An Alpine Drought Impact Inventory to explore past droughts in a mountain region

. Drought affects even mountain regions, despite a humid climate. Droughts’ damaging character in the past and an increasing probability in future projections call for an understanding of drought impacts in the European Alpine region. The European Drought Impact Report Inventory (EDII) collects text reports on negative drought impacts. This study presents a considerably updated EDII focusing on the study region of the greater ‘Alpine Space’. This ﬁrst version release of an Alpine Drought Impact Inventory (EDII ALPS ) classiﬁes impact reports into categories covering various affected sectors and enables 5 comparisons of the drought impact characteristics. We analyzed the distribution of reported impacts on the spatial, temporal and seasonal scale, and by drought type for soil-moisture and hydrological drought. For the spatial analysis, we compared the impact data located in the Alpine Space’ to entire Europe. Further, we compared impact data between different climatic and altitudinal domains (Northern vs. Southern region, pre-Alpine vs. high-altitude region), and between the Alpine countries. Compared to entire Europe, in the Alpine Space agriculture and livestock farming impacts are even more frequently reported, 10 especially in the Southern region. Public water supply is second most relevant sector, but overall less prominent compared to Europe, especially in spring when snowmelt mitigates water shortages. Impacts occurred mostly in summer and early autumn with a delay between those impacts initiated by soil-moisture and those by hydrological drought. The high-altitude


Introduction
Droughts are natural hazards, which can cause widespread and severe impacts on the environment and societies. Compared to other weather-related hazards, such as floods and storms, droughts are among the most damaging events in terms of affected people and economic costs (Wilhite, 2000a;UNISDR, 2009;UNDRR, 2019). The summer droughts of 2003, 2015 and 2018 have raised concerns about the vulnerability of the European water budget to climate change Teuling, 25 2018), because these events have affected more than 17 % of the European population (Mastrotheodoros et al., 2020). Because of the humid mountain climate with an annual total precipitation between 400 to beyond 3000 mm/year (Isotta et al., 2014) and the four major European rivers Po, Rhone, Rhine, and Danube, the Alps are also called the "Water towers for Europe" . Nonetheless, past droughts caused severe impacts such as limited supply of water for drinking, irrigation and hydropower generation (Haslinger et al., 2019). The predicted increase of drought frequency, duration and extent stresses 30 the relevance of systematic analysis of drought impacts and their cascading effects in mountainous areas. This is particularly relevant within the Mediterranean climate in the Southern parts of the Alpine regions, where recent drought events triggered water disputes and spread of multiple impacts (Yves et al., 2020). The need to understand the role of drought impacts in Europe's mountainous region is stressed by the fact that more than 170 million people live within the major river basins . Until now and to the best of our knowledge, only the expert paper by the "Water management in the The different drought types generally lead to a wide range of impacts, making an impact assessment more difficult compared to other disasters. D M is typically understood as the prime trigger. While D M impacts in lowlands may often have compound causes with heat waves, e.g. excess mortality as a result of cardiovascular diseases, in cooler mountain regions such direct 55 impacts are likely less relevant. Most direct drought impacts can be linked to either D SM or D H . For example, low soil moisture typical for D SM initiates reduced vegetation health or crop quality, whereas low discharge and/or groundwater storage typical for D H causes problems in Public water supply (Wilhite and Vanyarkho, 2000). Impacts not directly induced by D SM or D H , are also known as 2nd-level or indirect impacts (Wilhite and Vanyarkho, 2000), such as increased costs due to supplementary irrigation.
