**1. Introduction**

*Apis mellifera*, the western honey bee, is not only an important pollinator for wild flowers but also crucial for the pollination of the world agricultural production. It was calculated in 2005 that the estimated economic value of pollinating insects in the European Union equals 14.2 billion euro [1]. The threats for honey bees are increasingly well studied and understood and range from abiotic stressors, such as pesticides, to biotic stressors [2–4]. The latter include parasites, pathogens, and pests, like the parasitic mite *Varroa destructor* or monoculture plantings that influence the quality and richness of forage sources [5]. Further, synergistic effects of single factors may add up to a threat that is greater than the sum of its individual factors. Beekeepers try their best to support their colonies by optimizing their hive managemen<sup>t</sup> to help them cope with the environmental conditions they are facing. Comparisons of hive managemen<sup>t</sup> practices of large data sets collected from beekeepers have demonstrated different strategies and consequently different overwintering success [6–9]. For example, efficient treatments of colonies against varroa mite are necessary, as this parasite is known to reduce winter survival [10–12].

Honey bee monitoring via citizen scientist beekeepers or crowdsourcing has been carried out for a couple of years in Austria [13] and other countries [6,7]. It helps to identify potential risk factors and provides a data base to gain a better understanding of honey bee colony losses. Thanks to the nonprofit honey bee research association COLOSS (Prevention of honey bee COlony LOSSes) [14], monitoring of honey bee colonies is done in various European and some non-European countries

following international standards [15]. This facilitates the comparison between countries and a joint identification of risk factors [16,17]. The analysis of this data has already revealed successful practices for colony managemen<sup>t</sup> and identified unfavourable biotic factors, such as crops which negatively influence colony survival [18]. In Austria, a substantial data set, which has been acquired over several years, has already been used to investigate the effect of weather [19] and land use [20] on colony losses.

In contrast to previous COLOSS publications [7,16] that have combined loss rates from countries with quite different environmental conditions and hence are limited in analyses of regional risk factors, we present a complete examination of the data collected within COLOSS for Austria. This complements the last comprehensive risk analysis on Austrian honey bee colony winter mortality published a while ago [13]. We present a study using a large data set obtained by crowdsourcing for winter 2018/2019 in Austria and a substantially improved methodology. The aim is to report internationally comparable loss rates, to identify factors which have negative or positive effects on honey bees, and to reappraise different hive managemen<sup>t</sup> techniques.

#### **2. Materials and Methods**

#### *2.1. Survey Design and Response Rate*

Our survey was based on the questions of the international COLOSS questionnaire, which has been translated to German. The questionnaire asks for the number of honey bee colonies wintered and the number of colonies lost with three possible categories (colony dead or reduced to a few hundred bees, lost due to natural disaster (i.e., flooding, falling trees, and vandalism), or alive but with unsolvable queen problems (drone laying queen or no queen at all)) [15]. The number of colonies lost due to natural disaster was not included in the total loss calculation and risk analysis because it is not directly related to biological (i.e., age of queen bees) or operational risk factors in the survey.

Questions on hive managemen<sup>t</sup> practices are also part of the COLOSS international survey and were used unmodified. These questions concerned topics like the number of the wintered colonies that had a new queen (born and mated 2018), observed queen problems in colonies during foraging season compared to previous seasons, certified organic beekeeping, queens bred from varroa tolerant/resistant stock, hives fabricated from synthetic materials, insulated hives, screened bottom boards, purchase of wax from outside of the own operation (a measure of professionalism and wax quality), natural comb without foundation, small brood cell size, migratory beekeeping, replacement of brood frames (in relative percentage categories), notice of bees with crippled/deformed wings (often, seldom, none, and do not know), and foraging crops perceived by the beekeeper. As for the foraging crops, the participants were asked if the majority of their bee colonies had significantly foraged on one or more of the asked crops (only crops relevant for Austria were listed). The selection of some of these operational factors followed suggestions from beekeepers in previous surveys or discussion groups and were accepted for international use in COLOSS.

An important part of the survey was dedicated to varroa control. Beekeepers could identify if and in which month they monitored varroa infestation level. Additionally, the methods and application time of varroa control were surveyed.

The estimated percentage of beekeepers participating in the survey was calculated with the number of beekeepers and colonies registered at the national beekeeping association (Table 1). Data collection was carried out via an online survey (LimeSurvey Version 3.16.1+) and a questionnaire published in a beekeeping journal and physically handed out at beekeeper meetings. This ensured that also beekeepers without an internet connection had access to the survey [15]. Overall, the survey followed the guidelines of the COLOSS Project rather closely [15]. To guarantee the protection of the participants private data, a data privacy protocol was established between the people involved in the project. All private data was removed when working on the analysis of received survey data and only used to enquire questionable data.


**Table 1.** Number and percentage of survey participants in relation to the amount of registered beekeepers with the national beekeeping association "Biene Österreich" (2018): Commercial beekeepers are subsumed as "not specified" due to the absence of state data.

#### *2.2. Data Validation and Error Control*

If beekeepers did participate via paper questionnaire, the data was manually transferred to a Microsoft Excel file, where all survey data was collected. Automatic checks with simple formulas in Excel were used to minimize processing errors and to highlight possible invalid responses, i.e., more colonies lost than existent, as described in References [15,16]. These contradictory entries or multiple entries of the same beekeeper were removed. The survey in the beekeeping journal did not cover all the questions that were asked in the internet survey, and omitted answers also led to a reduced number of responses for several questions. If the response for one factor was not enough for statistical analysis, the data was not used. The participants did not disclose the exact location of their apiaries, only the rough area of their main wintering apiary, i.e., district and zip code at minimum. Coordinates for plotting were generated with a batch geolocation finder and, if not feasible, by a manual search. The resulting locations were plotted on a district map and tested for correct assignment to districts as given by the participants in the survey, to minimize incorrect geolocations. The estimation of elevation for the apiary locations was carried out with the topography model SRTM-3v4 via a web service [21].
