Introduction

The progressive increase in the human population, as well as human-mediated activities, e.g., land-use change due to an expansion of agriculture (Laurance et al. 2013), habitat degradation (Kerr and Deguise 2004), invasion of alien species (Clavero and García-Berthou 2005), and plastic pollution (Chae and An 2018), have negative consequences on species survival and ecosystem functioning worldwide (Pereira et al. 2010). However, one of the most significant global biodiversity threats is climate change (Sala et al. 2000). Climate change alters the key abiotic agents of selection, i.e., CO2, temperature, and precipitation (Arias et al. 2021), transforming current ecosystems and food webs (Schmitz et al. 2003; Alava et al. 2017; Bartley et al. 2019; Pound et al. 2020). It may affect species range shifts, leading to a spatial mismatch between organisms and decoupling their interactions (Nakazawa and Doi 2012; Miller-Struttmann et al. 2015; Damien and Tougeron 2019). Climate change may directly or indirectly influence plant and animal species distribution shifts (Travis et al. 2013). Understanding the effect of future climate change on biodiversity is important not only in conservation biology (Tam and McDaniels 2013), but also for human health and wellbeing (Jones et al. 2012; McMichael 2013), and the global economy (Newell and Paterson 2010). Projected range expansion of insect pests (Battisti and Larsson 2015), vectors of human diseases (Bueno Marí and Jiménez-Peydró 2013), and invasive alien species (Anibaba et al. 2022) are prime examples of this importance.

Insects are sensitive to increased temperature (Beck 1983), and consequently to the effects of global climate change (Halsch et al. 2021). Global warming affects insect ecology, physiology, and behavior (Pelini et al. 2009) and causes mismatches in their trophic mutualistic and antagonistic interactions with plants and other animals (Blois et al. 2013). Due to rapid climate change, several life-history traits such as survival and growth rate are likely to change (Bale et al. 2002). This might also alter the population size and temporal and spatial dynamics of insect communities (Pelini et al. 2009; Raven and Wagner 2021) and cause shifts in their ranges (Arribas et al. 2012; Hill et al. 2017; Martín‐Vélez and Abellán 2022).

Global warming is also linked with changing predator–prey interactions (Laws 2017). Changes in local community composition over space and time due to shifting species ranges or the creation of novel communities may have consequences in population dynamics for both predator and prey. This may affect ecosystems in many ways, as seen with the predator effect on pollination — plant interactions (Damien and Tougeron 2019). Several studies have shown that predators can be a major source of mortality of pollinators, decreasing pollinator population size as well as reducing the pollination effectiveness on seed set by reducing the density of pollinators (Dukas 2001a, b, 2004; Dukas and Morse 2003; Muñoz and Arroyo 2004; Jones and Dornhaus 2011). Understanding how climate change will contribute to the range expansion of predators could help prevent the local extinction of pollinator communities.

Evidence suggests that climatic changes will significantly, and in some cases irreversibly, harm the economy and many ecosystems alike (Willetts et al. 2010)⁠. Widely available maps of current and future climatic conditions and the rapidly increasing number of species observations in scientific and citizen science databases allow for increasingly extensive studies of spatio-temporal changes in species distributions (Puchałka et al. 2021, 2022). Species distribution modeling is a widely used tool for predicting potential changes in the distribution range of plant and animal species and detecting niche shifts with climate change (Guisan and Thuiller 2005; Hill et al. 2017; Dyderski et al. 2018). Recent studies of the effect of climate change on future species distribution have focused on native and on potentially invasive alien insect pests in agriculture and forest ecosystems (Hlásny and Turcani 2009; Cudmore et al. 2010; Qin et al. 2019). Future projections suggest a northwards range expansion of pest insect species in the Northern Hemisphere and southwards in the Southern Hemisphere, where a more suitable niche will be available (Qin et al. 2019; Cornelissen et al. 2019; Ramasamy et al. 2022). For example, Cornelissen et al (2019) predicted that the distribution of Aethina tumida, an invasive pest of the honeybee endemic to sub-Saharan Africa, will expand into the Northern Hemisphere and potentially become a global threat. In contrast, one of the most important insect groups, pollinators, that provide an ecosystem service for the human food production sector, seems to be mostly negatively affected by climatic warming (Potts et al. 2010; Goulson et al. 2015; Imbach et al. 2017).