In mountain regions economic sectors and priorities or activities may differ. These indirect impacts are the least tangible and 60 often related to D SE . In order to link drought impacts specifically to drought types, D SM and D H are the most evident types (Stagge et al., 2015). Despite the challenge to identify drought impacts, several efforts have been made predominantly focusing on the agricultural sector (Logar and van den Bergh, 2013;Poljanšek et al., 2019;Cammalleri et al., 2020), but not specifically on mountain regions or mountain-to-foothill transitions. Stahl et al. (2012) introduced the European Drought Impact Report Inventory (EDII) to widen the perspective to the broad scope of drought impacts including more sectors, such as Public water (12) as the link between drought, impact and management in these categories is often inconclusive and other databases, such as the Forest Fire Information system (EFFIS, 2020) are more comprehensive. The final categorization enabled the analysis of spatial distribution, temporal trends and differences by drought and impact type. The resulting dataset of this systematic 140 collection and classification of impact data within the Alpine Space in its first version is: EDII ALPS V1.0 available from the FreiDok repository (doi: tbd, last update 8th of January, 2021).
The distributions of all reports impact categories from EDII ALPS were compared with those of entire Europe (EDII EU ).
Within EDII ALPS , we compared pre-Alpine and high-Alpine, Northern and Southern Alpine regions, as well as political units: countries (respectively the part of the country within the Alpine Space), and NUTS 2 and 3 regions (Fig. 1). Because the 145 total numbers differ among these areas, we additionally focused on relative proportions ('fractions') of the 13 categories, i.e.
information that is independent of the overall data availability.
For the investigation of temporal trends, we aggregated EDII ALPS data per year and compared the proportions of the most frequent impact categories for the drought events 1976, 2003, 2015 and 2018. For the seasonal analysis, we pre-processed impact data as follows: Within the data collection, each impact was assigned to a month or season, in which the impact started 150 to occur. Most of the time, information about the end of an impact was unavailable. In this case we assumed that the impact only occurred in that month or season and not beyond. When a starting season is given for an impact, we assigned the season "spring" to the months March, April, May (MAM), "summer" to June, July, August (JJA), "autumn" to September, October, November (SON), and "winter" to December, January, February (DJF) for monthly impact data.
Different drought types may lead to different impacts, with D SM typically related to impacts in agriculture and D H typically 155 related to impacts within a range of several water uses, such as the Public water supply. In this study, we focussed on the D SM and D H type. D M often does not lead to impacts directly and D SE is challenging to relate to specific impacts, as they are the least tangible. Additionally, impacts on agriculture and on various water uses are among the most relevant concerning drought . To analyse timing, amount and the relevance of specific impacts, we re-grouped impact data to the D SM or D H type using the subtype category for assignment. For instance, within the category Energy and industry (4) subtypes are e.g. 'Reduced 160 hydropower production' (4.1) and 'Impaired production' [...] (4.2) . Regardless of the corresponding primary category, these subtypes were assigned to D SM and/or D H . For example, 'Reduced hydropower production' (4.1) is a result of low discharge and is therefore assigned to D H , whereas 'reduced productivity of annual crop cultivation' (1.1) is a result of low soil moisture and therefore assigned to D SM . Based on expert knowledge, four different people identified independently all 96 subtypes to D SM and/or D H . The newly grouped D SM and D H impacts were then analysed for seasonal occurrence in 165 the different domains (entire Europe, Alpine Space, pre-Alpine and high-altitude region, Southern and Northern region). We calculated smoothed seasonal "impact regimes" as loess curves by local regression (Cleveland, 1979).

Hypothesis testing
The following hypotheses on spatial differences were tested: the distribution of impact categories reported for the Alpine Space differed from that of the entire European region, the high-altitude region from that of the pre-Alpine region, the Northern region 170 from that of the Southern region. Additionally, we tested the hypothesis that the distribution of reported impact categories of For the statistical analyses, we used the compiled impact data as count data. We applied the Pairwise Wilcoxon Rank Sum Test to test whether the fraction of the counted data were significantly different. The test analysed if the central tendencies of the distributions between the groups differed (Cuzick, 1985). With the p-value ≤ 0.05, we rejected the null hypothesis that tendencies among the tested groups are the same and concluded a statistically significant difference between them. In addition, this test allowed us to identify which group(s) was (were) significantly different, if we tested more than two.

