High-resolution maps of Swiss apiaries and their applicability to study spatial distribution of bacterial honey bee brood diseases

Apiaries

We obtained geodata (coordinates of apiaries) from the Swiss Federal Office for Agriculture (FOAG). However, among experts, the quality of these FOAG data is doubted. In order to verify the actual quality, we additionally collected raw data for every canton at their agricultural or veterinary offices in the end of 2016. For our main analyses, we created a data set with active apiaries (‘Switzerland_active’) containing mainly the cantonal data (not the FOAG data). Since data were not available for some cantons (Appenzell Outer-Rhodes, Appenzell Inner-Rhodes, Basel-City, Basel-District, Jura and Geneva), the cantonal data set was completed by the data of the FOAG resulting in the final ‘Switzerland_active’ data set. Hence, we had three data sets containing active apiaries in Switzerland in 2016:

1. FOAG_active (from Swiss Federal Office for Agriculture): used for ‘quality-index’

2. Cantonal_active (from cantonal authorities): used for ‘quality-index’

3. Switzerland_active (Cantonal_active completed by FOAG_active): used for main analyses

In the beginning, the geodata were cleaned up from errors. Geodata with incorrect coordinates or duplicates were removed or corrected, if possible. Depending on the canton, 1–5% of all geodata were deleted. The canton of Geneva (GE) had very low quality geodata, leading to the removal of 89% of the data points. For further calculations, inactive apiaries (not actively used by beekeepers) were removed and only the active ones were used. The metadata concerning occupation of apiaries (active/inactive) were nearly always available in the datasets. If not, we assumed that just the active ones were recorded.

The quality of the FOAG data set was quantified in GIS (ArcGIS Desktop 10.3, ESRI Inc.) by comparing the FOAG data set (‘FOAG_active’) with the data set of the cantonal authorities (‘Cantonal_active’). For this purpose, a ‘quality-index’ was calculated for each canton or municipality based on the ‘swissBOUNDARIES3D’ data set (Swisstopo, 2017). The ‘quality-index’ gives the proportion of the number of apiaries of the FOAG data set, relative to the number of apiaries of the cantonal data set. Thus, a ‘quality-index’ of 100% means that there is the same number of apiaries in both data sets. A ‘quality-index’ >100% indicates more FOAG-apiaries; <100% indicates more apiaries in the cantonal data set.

Based on the ‘Switzerland_active’ data set, maps were generated in GIS, representing the density of apiaries in 2016. Therefore, at cantonal (average size: 1,588 km²), district (average size: 266 km²) and municipal (average size: 18 km²) level ‘density-indices’ were calculated, giving the number of apiaries per area (km²) (Figs. 1 and 2). In addition to these classic geographic area analyses, an agricultural landscape type based analysis was run using the 55 different types of agricultural landscape defined for Switzerland (Federal Office for Agriculture of Switzerland, 2009; Szerencsits et al., 2009) (Fig. 3). To estimate dynamics of apiary densities during the last decade, the current data set at cantonal level was compared to older data (Fluri, Schenk & Frick, 2004). A GIS-map was used to illustrate the percentage change.

In order to investigate possible connections between the distribution of apiaries and occurrences of bee brood diseases (AFB and EFB), the distance to the nearest apiary has to be taken into consideration as well as apiary density. Therefore, GIS-maps were created, which specify the ‘average minimum distance’ to the nearest apiary per study area.

Apiary densities might directly be linked to human settlement or urban areas. To test the hypothesis that higher apiary densities correlate with higher human population densities (higher likelihood for beekeeping activity), a log–log transformed linear model (regression analysis in RStudio 1.0.136 (R 3.4.3); R Core Team, 2018; RStudio Team, 2017) was used, including the calculated apiary ‘density-indices’ and human population density on district level. Human population data came from ‘swissBOUNDARIES3D’ (Swisstopo, 2017) and were processed in ArcGIS.

