regard to its long-term climate consistency. Changes in station locations and measurement devices have introduced inhomogeneities to some of the records. Moreover, variations in the station network over time can significantly compromise the consistency of the grid dataset. Such effects were observed in other grid datasets relying on timevarying station networks (e.g. Hofstra et al., 2009, 2010). 2013 Royal Meteorological Society We expect that inhomogeneities are relatively more evident at small spatial scales (near the effective resolution), but tend to average out when considering mean values over larger domains. Indicators of daily precipitation extremes, such as the frequency of a threshold exceedance, may be particularly sensitive to variations in network density, due to the smoothing of the interpolation. We therefore call for caution when applying the present dataset for purposes with a strong requirement on long-term consistency. Remedy for this caveat may be sought through statistical reconstruction of precipitation fields from a time-invariant homogeneous station dataset, for example by using the technique proposed by Schmidli et al. (2001, 2002) and Schiemann et al. (2010). This would inevitably be accompanied by a reduction in effective resolution. Hence, such a climate consistent grid dataset would not supersede the present grid dataset but rather exist as a complement that satisfies other requirements. Gridded datasets have become a popular and compact basis for describing the climate of a region. The present grid dataset offers a valuable resource of information on the precipitation climate of the Alpine region. Indeed the example analyses of this study have provided new insights into the distribution of mesoscale daily precipitation indicators, which would have been difficult to identify without a trans-national consistent data source. The dataset also offers itself as a basis for numerous applications, for example as input into quantitative models of environmental sub-systems, for statistical downscaling of climate change scenarios, the evaluation of regional climate models and many more. An application of this dataset for the evaluation of global and continental observation datasets is currently underway. The gridded Alpine precipitation dataset is available for research from http://www.meteoswiss.ch. Acknowledgements The research leading to these results has received funding from the European Union, Seventh Framework Programme (FP7/2007-2013) under grant agreement n◦ 242093. We are grateful to the following people for assistance and valuable discussions: Christof Appenzeller, Marco Gaia, Mark Liniger and Christian Lukasczyk (all MeteoSwiss, Switzerland), Claudio Cassardo (Univerity of Turin, Italy), Jean-Pierre Céron (Météo-France, France), Albert Klein Tank (Royal Netherlands Meteorological Institute, Netherlands) and Christoph Schär (ETH Zürich, Switzerland). We also thank two anonymous reviewers for their valuable comments. References Adler RF, Huffman GJ, Chang A, Ferraro R, Xie P-P, Janowiak J, Rudolf B, Schneider U, Curtis S, Bolvin D, Gruber A, Susskind J, Arkin P, Nelkin E. 2003. The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979present). Journal of Hydrometeorology 4: 1147–1167. Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS Ahrens B, Jaun S. 2007. On evaluation of ensemble precipitation forecasts with observation-based ensembles. Advances in Geosciences 10: 139–144. Amt der Vorarlberger Landesregierung. 