Spatial differences across domains and countries
At the time of this study EDII EU comprised in total around 10,600 reported drought impacts for NUTS 2 and 3 regions. The first version of EDII ALPS summed up to more than 3,200 impacts. We observed substantial differences between the amounts of reported impacts across the domain ( Fig. 2) with most reports located in the Northern and pre-Alpine region, followed by the 185 Southern and high-altitude region. The Pairwise Wilcoxon Rank Sum Test depicted significant differences between the impact category distributions of the Alpine Space and Europe, but not between the other pairs (Table 1). Among the Alpine countries, the test identified Slovenia to be significantly different from Austria and Germany. At the smaller scale, several NUTS 2 regions located in Italy differed to NUTS 2 regions in Austria, Switzerland and Germany.
In addition to the test results, relative distributions provide further information. The EDII ALPS and EDII EU report the same 190 two most frequent impact categories: Agriculture and livestock farming (EDII ALPS : 48 %, EDII EU : 31 %) and Public water supply (EDII ALPS : 21 %, EDII EU : 25 %). More than half of all reports on drought impacts belong to these two sectors in both regions. With 61 % this dominance is even stronger in the EDII ALPS (EDII EU : 56 %). The fraction of reported impacts in Agriculture and livestock farming was clearly higher in all domains in the Alpine Space compared to the average of Europe.
Especially, the Southern region presented almost 60 % of the reported impacts in this category, with more than half of all reports with 31 %, and on country-level France stood out with 39 %, both exceeding the European average. The most common subtype in this category was 'Bans on domestic and public water use' (7.3).
The third most frequent category in the EDII EU was clearly identified with Freshwater ecosystems (11 %). For the EDII ALPS ' third rank, we identified Forestry (7 %) with a slightly higher fraction than Freshwater ecosystems (6 %). 17 % of all entries the European average. Most of these impacts were located at the Rhine river and most frequent with the subtype 'Increased mortality of aquatic species' (9.1). The EDII ALPS ' impacts on Forestry (7 %) were as well mostly located in the German (19 %) and Swiss (10 %) part of the Alpine Space. In the EDII EU the category Forestry is ranked 6th with a fraction of 5 %. In both the EDII EU and EDII ALPS we identified the same most relevant subtype 'Reduced tree growth and vitality' (2.1).
The EDII ALPS presented these four described categories Agriculture and livestock farming, Public water supply, Freshwater 210 ecosystems and Forestry among the most important ones for the domains, but as well on the country-level with the following differences. As already mentioned, reports located in Slovenia were clearly dominated by the category Agriculture and livestock farming (96 %). More than 70 % of the impacts located in France (77 %) , Italy (78 %), and Austria (72 %) reported drought affecting Agriculture and livestock farming and Public water supply with switching relevance. Impacts located in Germany and Switzerland presented less dominance by these two categories, as Forestry and Freshwater ecosystems played as well a major 215 role.
Regarding the other categories and subtypes, we observed smaller differences. Impacts related to Waterborne transportation presented the 5th highest fraction for both, the EDII EU (9 %) and EDII ALPS (5 %). The relevance was lower in the Alpine Space, but high in the French part with 14 % and in the high-altitude region with 7 % of all impacts related to this category.
Whereas the impacts across entire Europe related most to the subtype 'Impaired navigability of streams (reduction of load 220 [...])' (5.1), the majority of the impacts in the Alpine Space were not tangible to a specific subtype. They related to 'Others' (5.3), with a majority from the French database, Propluvia mapping 'measures prohibiting valve operation, nautical activity', which could not be clearly assigned to any subtype of the category Waterborne transportation. In the EDII EU and EDII ALPS , the category Water quality presented a fraction of 7 % and 4 %, both with the most frequent subtype '(Temporary) Water quality deterioration / problems of surface waters [...]' (8.2). Thus, the frequency of this category was lower in the Alpine Space, and 225 mostly located in Italy with 7 % meeting the fraction by the EDII EU . The categories Air quality, Freshwater aquaculture and fisheries, Tourims and recreation, Terrestrial ecosystems, Energy and industry, Human health and public safety and Conflicts did not depict an obvious relevance in any of the analysed regions.