Bee brood diseases

The data on honey bee brood diseases in Switzerland were provided by the InfoSM database of the Swiss Federal Food Safety and Veterinary Office—FSVO (FSVO, 2018). The data set used in this study contains all reported epidemiological cases of AFB and EFB since 1991 with information on diagnosis date and place of occurrence (canton, district and municipality). Based on the number of cases for each canton and the number of apiaries per canton, a ‘disease-index’ was calculated using the following mathematical model (mathematical derivation is given in Table S1): ${\displaystyle \text{‘disease-index’}\left[\text{\%}\right]=\left(1-{\left(1-\frac{\#cases}{\#months\times \#apiaries}\right)}^{\#months}\right)\times 100}$

The ‘disease-index’ shows the probability that a particular apiary in the study area was affected at least once during a given period. Testing the hypothesis that bee brood diseases occur more frequently in regions with higher apiary density, a linear model (regression analysis in RStudio) was used including the calculated ‘density-’ and ‘disease-indices’.

However, regression analysis at cantonal level should be interpreted with caution as apiary density within cantons is often distributed very inhomogeneously (Fig. 2). To account for this skewed distribution three cantons were examined at higher resolution (district level). For the following reasons, the cantons of Aargau (AG), Bern (BE) and Solothurn (SO) have been selected as case study objects: (1) very inhomogeneous apiary distribution; (2) extreme values (maxima, minima) for apiary densities in national comparison; (3) high number of reported epidemiological cases and number of apiaries (large sample size); (4) high data quality (‘quality-index’ approximately 100%) and (5) AG, BE, SO are a contiguous study area (exclaves –non-contiguous areas –are not included in the case study).

Apiaries

Comparing the FOAG data (‘FOAG_active’) with the cantonal data (‘Cantonal_active’), to evaluate the quality of the FOAG data, showed that in almost half of all cantons the number of registered apiaries differ only in less than ±5% (Figs. S1 and S2). The other half of the cantons show larger differences, in some cases the ‘quality-index’ is even lower than 50% (e.g., Lucerne: 37%; Zug: 41%; and Neuchâtel: 49%). This locally high discrepancy between FOAG and the cantonal data resulted in the generation of a combined, corrected data set called ‘Switzerland_active’ (for details see ‘Material and Methods’).

This corrected data set used for further analyses, consisting primarily of the cantonal data, comprised in total 23,055 apiaries (for details see Table 1). Based on this data set, an average apiary density (‘density-index’: DI) of 0.56 apiaries per km² was calculated for Switzerland for the year 2016. The highest densities were found in northern central Switzerland (maximum values: Zurich, DI 1.27 and Zug, DI 1.56; Table 1, Fig. 1). The cantons with the lowest ‘density-indices’ are those with high alpine share (minimum values: Uri, DI 0.12 and Grisons, DI 0.16; Table 1, Fig. 1).

In contrast, on the high-resolution map on municipal level (Fig. 2), areas with occasionally very high apiary density became visible, which were, due to the low resolution, not displayed on the map at cantonal level (Fig. 1). For instance, in the center of the canton of Bern, an area with consistently high apiary density can be found. There, the municipalities have a density of more than two apiaries per km² (Fig. 2). In a general comparison, it became obvious that in the Swiss midlands and in the pre-Alpine regions, the density of apiaries is higher on average, than in the Alps and the Jura (Table 1; Figs. 1 and 2, S3) (for details according to spatial classification of Swiss landscapes see Szerencsits, 2013). These differences that are related to different regions and human usage were visible, with the highest resolution, on the map including Swiss agricultural landscape types (Fig. 3A). Highest apiary densities are in regions used for fruit and vine cultivation, settlements (e.g., Bern), mixed use and fodder cultivation; whereas lowest densities can be seen for mountainous areas and lakelands (Table 2, Fig. 3B). These differences are highly significant (Kruskal-Wallis ANOVA: H = 410.78, df = 8, p < 0.0001) comparing apiary ‘density-indices’ between nine landscape classes, summarizing the 55 different types of agricultural landscape defined for Switzerland (Table 2, Fig. 3B).