2005. Das Starkregen- und Hochwasserereignis des August 2005 in Vorarlberg. Amt der Vorarlberger Landesregierung, 58 pp. Annoni A, Luzet C, Gubler E, Ihde J. 2003. Map Projections for Europe. Institute for Environment and Sustainability: European Communities. Anquetin S, Yates E, Ducrocq V, Samouillan SK, Chancibault K, Davolio S, Accadia C, Casaioli M, Mariani S, Ficca G, Gozzini B, Pasi F, Pasqui M, Garcia A, Martorell M, Romero R, Chessa P. 2005. The 8 and 9 September 2002 flash flood event in France: a model intercomparison. Natural Hazards and Earth System Sciences 5: 741–754. Arkin PA, Xie P. 1994. The Global Precipitation Climatology Project: first algorithm intercomparison project. Bulletin of the American Meteorological Society 75: 401–419. Auer I, Böhm R, Jurkovic A, Orlik A, Potzmann R, Schöner W, Ungersböck M, Brunetti M, Nanni T, Maugeri M, Briffa K, Jones P, Efthymiadis D, Mestre O, Moisselin JM, Begert M, Brazdil R, Bochnicek O, Cegnar T, Gajic-Capka M, Zaninovic K, Majstorovic Z, Szalai S, Szentimrey T, Mercalli L. 2005. A new instrumental precipitation dataset for the greater alpine region for the period 1800–2002. International Journal of Climatology 25: 139–166. Baumgartner A, Reichel E, Weber G. 1983. Der Wasserhaushalt der Alpen. Oldenburg: München. Bayerisches Staatsministerium für Umwelt, Gesundheit und Verbraucherschutz. 2005. August Hochwasser 2005 in Südbayern, 27 pp. Bebber WJ van. 1891. Die Zugstrassen der barometrischen Minima nach den Bahnenkarten der Deutschen Seewarte für den Zeitraum 1875–1890. Meteorologische Zeitschrift 8: 361–366. Begert M, Schlegel T, Kirchhofer W. 2005. Homogeneous temperature and precipitation series of Switzerland from 1864 to 2000. International Journal of Climatology 25: 65–80. Behrendt J. 1992. Dokumentation ROUTKLI: Beschriebung der Prüfkriterien im Programmsystem QUALKO (to be obtained from Deutscher Wetterdienst, Abteilung Klimatologie, Offenbach a.M., Germany), 12 pp. Bougeault P, Binder P, Buzzi A, Dirks R, Houze R, Kuettner J, Smith RB, Steinacker R, Volkert H. 2001. The MAP special observing period. Bulletin of the American Meteorological Society 82: 433–462. Brunetti M, Lentini G, Maugeri M, Nanni T, Simolo C, Spinoni J. 2012. Projecting North Eastern Italy temperature and precipitation secular records onto a high-resolution grid. Physics and Chemistry of the Earth 40: 9–22. Ceschia M, Micheletti S, Carniel R. 1991. Rainfall over Friuli-Venezia Giulia: high amounts and strong geographical gradients. Theoretical and Applied Climatology 43: 175–180. Daly C, Gibson WP, Taylor GH, Johnson GL, Pasteris P. 2002. A knowledge-based approach to the statistical mapping of climate. Climate Research 22: 99–113. Daly C, Halbleib M, Smith JI, Gibson WP, Doggett MK, Taylor GH, Curtis J, Pasteris PP. 2008. Physiographically sensitive mapping of climatological temperature and precipitation across the conterminous United States. International Journal of Climatology 28: 2031–2064. Daly C, Neilson RP, Phillips DL. 1994. A statistical-topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology 33: 140–158. DeGaetano AT, Wilks DS. 2008. Radar-guided interpolation of climatological precipitation data. International Journal of Climatology 29: 185–196, DOI: 10.1002/joc.1714 EEA (European Environment Agency). 