Temporal trends and drought years
Before the year 1950 the EDII ALPS only contains a small number of reported impacts (n: 244), covering eight categories, 230 dominated by Agriculture and livestock farming (n: 272) and followed by Public water supply (n: 23), Energy and industry

Seasonal patterns
All reported drought impacts occurred mostly in summer, followed by either spring (Southern region) or autumn (all other domains), and fewest in winter (Fig. 4). Significant seasonal differences were found in Europe and in the pre-Alpine and 255 high-altitude region in the Alpne Space (Table 2), always presenting the winter significantly differing from the summer. We identified low p-values for the summer differing from spring, but not between summer and autumn.
Although the seasonality of EDII ALPS and EDII EU is rather similar for the domains, some categories show differences.
Impacts related to Agriculture and livestock farming occurred in all seasons in EDII ALPS and it is the dominating category throughout the year excluding winter. Agriculture and livestock farming in the entire EDII EU on the other hand dominates 260 only during summer and Public water supply is the most dominant category at other times of the year (Fig. 4a,b). Regarding Agriculture and livestock farming, we observed most of these impacts in the summer in all subdomains as well, with a strong increase before and a decrease after the summer months. The pre-Alpine and Southern region show a clear increase of impacts in March and April (Fig.4c,f), and the high-altitude region has substantially higher counts in September and October (Fig. 4d).
In the Alpine Space, the spring and summer impacts in this category related most to the subtype 'Reduced productivity of [...] 265 crop cultivation' (1.1, 1.2), whereas the impacts in autumn related most to 'Reduced availability of irrigation water' (1.4). The EDII ALPS reported impacts in Public water supply were less frequent compared to the ones by the EDII EU , and especially in the first months of the year till May. The high-altitude region shows the most different pattern compared to Europe (Fig. 4b,d).
The monthly sums of this category do not increase in spring, but start to accumulate from May to August, and were less but still frequently reported in autumn. Especially in November and December, the reported impacts showed the same fractions as 270 those related to Agriculture and livestock farming. In the EDII ALPS the most frequent subtype 'Bans on domestic and public water use' (7.3) dominated this category, wheres in the EDII EU , the subtypes 'Local [...] and region-wide water supply shortage / problems' (7.1, 7.2) were as well prominent.
The EDII ALPS ' category Freshwater ecosystems reported most impacts in summer and autumn and almost none in spring and winter, with reports mostly located in the Northern region respectively pre-Alpine region (Fig. 4c,e). The EDII EU presented as 275 well most counted reports in the summer months, but the fractions of this category spread more equally across the seasons.
Further, we identified most impacts related to Forestry to occur in summer for both, in the Alpine Space and in Europe. Within the sub-regions of the Alpine Space, we observed the fraction of Forestry impacts in relative terms to be varying. We depicted the lowest relative fractions in autumn, which is due to higher counts in spring and summer especially in the Northern region ( Fig. 4e) and higher counts in winter in the high-altitude region (Fig. 4d). In the Alpine Space and in Europe, we found most 280 frequent reports in Waterborne transportation from high summer to early September and with highest seasonal fraction in autumn (Fig 4a,b). Impacts related to Tourims and recreation differed in their seasonal distribution between the domains. In the Alpine Space, these impacts were mostly winter impacts with a majority located in the high-altitude and Southern region ( Fig. 4a,d,f). Additionally, the high-altitude region showed higher fractions in spring for this category. Though the pre-Alpine and Northern region reported most impacts in summer, similar to Europe, these records are few compared to the more frequent 285 categories, such as Agriculture and livestock farming and Public water supply. We also found a few impacts on Air quality in the EDII ALPS and EDII EU . In the Alpine Space, this category together with Tourims and recreation were the only ones reported most in autumn (Air quality) and winter (Tourims and recreation).