Climate change may increase the expansion of invasive pest species and diseases as well as having negative consequences on behavior and physiology that may lead to diversity loss in bees (Potts et al. 2010; Polce et al. 2014; Cornelissen et al. 2019; Carrasco et al. 2020; Martínez López et al. 2021). However, knowledge of the potential effect of climate change on bee predators, such as digger wasps (e.g., Philanthus spp.), is currently lacking (but see Eid and Abou-Shaara 2021).

The European beewolf Philanthus triangulum (Philanthidae) is a medium-sized digger wasp (female 13–17 mm, male 8–10 mm in length), whose distribution stretches from the Atlantic coast of Europe to the Middle East and from Scandinavia in the north to South Africa in the south (Bitsch et al. 1997; Blösch 2000; Pulawski 2021; Sann et al. 2018). In Europe, both males and females are active from mid-June to October. Males usually appear a week before females and stay at the edge of the breeding sites. Females choose sandy sites for their nests (in both horizontal and vertical surfaces) and dig their burrows using their legs and mandibles. Adult European beewolves feed on pollen and nectar from flowers but can also obtain nectar and pollen by squeezing them from the bodies of their prey (Blösch 2000). The food of the larvae consists of bees usually attacked at flowers, but they may also be taken in flight. The prey is paralyzed by being stung, always behind the first pair of legs (Rathmayer 1962). European beewolf is generally considered to be economically harmful because it mainly hunts honeybees (Apis mellifera). Tinbergen (1958) showed that an aggregation of European beewolf can capture several thousand honeybees per day. Simonthomas and Simonthomas (1980) concluded that an aggregation of 3000 European beewolf individuals, captured up to 30,000 honeybees per day. The number of burrows in nest aggregations can be up to 15,000 (Else 1995a, b). This observation was confirmed for another bee wolf species, Philianthus bicinctus, in northwest Wyoming, USA, where it was estimated that the wasp aggregation captured several hundred to a few thousand bumblebees per day (Dukas 2004). Hamm and Richards (1930) reported that European beewolf can also hunt solitary bees, e.g., Andrena flavipes or Lasioglossum zonulum (Smith 1858; Fahringer 1922). Some studies have shown that this species can locally affect beekeeping due to their use of honeybees as larval food (El-Borollosy et al. 1973; Simonthomas and Simonthomas 1980; Petrov 1996). The European beewolf can also impact populations of wild bees and other pollinators, especially where few honeybees are present (Blösch 2000). It is conceivable that the European beewolf may become a serious threat for the honeybees in areas where it was not previously present.

The distribution of European beewolf is susceptible to climatic fluctuations, e.g., temperature (Braestrup and Nielsen 1941; Hansen 1997). The entire population of this species may decline due to long-term adverse weather conditions (Blösch 2000)⁠. The European beewolf is distributed mainly in the mid latitudes and the tropics. Hence, we hypothesized that this species may benefit from predicted climate change and, in the future, could become a potential threat for honeybees in Europe. In this study, we aimed to determine the impact of climate change in the 2040–2060 and 2060–2080 periods on the potential distribution of the European beewolf, a specialized predator of bees.

Materials and methods

European beewolf occurrence data

The occurrences of European beewolf (Supplementary material, Fig. 1) were obtained from scientific papers, Global Biodiversity Information Facility (www.gbif.org) and confirmed identifications of georeferenced photographs on the webpage Biodiversidad Virtual (www.biodiversidadvirtual.org)⁠⁠⁠. The gaps in occurrence data were filled by the authors’ unpublished field observations. In total, we obtained 10,574 distribution records with associated geographical coordinates (latitude and longitude), which were loaded into a spatially referenced database. Among all records, 53.6% were collected after 2015, 84.4% after 2005, and 92.9% after 1990. To overcome any disadvantage of citizen science data collection resulting in uneven sampling that could affect analyses, we randomly selected a single record from each 0.25° grid raster cell for MaxEnt analysis (Rocchini and Garzon-Lopez 2017). This resulted in 1021 single observations to model current and future potential distribution of European beewolf.