2009. Regional climate change and adaptation. The Alps facing the challenge of changing water resources. EEA Technical Report, 143 pp, DOI: 10.2800/12552. Efthymiadis D, Jones PD, Briffa KR, Auer I, Bohm R, Schoner W, Frei C, Schmidli J. 2006. Construction of a 10-min-gridded precipitation data set for the Greater Alpine Region for 1800–2003. Journal of Geophysical Research-Atmospheres 111: D01105, DOI: 10.1029/2005JD006120 Erdin R, Frei C, Künsch HR. 2012. Data transformation and uncertainty in geostatistical combination of radar and rain gauges. Journal of Hydrometeorology 13: 1332–1346. Fliri F. 1974. Niederschlag und Lufttemperatur in Alpenraum. Wissenschaftliche Alpenvereinshefte 24, 110 pp. 2013 Royal Meteorological Society Fliri F, Schüepp M. 1983. Synoptische Klimatographie der Alpen zwischen Mont Blanc und Hohen Tauern. Wissenschaftliche Alpenvereinshefte 29, 686 pp. Frei C. 2006. Eine Länder übergreifende Niederschlags-Analyse zum August-Hochwasser 2005: Ergänzung zu Arbeitsbericht 211. Arbeitsberichte der MeteoSchweiz 213, 10 pp. Frei C, Christensen JH, Déqué M, Jacob D, Jones RG, Vidale PL. 2003. Daily precipitation statistics in regional climate models: Evaluation and intercomparison for the European Alps. Journal of Geophysical Research 108(D3): 4124, DOI: 10.1029/2002JD002287 Frei C, Germann U, Fukutome S, Liniger M. 2008. Möglichkeiten und Grenzen der Niederschlagsanalysen zum Hochwasser 2005. Arbeitsberichte der MeteoSchweiz 221, 19 pp. Frei C, Schär C. 1998. A precipitation climatology of the Alps from high-resolution rain-gauge observations. International Journal of Climatology 18: 873–900. Frei C, Schmidli J. 2006. Das Niederschlagsklima der Alpen: Wo sich Extreme nahekommen. promet, meteorologische Fortbildung (Deutscher Wetterdienst) 32(1/2): 61–67. Greminger PJ. 2003. Natural hazards and the Alpine Convention—Event analysis and recommendations. Federal Office for Spatial Development (ARE), 53 pp. Groisman PY, Legates DR. 1994. The accuracy of United States precipitation data. Bulletin of the American Meteorological Society 75: 215–227. Gyalistras D. 2003. Development and validation of a high-resolution monthly gridded temperature and precipitation data set for Switzerland (1951–2000). Climate Research 25: 55–83. Häggmark L, Ivarsson I, Gollvik S, Olofsson O. 2000. Mesan, an operational mesoscale analysis system. Tellus 52A: 2–20. Haiden T, Kann C, Pistotnik G, Bica B, Gruber C. 2011. The Integrated Nowcasting through Comprehensive Analysis (INCA) and its validation over the Eastern Alpine Region. Weather and Forecasting 26: 166–183. Haylock MR, Hofstra N, Klein Tank AMG, Klok EJ, Jones PD, New M. 2008. A European daily high-resolution gridded data set of surface temperature and precipitation for 1950–2006. Journal of Geophysical Research 113: D20119, DOI: 10.1029/2009JD011799 Hewitson BC, Crane RG. 2005. Gridded area-averaged daily precipitation via conditional interpolation. Journal of Climate 18: 41–57. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25: 1965–1978. Hilker N, Badoux A, Hegg C. 2009. The Swiss flood and landslide damage database 1972–2007. Natural Hazards and Earth System Sciences 9: 913–925. Hofstra N, Haylock M, New M, Jones PD. 2009. Testing E-OBS European high-resolution gridded data set of daily precipitation and surface temperature. Journal of Geophysical Research 114: D21101, DOI: 10.1029/2009JD011799 Hofstra N, Haylock M, New M, Jones PD, Frei C. 2008. The comparison of six methods for the interpolation of daily European climate data. Journal of Geophysical Research 113: D21110, DOI: 10.1029/2008JD010100 Hofstra N, New M, McSweeney C. 2010. The influence of interpolation and station network density on the distributions and trends of climate variables in gridded daily data. Climate Dynamics 35: 841–858. Holzkämper A, Calanca P, Fuhrer J. 2012. Statistical crop models: predicting the effects of temperature and precipitation changes. Climate Research 51: 11–21. International Innovation. 