Impacts related to drought types
We re-grouped 42 out of 96 subtypes by their related drought type D SM and D H (Fig. 5a). Twelve subtypes were classified as 290 D SM impacts with most of them from the categories Forestry and Agriculture and livestock farming. Further, we classified 32 subtypes as D H impacts with a majority from the categories Water quality, Public water supply, and Freshwater ecosystems.
Two subtypes were classified into both, the D SM group and the D H group: 'Lack of feed/water for livestock' (1.7) and 'Lack of feed/water for wildlife' (10.8), as low soil moisture as well as low discharge and/or groundwater storage are appropriate drivers.
For 54 subtypes neither D SM nor D H was considered a clear trigger. They were classified into a group of indirect impacts 295 that include impacts less tangible to specific drought conditions and strongly dependent on market situations, management and governance aspects. We found most of these ambiguous subtypes in the categories Energy and industry, Human health and public safety, Air quality, Conflicts, Freshwater ecosystems and the subtype 'Increased costs/economic losses' in several categories (all assignments are presented in Table S1).
With this classification scheme, the fraction of D SM and D H impacts differ clearly per domain (Fig 5). In the Alpine Space 300 36 % of all impacts were assigned to D SM and 43 % to D H . In Europe, D SM impacts were less (22 %) and D H impacts (56 %) more frequent (Fig. 5a,b). In the pre-Alpine region, 38 % of the impacts were assigned to D SM and 42 % to D H %. In the highaltitude region, we assigned less impacts to D SM (23 %) and more to D H (56 %) (Fig. 5c,d). The Northern and Southern region compared differed less with 35 % and 37 % assigned to D SM and 40 % and 49 % assigned to D H (Fig. 5e,f). The Southern region shows the greatest fraction of combined D SM and D H impacts (86 %) among all domains. In the D H impact group, the most frequent subtypes are 'Bans on domestic water use' (7.3), 'Local water supply shortage/problems' (7.1), 'Reduced availability of irrigation water' (1.4), 'Regional shortage of feed/water for livestock' (1.7), and 'Increased mortality of aquatic species' (9.1). We found the largest proportional differences between the pre-Alpine and highaltitude region among the subtypes 'Reduced availability of irrigation water' (1.4), and 'Bans on domestic and public water use' (7.3). Accordingly, we observed a shift among these subtypes between the Northern and Southern region, but additionally According to the total counts of the grouped impacts by drought types, the Alpine Space has a higher D SM impact peak occurring earlier in the year (June-July) than the D H impact peak (July-August) (Fig. 5a). In addition, the increase and decrease of the D H curve happens later in the year than that of the D SM curve. Thus, between March and July D SM impacts show higher 325 fractions, while between September and December D H impacts show higher fractions. The EDII EU contrasts these results (Fig.   5b), as the records across Europe depicted the higher peak of D H impacts in the same summer months as the D SM impact peak. Within the EDII ALPS the delayed D H peak and the higher fractions of D H impacts between September and December is confirmed by all subdomains (Fig. 5c,d,e,f). The most different pattern was found for the high-altitude region with the latest onset of the D H impact curve and a delayed peak between August and September (Fig. 5d). Furthermore, this was the only 330 domain within the Alpine Space with a higher D H impact peak, and subsequently, the highest fractions between July and December.