The comparison of the current density map with the historic data of Fluri, Schenk & Frick (2004) resulted in very heterogeneous percentage changes, varying between −70% (Basel-City) and +103% (Ticino) (Fig. S4). At the moment, putative reasons explaining this high variance are highly speculative and will not be included in the current study. The most obvious factor explaining this observation, is the difference in apiary registration, being compulsory nowadays.

The ‘average minimum distance’ to the nearest apiary is 459.6 m. On cantonal level it varies between 385 m (Zug) and 655 m (Obwalden) (Table 1, Fig. S5). In principle, cantons with a higher ‘density-index’ tend to have a smaller ‘average minimum distance’ (linear regression; R² = 0.51, p < 0.0001). However, this is not always the case as apiaries are not distributed homogeneously. In mountain regions, for instance, apiaries are often arranged linearly along the agricultural and populated valleys (Fig. 3A, S6, S7). This is evident in the canton of Valais: the ‘average minimum distance’ is very low (396 m), although the ‘density-index’ is not high (DI 0.27) (Table 1).

Comparing apiary density on high resolution level (district) and the human population density, a highly significant correlation was detectable (linear regression; R² = 0.64, p < 0.0001). The higher the human population density, the more likely higher numbers of beekeepers can be seen, resulting in higher apiary densities (Fig. S8).

Bee brood diseases

In the period between 1991 and 2016, 9500 cases of bee brood diseases were registered in Switzerland. 81% (7221) of all cases were EFB, 19% (2279) were AFB (Table 1). The temporal analysis of the data confirms previous statements by Roetschi et al. (2008) that EFB has been growing strongly in Switzerland since 1999. It peaked in 2009/10 with 796 cases in 2009 and 987 cases in 2010 respectively. Since 2010, the number of EFB cases is decreasing down to 347 (2015) and 371 (2016) (Figs. S9, S10).

AFB is still declining since 2003 (Fig. S11). Between 1991 and 2003, there were 1403 cases of AFB in Switzerland (62% of all outbreaks), and from 2004 to 2016 there were 876 AFB cases (38% of all outbreaks). While in 1991, 75% of all brood disease cases fell to AFB (65 cases) and only 25% to EFB (21 cases), the increase in EFB cases showed that from 2010 until now, only about 10% of all cases were AFB and about 90% EFB (Fig. S9).

Spatial investigations revealed that in the case of AFB no canton stands out by a higher number of cases than other cantons (Table 1, Fig. S11 ). In stark contrast, a canton stands out clearly in the case of EFB: the canton of Bern with 3104 EFB cases between 1991 and 2016, 43% of all EFB cases in Switzerland (Table 1, Fig. S10). However, the EFB epizootic in Switzerland around the year 2010 was not just a phenomenon in this canton. In most other cantons there was also a high incidence of EFB during this time. Consequently, the disease data (AFB and EFB) of all cantons can be used as a resilient model input to investigate possible interactions between apiary density and disease occurrence. Table 1 and Table S2 show the variables to be analyzed: the density of apiaries and ‘disease-indices’ for AFB and EFB at cantonal (Table 1) and district (Table S2) level.

On cantonal level, there is hardly any evidence for a correlation between apiary density (‘density-indices’) and occurrence of both bee brood diseases (‘disease-indices’) (linear regression; R² < 0.001, p > 0.9, Table 3). Since apiaries within cantons are distributed very heterogeneously, statistical analyses at cantonal level have to be interpreted with caution. Therefore, further analyses were conducted on district level for three cantons. In the case of AFB, a positive but not significant interaction could be observed (linear regression; R² = 0.08, p = 0.14, Table 3). Contrary, in the case of EFB, a highly significant correlation (linear regression; R² = 0.31, p = 0.002, Table 3) between apiary density and occurrence of EFB was detectable. For the three cantons tested on district level (AG, R² = 0.54; p = 0.009; BE, R² = 0.69, p = 0.003 and SO, R² = 0.41, p = 0.09), the pattern of higher apiary densities (‘density-indices’) correlating with higher EFB occurrence (‘disease-indices’) was nearly the same (Table 3).