2011. Environment Monitoring a changing climate. In Confronting the Pressures on Ecosystem Services. International Innovation. August 2011, 16–18. Available at www.researcheurope.com/index.php/international-innovation/. Klein Tank AMG, Wijngaard JB, Konnen GP, Bohm R, Demaree G, Gocheva A, Mileta M, Pashiardis S, Hejkrlik L, Kern-Hansen C, Heino R, Bessemoulin P, Muller-Westermeier G, Tzanakou M, Szalai S, Palsdottir T, Fitzgerald D, Rubin S, Capaldo M, Maugeri M, Leitass A, Bukantis A, Aberfeld R, Van Engelen AFV, Forland E, Mietus M, Coelho F, Mares C, Razuvaev V, Nieplova E, Cegnar T, Antonio Lopez J, Dahlstrom B, Moberg A, Kirchhofer W, Ceylan A, Pachaliuk O, Alexander LV, Petrovic P. 2002. Daily dataset of 20th-century surface air temperature and precipitation series for the European climate assessment. International Journal of Climatology 22: 1441–1453. Klein Tank AMG, Zwiers FW, Zhang X. 2009. Guidelines on: Analysis of extremes in a changing climate in support of informed decisions for adaptation. World Meteorological Organization, Report WCDMP-72, WMO-TD 1500, Geneva, Switzerland, 52 pp. Int. J. Climatol. (2013) F. A. ISOTTA et al. Kleinn J, Frei C, Gurtz J, Lüthi D, Vidale PL, Schär C. 2005. Hydrological simulations in the Rhine basin, driven by a regional climate model. Journal of Geophysical Research 110: D04102, DOI: 10.1029/2004JD005143 Klok EJ, Klein Tank AMG. 2008. Updated and extended European dataset of daily climate observations. International Journal of Climatology 29: 1182–1191, DOI: 101002/joc.1779 Konzelmann T, Wehren B, Weingartner R. 2007. Niederschlagsmessnetze. Hydrological Atlas of Switzerland , HADES, available from University of Bern, Plate 2.1. Krajewski WF. 1987. Cokriging radar-rainfall and rain gage data. Journal of Geophysical Research 92(D8): 9571–9580. Legates DR, Willmott CJ. 1990. Mean seasonal and spatial variability in global surface air temperature. Theoretical and Applied Climatology 41: 11–21. LHG. 1991. Ursachenanalyse der Hochwasser 1987. Mitteilung der Landeshydrologie und -geologie 14, 184 pp. Available from www.bafu.admin.ch. Lussana C, Salvati MR, Pellegrini U, Uboldi F. 2009. Efficient high-resolution 3-D interpolation of meteorological variables for operational use. Advances in Science and Research 3: 105–112, DOI: 10.5194/asr-3-105-2009 Machguth H, Paul F, Kotlarski S, Hoelzle M. 2009. Calculating distributed glacier mass balance for the Swiss Alps from regional climate model output: a methodical description and interpretation of the results. Journal of Geophysical Research 114: D19106, DOI: 10.1029/2009JD011775 Martius O, Zenklusen E, Schwierz C, Davies HC. 2006. Episodes of Alpine heavy precipitation with an overlying elongated stratospheric intrusion: a climatology. International Journal of Climatology 26: 1149–1164. MeteoSwiss. 2006. Starkniederschlagsereignis August 2005. Arbeitsberichte der MeteoSchweiz 211, 63 pp. Mounier F, Lassègues F, Gibelin A-L, Céron J-P, Veysseire JM. 2012. Radar-guided control and interpolation of raingauge precipitation data over France. EURO4M report 2012 . Available at www.euro4m.eu/Publications Neff EL. 1977. How much rain does a rain gage gage? Journal of Hydrology 35: 213–220. Paulat M, Frei C, Hagen M, Wernli H. 2008. A gridded dataset of hourly precipitation in Germany: its construction, application and climatology. Meteorologische Zeitschrift 17: 719–732. Pavan V, Antolini G, Agrillo G, Auteri L, Barbero R, Bonati V, Brunier