Drought impacts across the European Alpine region
Although the Alpine Convention has started to raise the topic, drought impacts in different regions of the Alpine and pre-Alpine 335 area have not yet been formally compared. Assembling the EDII ALPS , an inventory of drought impact reports, now enabled a first overview and some regional comparisons. Any collection of drought impact data is a challenging task because of the difficulties related to a clear definition of a drought impact's onset and termination for multifaceted causes (Bachmair et al., 2015). The chosen data sources proved to be suitable, as impacts were clearly linked to the droughts occurrence as the cause for the collected impact report. Some impacts can be measured and are therefore easier to collect and hence more consistent 340 through time and space (e.g., the agriculture yield losses), but most of them are hard to quantify (Logar and van den Bergh, 2013). For most of the impacts no continuous data is available, for which the text-reports proved to be a suitable surrogate and are worthwhile to collect (Bachmair et al., 2016;Hayes et al., 2012). Nonetheless, not all impacts are reported or only locally, in which case the compiled information in EDII ALPS V1.0 may still have gaps. Despite these barriers, the amount of impact may include Waterborne transportation to take place in the lowlands and Water quality benefitting from less pollution and environmental degradation. Other impact categories did not show substantial differences.

Drought events and temporal impact trends
Reported impacts suggest a diversification over the main drought years from 1976, 2003, 2015 to 2018, confirming observa-385 tions in earlier studies with the EDII EU . lower, but more continuous mortality of tree species occurring delayed due to several interdependent physiological mechanisms. We observed a similar but less prominent pattern for the category Waterborne transportation between the years 2003 and 2018 that might correspond to a longer lasting hydrological drought, with less discharge and/or groundwater storage over the years.
Over the whole time period, the number of collected drought impacts increased, especially after 2000 and 2010. This trend is 405 influenced by (1) general reporting behaviour changes with digitization and online media availability (regardless of the sources, such as scientific or newspaper articles or governmental reports), (2) accessibility to drought reports in the recent past being easier than to historic information ( as an example how international collaboration can customize existing databases such as the EDII with a moderate effort to make them useful in regions that have previously not really been a focus.
Our study presented the European Alpine region vulnerable to drought, despite the water-rich character. We compiled a great amount of impacts mostly related to agriculture and livestock farming followed by public water supply. These affected sectors are firmly established in the region, wherefore adaption and management strategies are essential for the future climate regimes.

490
Apart from the most relevant sectors, we found a surprising diversity of impacts covering a wide range of environmental, economic and societal effects that confirm the multifaceted character of drought in the Alpine Space. This growing diversity over time is likely due to the increasing complexity of the socio-economy in the Alpine Space with various sectors exposed and/or vulnerable to drought. In addition, the number of impacts increased substantially over time.
Key characteristics of drought impacts in the region are that impacts mostly occur in summer and early autumn regardless 495 of region, climatic condition or altitude. Typical to EDII ALPS are also some winter impacts related to ski tourism while spring impacts occur mostly in the southern region. The regions' specific snow accumulation in winter likely mitigates water shortages through snowmelt contributions in spring and early summer. Further, we could prove the possibility to link impacts to hydrological drought and soil-moisture drought in order to analyse drought specifics in different hydrological and climatic conditions. For the mountainous regions, we could demonstrate the delayed effect between impacts caused by soil-moisture 500 drought and hydrological drought, especially for the higher elevated region. All these seasonal effects of water redistribution and demand are essential to understand the hazard potential in different climatic and altitudinal zones for which our study presents a good starting point. Despite some biases in the current database, the amount of impacts we compiled in the EDII ALPS should raise awareness. Future climate predictions with increased drought severity, less snow and shift in precipitation patterns, suggest the European Alpine region will be vulnerable to drought impacts. This has to be taken into account for drought risk 505 and management assessments.
Data availability. We plan to publish the EDIIALPS data with a DOI.
Author contributions. Ruth Stephan, Mathilde Erfurt and Kerstin Stahl, designed the research. All co-authors provided data. Ruth Stephan moderated all EDIIALPS entries, carried out the analysis and created the graphs, maps and tables in the manuscript. Ruth Stephan prepared the manuscript with contributions from Mathilde Erfurt and reviews from all co-authors.