Apiaries

The distribution of apiaries at municipal level is very heterogeneous (Fig. 2). At first sight, Alpine cantons (e.g., Grisons and Uri) have a very low apiary density. This is in line with Crane (1992) who found that the upper limit of permanent colonies in the Alps is at an altitude of about 1500 m above sea level. Fluri, Schenk & Frick (2004) and Fluri & Frick (2005) assumed that the low apiary density is linked with the generally lower human population density in these areas and thus a lower number of potential beekeepers. Here, we found as well a very high correlation (Fig. S8) and the lowest apiary density in mountainous areas (Table 2, Figs. 3A, Fig. 3B), supporting this assumption. The spatial analysis in our study revealed that in the Alpine area of Switzerland, apiaries are arranged linearly along the agricultural and populated valleys. In addition, the limited number of suitable sites for apiaries (steep terrain, late sun exposure in spring, avalanche risk) and reduced economic advantages (limited honey production, starvation risk during summer) might be additional reasons for the lower number of beekeepers.

The high apiary densities near midland lakes (e.g., Thurgau, Zug and Zurich) might be explained by the high human population density close to these lakes, but also by the topographically favorable conditions for apiaries and the abundance of foraging sources for bees (Klein et al., 2007), e.g., fruit trees (Lake Constance), vegetable production (Moos near Lake Biel) and vineyards (Lake Biel, Lake Geneva). Even that grapes (Vitis vinifera) are self-pollinated and do not depend on honey bees (Kassemeyer & Staudt, 1981), bee colonies are kept in the vicinity of vineyards, as honey bees potentially help to control the grape moth (Eupoecilia ambiguella, Lobesia botrana) and pollinate fruit trees, which are often planted at the edges of vineyards (Tautz & Rostas, 2008). Both assumptions could be confirmed by the highest apiary densities in areas with settlement, fruit and wine cultivation (Table 2, Figs. 3A, 3B).

Within the scope of this study, the apiary density was calculated based on the number of apiaries divided by the surface of Switzerland, the canton, the district and the municipality, respectively. Fluri, Schenk & Frick (2004) also calculated honey bee densities. They used the number of colonies, the total surface of Switzerland (41,285 km²) and the cantons, and found 4.7 colonies per km² in 2003, with mountains and lakes included. Although the numbers from their study (colony density) are not directly comparable to our study (apiary density), the apiary density in 2003 can be estimated by assuming that each beekeeper owns one apiary with approximately 10 colonies (Fluri, Schenk & Frick, 2004). This means in 2003 Switzerland had an average apiary density of 0.47 per km², being slightly below the density calculated here (0.56 apiaries per km²).

According to Fluri, Schenk & Frick (2004) and the FAO (2018), the number of Swiss colonies has been declining since World War II. This development should also be observed in the number of apiaries. However, comparing the data from Fluri, Schenk & Frick (2004) with the cantonal data of 2016 of this study, not such a trend became obvious (Fig. S4). Most of the cantons have a much higher density of apiaries in the more recent data, though these unexpected results call into question the quality of the data, as indicated by Moritz & Erler (2016) for comparable data sets. However, we have to emphasize that in 2003 it was not yet mandatory to register apiaries. Therefore, it can be assumed that not all apiaries were recorded at this time.

Bee brood diseases

AFB and EFB, both caused by Gram-positive bacteria (AFB: Paenibacillus larvae; EFB: Melissococcus plutonius), are widely distributed in Switzerland (Forsgren et al., 2018; Gillard, Charriere & Belloy, 2008; Roetschi et al., 2008) and both diseases are potentially lethal to the brood of infected colonies (Forsgren, 2010; Forsgren et al., 2018). Adult bees can act as carriers and spread (in combination with the beekeeper) the disease between colonies and apiaries (Belloy et al., 2007; Lindström, Korpela & Fries, 2008).