F, Cacciamani C, Cazzuli O, Cicogna A, De Luigi C, Maraldo L, Marigo G, Millini R, Panettieri E, Ratto S, Ronchi C, Saibanti S, Sulis A, Tomei F, Tomozeiu R, Torlai I, Villani G. 2013. The ARCIS project. Italian Journal of Agrometeorology (in press). Perčec Tadić M. 2010. Gridded Croatian climatology for 1961–1990. Theoretical and Applied Climatology 102: 87–103, DOI: 10.1007/ s00704-009-0237-3 Quintana-Segui P, Le Moigne P, Durand Y, Martin E, Habets F, Baillon M, Canellas C, Franchisteguy L, Morel S. 2008. Analysis of nearsurface atmospheric variables: Validation of the SAFRAN Analysis over France. Journal of Applied Meteorology and Climatology 47: 92–107. Rauthe M, Steiner H, Riediger U, Mazurkiewicz A, Gratzki A. 2013. A Central European precipitation climatology—Part I: Generation and validation of a high-resolution gridded daily data set (HYRAS). Meteorologische Zeitschrift(accepted) . Richter D. 1995. Ergebnisse methodischer Untersuchungen zur Korrektur des systematischen Messfehlers des HellmannNiederschlagsmessers. Bericht Deutschen Wetterdienstes 194, 93 pp. (To be obtained from German Weather Service, Offenbach a.M., Germany.) Rubel F, Rudolf B. 2001. Global daily precipitation estimates proved over the European Alps. Meteorologische Zeitschrift 10: 407–418. Schädler B, Weingartner R. 2002. Ein detaillierter hydrologischer Blick auf die Wasserressourcen der Schweiz. Wasser, Energie, Luft 94(7/8): 189–197. Schär C, Davies TD, Frei C, Wanner H, Widmann M, Wild M, Davies HC. 1997. Current Alpine climate. In Views from the Alps: Regional Perspectives on Climate Change, Cebon P, Dahinden U, Davies HC, Imboden D, Jäger C (eds). MIT Press: Boston, MA; 20–72. Shepard DS. 1984. Computer mapping: the SYMAP interpolation algorithm. In Spatial Statistics and Models, Gaile GL, Willmott CJ (eds). D. Reidel Publishing Company: Dordrecht, Netherlands; 133–145. 2013 Royal Meteorological Society Scherrer S, Frei C, Croci-Maspoli M, van Geijtenbeek D, Hotz C, Appenzeller C. 2011. Operational quality control of daily precipitation using spatio-climatological plausibility testing. Meteorologische Zeitschrift 20: 397–407. Schiemann R, Liniger MA, Frei C. 2010. Reduced space optimal interpolation of daily rain gauge precipitation in Switzerland. Journal of Geophysical Research 115: D14109, DOI: 10.1029/2009 JD013047 Schmidli J, Frei C, Schär C. 2001. Reconstruction of mesoscale precipitation fields from sparse observations in complex terrain. Journal of Climate 14: 3289–3306. Schmidli J, Goodess CM, Frei C, Haylock M, Hundecha Y, Ribalaygua J, Schmith T. 2007. Statistical and dynamical downscaling of precipitation: An evaluation and comparison of scenarios for the European Alps. Journal of Geophysical Research 112: D04105, DOI: 10.1029/2005JD007026 Schmidli J, Schmutz C, Frei C, Wanner H, Schar C. 2002. Mesoscale precipitation variability in the region of the European Alps during the 20th century. International Journal of Climatology 22: 1049–1074. Schmutz C, Arpagaus M, Clementi L, Frei C, Fukutome S, Germann U, Liniger M, Schacher F. 2008. Meteorologische Ereignisanalyse des Hochwassers 8. bis 9. 