Climatic data — predictors of the potential distribution of the European beewolf

We obtained 19 bioclimatic variables (Table 1) (2.5′ resolution) (Booth 2018)⁠ from the WorldClim 1.4 database (www.worldclim.org); (Hijmans et al. 2005). These data on annual, monthly, and/or seasonal temperature and rainfall indices are commonly used in modeling potential geographical ranges (Booth et al. 2014; Booth 2018) ⁠and are the factors considered to restrict the geographic distribution of species (Pearson and Dawson 2003; O’Donnell and Ignizio 2012; Sullivan 2014). To limit the number of variables and to avoid collinearity, we assessed the correlation matrix of climatic data for European beewolf occurrences. Then, from each pair of variables containing similar information (e.g., bio5 and bio10, or bio6 and bio11), we removed the one with |r|> 0.7 (Puchałka et al. 2021). Taking into account information about the studied biology of the species, we can justify the choice of climate variables as follows. Bio2, 3, 7, and 10 are variables that preface the amplitude of daily, seasonal, and annual temperature ranges, which may be important to the survival of populations of this Mediterranean species, which may be affected by temperature decreases (Braestrup and Nielsen 1941; Leclercq 1944). Bio17 may determine population survival because snow cover in colder regions can protect nests from low temperatures. Meanwhile, bio18 and 19 inform on the amount of precipitation during Europe’s warmest and wettest period of the year, which falls in summer, the period of imago activity (Simonthomas and Poorter 1972). This reduced the dataset to seven variables for modeling the potential distribution of European beewolf (Table 1).

Table 1 Bioclimatic variables used in the potential distribution modelling of Philanthus triangulum. The finally used in the analysis maps were marked with asterisks
Full size table

Potential distribution modeling

We used the MaxEnt model to model the potential distribution of European beewolf. This is the presence-only method, which uses point occurrence data and pseudoabsences (Phillips et al. 2006; Elith et al. 2011). In contrast to parametric models and other classification tools, MaxEnt can work without true absence data in the theoretical assumptions, but instead, it uses background data (so-called pseudoabsences). As MaxEnt analyzes patterns of presences distinct from the background data, the prevalence of background points makes the model more conservative, requiring a stronger signal than would be the case for equal proportions of presences and pseudoabsences (Elith et al. 2011). Prior to analyses, we split the dataset into a training set (80% of observations) used for model development, and an independent validation set (20% of observations). This split allowed us to prevent overfitting when validating the model using the training set. We used default MaxEnt settings with 10,000 randomly selected pseudoabsence points. We assessed model quality using area under receiver operator curve (AUC) as a metric depending on true positive and true negative rates (i.e., rates of positive and negative overlapping the real and predicted occurrences). The output of the MaxEnt model is a measure of climatic suitability for studied species in a particular raster cell. To obtain presence-absence information, we calculated the threshold — the probability value with the highest sum of the sensitivity (true positive rate) and specificity (true negative rate). Such an approach balances false negatives and false positives (Fielding and Bell 1997). MaxEnt also returns variable importance, expressed as percentage contribution to the model. We used dismo package (Hijmans et al. 2020) for MaxEnt model development and raster (Hijmans 2020) and sf (Pebesma 2018) packages for spatial data processing.

We developed a prediction of the species range shifts for two time horizons: 2040–2060 (hereafter 2050) and 2060–2080 (hereafter 2070). Models were created for two scenarios of increasing climate change magnitude, reflected in the 5th Assessment Report of IPCC by the respective representative concentration pathways (RCPs), based on the difference between the pre-industrial level of radiative forcing (in W m−2) and that predicted for the year 2100 (van Vuuren et al. 2011; Harris et al. 2014; IPCC 2014). The more optimistic variant (RCP4.5) predicts an increase in CO2 concentration in the atmosphere to 650 ppm and an average temperature increase of 1.0–2.6 °C by 2100. This is a scenario predicting intensive economic growth, but assuming emission reduction. The pessimistic scenario (RCP8.5) assumes a CO2 concentration of 1350 ppm and an average temperature increase of 2.6–4.8 °C by 2100 — this is a scenario of intensive economic growth with intensive use of fossil fuels (van Vuuren et al. 2011; Harris et al. 2014; IPCC 2014). To account for uncertainty across different global circulation models (GCMs) used in predicting future values of bioclimatic variables (Thuiller et al. 2019; Paź-Dyderska et al. 2021)⁠, we used three GCMs and averaged the output to provide a more robust forecasts of range shifts. The three GCMs were — HadGEM2-ES (Jones et al. 2011), IPSL–CM5A-LR (Dufresne et al. 2013), and MPI-SM-LR (Giorgetta et al. 2013), reflecting, respectively, low, moderate, and high levels of change in climate conditions depending on scenarios of CO2 concentration (Goberville et al. 2015)⁠. The uncertainty of predictions was estimated using standard deviation (SD) among GCMs for RCPs (see Supplementary material).