Belloy et al. (2007) showed in a field study that almost 40% of bee colonies, between 500 and 1,000 m from a bee colony with EFB symptoms, were infected with M. plutonius. This is about the same distance which was found in this study (using data for all of Switzerland) as average minimum distances between apiaries in Switzerland (350–650 m) (Table 1). With the assumption that within apiaries colony distances are usually much lower than between apiaries, the data of both studies showed that pathogen transmission can be highly successful between apiaries and will be even more successful between colonies of each apiary. Belloy et al. (2007) and Forsgren et al. (2005) pointed out that EFB epizootics in Switzerland might have benefited from the high density of bee colonies, which promotes the transmission of the pathogens by adult bees. Further, Forsgren et al. (2005) supposed that the high density leads to nutrition competition between colonies and increases the risk for EFB. According to Belloy et al. (2007), an increased density leads to shorter distances between apiaries, which means that M. plutonius can spread easier. Hence, the spatial distribution of apiaries seems to support pathogen transmission by bees and might explain partly the EFB epizootics in Switzerland. Although there is hardly any statistical correlation between EFB outbreak and apiary density at cantonal level, the case study at district level shows a highly significant, positive correlation between the density of apiaries and the probability of disease outbreaks (Table 3).

However, if there are no, or hardly any, infectious M. plutonius bacteria in a region, even a high apiary density does not lead to an increased probability of disease outbreaks. This can be seen in the cantons of Basel-City (0 EFB cases), Jura (2), Neuchâtel (1), Schaffhausen (5) and Ticino (15), and might explain why there was no correlation between apiary density and disease probability in the analysis at cantonal level. The absence of a clear effect at cantonal level might also be explained by the fact that some honey bee colonies may be disease resistant (Spivak & Reuter, 2001). In this case, the disease would be spatial enzootic (as known from other European countries, (Forsgren et al., 2018)) whereas in other regions it remains epizootic. Furthermore, the pure presence of the causative agent does not automatically cause a disease outbreak (Belloy et al., 2007; Forsgren et al., 2005). To test the hypothesis that an increased density may not necessarily lead to an increased probability of infection, but to a faster spread of a regional epizootic, all cases of colony infection would have to be provided with precise coordinates (Datta et al., 2013).

This study shows that the assumption of an increased probability of disease outbreaks for bacterial bee brood diseases in regions with high apiary density seems not to apply in the case of AFB. There is no significant correlation at either district or cantonal level (Table 3). It should be noted that in the case of AFB, sample size is relatively small, leading to lower statistical power and the result must be interpreted with caution. Another explanation for the differences between EFB and AFB might be differences in handling the two bee brood diseases. In the case of AFB, a restricted area is defined, which normally covers an area within a radius of 2 km from the contaminated status. All apiaries within the restricted area are checked and any import and export of bees and honeycombs is prohibited for at least 30 days. In the case of EFB, the radius is only 1 km, covering an area four times smaller than the restricted area in the event of an AFB case (TSV, 2018). Taking into account the foraging range of honey bees, the restriction circle of 1 km around an EFB infected apiary might be insufficient to prevent pathogen transmission. A. mellifera has a normal foraging range of up to 1.5 km (Beekman et al., 2004; Danner et al., 2016; Danner et al., 2017; Visscher & Seeley, 1982). Lindström, Korpela & Fries (2008) could show that AFB is transmitted almost exclusively within 2 km. The restriction circle of 2 km around an AFB-affected apiary should therefore be sufficiently large to limit the spread of the bee brood disease under normal conditions, even though a complete prevention cannot be guaranteed. In future studies, restriction circles for bacterial bee brood disease outbreaks have to be compared for different European countries and their effectiveness to prevent pathogen transmission. Restriction circles even larger than 2 km might be a future solution. However, with the high apiary density of Switzerland, economic consequences should be considered carefully.