2007. Arbeitsberichte der MeteoSchweiz 222. Schwarb M. 2000. The Alpine Precipitation Climate: Evaluation of a High-Resolution Analysis Scheme using Comprehensive Rain-Gauge Data. Dissertation ETH 13911. Schwarb M, Daly C, Frei C, Schär C. 2001. Mean annual and seasonal precipitation in the European Alps 1971–1990. Hydrological Atlas of Switzerland , available from University of Bern, Bern, Plates 2.6 and 2.7. Sevruk B. 1985. Systematischer Niederschlagmessfehler in der Schweiz. Der Niederschlag in der Schweiz, Beiträge zur. Geologischen Karte der Schweiz-Hydrologie 31: 65–75. Sevruk B. 2005. Rainfall measurement: gauges. In Encyclopedia of Hydrological Sciences. Part 2, Hydrometeorology, Chapter 40, vol. 1. Anderson MG (ed). Wiley & Sons Ltd: Chichester, UK. 8 pp. Sevruk B, Zahlavova L. 1994. Classification system of precipitation gauge site exposure: evaluation and application. International Journal of Climatology 14: 681–689. Smiatek G, Kunstmann H, Knoche R, Marx A. 2009. Precipitation and temperature statistics in high-resolution regional climate models: evaluation for the European Alps. Journal of Geophysical Research 114: D19107, DOI: 10.1029/2008JD011353 Steinacker R. 1988. Die alpinen Hochwasserereignisse des Sommers 1987 und ihre meteorologischen Rahmenbedingungen. Österreichische Wasserwirtschaft 40(5/6): 129–134. Suklitsch M, Gobiet A, Leuprecht A, Frei C. 2008. High-resolution sensitivity studies with the Regional Climate Model CLM in the Alpine Region. Meteorologische Zeitschrift 17: 467–476. Tustison B, Harris D, Foufoula-Georgiou E. 2001. Scale issues in verification of precipitation forecasts. Journal of Geophysical Research 106: 11775–11784. UVEK. 2008. Hochwasser 2005 in der Schweiz—Synthesebericht zur Ereignisanalyse. Eidgenössicher Departement für Umwelt, Verkehr, Energie und Kommunikation UVEK , 24 pp. Available at www.bafu.admin.ch. Vidal J-P, Marin E, Franchistéguy L, Baillon M, Soubeyroux J-M. 2010. A 50-year high-resolution atmospheric reanalysis over France with the Safran system. International Journal of Climatology 30: 1627–1644. Villarini G, Mandapaka P, Krajewski WF, Moore R. 2008. Rainfall and sampling uncertainties: a rain gauge perspective. Journal of Geophysical Research-Atmospheres 113: D11102. Viviroli D, Weingartner R, Messerli B. 2003. Assessing the hydrological significance of the world’s mountains. Mountain Research and Development 23: 32–40, DOI: 10.1659/0276-4741(2003) 023[0032:ATHSOT]2.0.CO;2 Viviroli D, Zappa M, Schwanbeck J, Gurtz J, Weingartner R. 2009. Continuous simulation for flood estimation in ungauged mesoscale catchments of Switzerland - Part I: Modelling framework and calibration results. Journal of Hydrology 377: 191–207. Vogel R. 2013. Quantifying the uncertainty of spatial precipitation analyses with radar-gauge observation ensembles. Scientific Report MeteoSwiss 95, 80 pp. Weingartner R, Viviroli D, Schädler B. 2007. Water resources in mountain regions: a methodological approach to assess the water Int. J. Climatol. (2013) CLIMATE OF DAILY PRECIPITATION IN THE ALPS balance in a highland-lowland-system. Hydrological Processes 21: 578–585. Widmann M, Bretherton CS. 2000. Validation of mesoscale precipitation in the NCEP reanalysis using a new gridpoint dataset for the northwestern US. Journal of Climate 13: 1936–1950. Wijngaard JB, Klein Tank AMG, Können GP. 2003. Homogeneity of 20th century European daily temperature and precipitation series. International Journal of Climatology 23: 679–692. 2013 Royal Meteorological Society Wüest M, Frei C, Altenhoff A, Hagen M, Litschi M, Schär C. 2010. A gridded hourly precipitation dataset for Switzerland using rain-gauge analysis and radar-based disaggregation. International Journal of Climatology 30: 1764–1775. Yang DQ, Elomaa E, Tuominen A, Aaltonen A, Goodison B, Gunther T, Golubev V, Sevruk B, Madsen H, Milkovic J. 1999. Wind-induced precipitation undercatch of the Hellmann gauges. Nordic Hydrology 30: 57–80. Int. J. Climatol. (2013)