Results

Our model showed an overall good performance, expressed by a high AUC (0.864) assessed using a validation dataset. The threshold of occurrence probability, assessed as the point with the highest sum of sensitivity and specificity, was at 0.533. The most important bioclimatic factors influencing the occurrence of European beewolf were “annual temperature range” (bio7), “mean temperature of the warmest quarter” (bio10), and “precipitation of the warmest quarter” (bio18) with importance > 5%; all other variables had importance < 5% (Fig. 1). The probability of European beewolf occurrence decreases with increased annual temperature range (bio7) and drops with increase of precipitation of the warmest quarter (bio18). When it comes to mean temperature of warmest quarter (bio10), it seems that below some threshold value, the probability of European beewolf occurrence is negligible. Above this threshold is a narrow margin of optimum conditions above which probability of occurrence again slowly decreases (Fig. 2). All GCMs variants analyzed in the context of each RCP indicate that the European beewolf will gain more climatically optimal areas for colonization than it will lose (Fig. 3), suggesting that it may become a more significant bee predator in the coming decades. Uncertainty of predictions among GCMs, expressed by the SD, was highest in Eastern Europe and ranged from < 0.001 to 0.538, with medians of 0.044, 0.054, 0.063, and 0.089 for 2050 RCP 4.5, 2050 RCP 8.5, 2070 RCP 4.5, and 2070 RCP 8.5 (Supplementary material, Fig. 2).

Fig. 1
figure 1

Variable importance, expressed as its contribution to the MaxEnt model. Explanations: bio2, mean monthly temperature range; bio8, mean temperature of wettest quarter; bio10, mean temperature of warmest quarter; bio11, mean temperature of coldest quarter; bio15, precipitation seasonality (CV, coefficient of variation); bio16, precipitation of wettest quarter; bio18, precipitation of warmest quarter; bio19, precipitation of coldest quarter

Full size image
Fig. 2
figure 2

Marginal responses of Philanthus triangulum climatic suitability to particular bioclimatic variables (see Table 1 for abbreviations). The red line indicates the predicted response while rugs at the bottom of each panel reflect the distribution of each bioclimatic variable

Full size image
Fig. 3
figure 3

Percentage distribution range shift predicted for six climate change scenarios for three global circulation models and two time-horizons: 2050 (a and b) and 2070 (c and d), related to the current real distribution (number of currently occupied grid cells; a and c), and related to the current potential (predicted) distribution (b and d)

Full size image

The climatically suitable areas cover almost all of Europe, except for high mountain ranges, reach the southern and coastal borders of Finland, Sweden, and Norway, southern parts of the British Isles, westwards to the Coastal regions of Portugal, and in the south to the Greek islands. As predicted by the model, this latter area will become increasingly unsuitable (Fig. 4). Regardless of the accepted scenario, the areas potentially suitable for occupation by European beewolf in the north-eastern part of the Iberian Peninsula will be significantly reduced. According to the fitted model, the species will also lose climatically suitable areas from the central part of the peninsula where it already occurs in suboptimal environments. A similar situation will apply to the southern part of France and to northern and central Italy. However, the greatest loss of climatically suitable habitats for European beewolf will be in south-eastern Europe, from the Adriatic Sea coast to the Black Sea coast.

Fig. 4
figure 4

Current and averaged projected ranges of Philanthus triangulum for two climate change scenarios: optimistic (RCP4.5) and pessimistic (RCP8.5) for 2050 and 2070, respectively. Green area: current and future projected ranges; blue: potential expansion range; orange: potential range contraction; gray: climatically unsuitable in all climatic scenarios. Spatial resolution: 2.50 degrees

Full size image

Averaged forecasts for the pessimistic scenario (RCP8.5) suggest that by 2070, following the loss of climatically suitable habitats (range contraction) in the southern part of Europe below 46°N, the European beewolf will remain only in areas relatively close to the shores of large water bodies. In the north, we predicted range expansion, especially in the British Isles, Scandinavian Peninsula, and from the Kola Peninsula as far east as Moscow and Novaya Zemlya. However, the scale of the predicted loss of climatically suitable habitats from southern Europe is small compared to the expected expansion towards the north. Even in the optimistic variant to 2050, the climatically suitable areas for European beewolf will extend beyond the Arctic Circle as far as the northern part of the Kola Peninsula. The habitat conditions for European beewolf will quickly become more climatically suitable in the south of Finland and the central part of Sweden. These areas will be within the optimum zone of European beewolf habitat requirements by 2050, regardless of the accepted scenario. The situation will be slightly different in Norway, where optimal conditions are forecasted mainly along the coast up to ~ 70°N. In the western part of Europe, changes in habitat conditions will expose Ireland and Scotland, which currently do not have conditions climatically suitable for European beewolf, to rapid colonization by 2050 in each analyzed scenario.

In the forecasts for both 2050 and 2070, all variants assume the expansion of the species range, especially towards the north. In 2050, two GCMs under the pessimistic scenario predict a greater expansion than in the optimistic scenario, while the MPI-SM-LR GCM shows similar values. Regardless of the baseline point of reference (either the actual or potential current distribution), the projections in both scenarios (pessimistic and optimistic) in 2050 and 2070 are similar (Fig. 3). Estimates based on the graphs indicate that the climatically suitable areas for European beewolf compared to today will increase by c. 50% by 2050 and by c.100% by 2070. It also appears that the distribution of European beewolf will increase substantially compared to the current situation (only the HadGEM2-ES model predicts a slight reduction in the frequency of occurrence in all variants).

Discussion

Here, we present the first study to assess the impact of global warming on a specialized predator of the honeybee on a European scale. Our projections predicted by various climatic scenarios suggest that the range of the European beewolf in the 2050 and 2070 scenarios will shift to the north-east. A similar pattern has been found in studies of other plant and animal species occurring in Europe (Dyderski et al. 2018; Iannella et al. 2020; Puchałka et al. 2021; Brygadyrenko et al. 2021).

Our future climatic predictions suggest that European beewolf will increase its distribution range to the north with simultaneous population reduction from the Mediterranean region, where the climatic conditions will be unsuitable for many insect species (e.g., Menéndez 2007; Vanhanen et al. 2007; Sánchez-Guillén et al. 2013). Based on previous empirical observations, European beewolf has increased its range since 1930, which might be a result of climatic changes (Leclercq 1960, 1973). The expansion of the European beewolf was recorded in the Scandinavia in 1942–1962 (Erlandsson 1962), and in North-West Germany in 1940–1977 (Haesler 1977)⁠. In the UK, the European beewolf was extremely rare. However, since 1980, the population there has expanded dramatically, occurring now in southern and mid-northern Britain (Bantock 2010). Moreover, in Sweden and Denmark, the European beewolf was more abundant in 1940 than in previous years, possibly due to the relatively high summer temperatures of 1937–1939 (Braestrup and Nielsen 1941)⁠⁠. In the Netherlands, Koster (1985)⁠ reported that the European beewolf was more common in 1979–1984 than in 1971. European beewolf is a rather Mediterranean species (Leclercq 1944), and it seems that the milder winters, which are a result of global warming, favor the northward spread of this species. A substantial part of the predicted changes in the European distribution of the European beewolf will be its expansion, mainly in a north-easterly direction. In Western Europe, the expansion of suitable area will cover nearly the entire British Isles (Fig. 4). A slightly different situation will occur in the south and in the center of Europe, where certain areas will be outside of the optimum habitat requirements of the species. This mainly concerns the central part of the Iberian Peninsula and the mountain ranges of central Europe.

Model predictions suggest that the most important bioclimatic variable determining the suitability of an area for the European beewolf was seasonal temperature range (Figs. 1 and 2). Temperature plays an essential role in the life of the European beewolf because it correlates with the availability of honeybees and hence the number of prey available to provision their nest. It can also prevent microbial spoilage of their food supplies in the ground (Strohm and Linsenmair 1997; Engl et al. 2016)⁠. Climate warming may have a substantial positive impact on the European beewolf potential niche availability and thus directly affect its abundance which, in turn, will contribute to the spread to the north and northeast of Europe. The future global warming scenarios project a rapid increase in climatic suitability for insect pests, especially in the Northern Hemisphere (Barbet-Massin et al. 2013; Hill et al. 2017; Qin et al. 2019). Thus, thermal climatic conditions play an important role in range shifts and in physiological adaptation for development (Vannote and Sweeney 1980) and survival (Kaiser et al. 2016; Maebe et al. 2021). Consequently, it has an impact on abundance and population size of insect pest species undergoing poleward range expansion due to warming (Økland et al. 2019; Qin et al. 2019; Cornelissen et al. 2019).

Currently, Philanthus spp. are of marginal concern to the beekeeping industry. However, empirical evidence suggests that aggregations of Philanthus species can capture from a hundred up to a thousand individuals of honeybee or other bee species per day (Tinbergen 1958; Simonthomas and Simonthomas 1980; Dukas 2004). Previous evidence suggests that novel predation by invasive wasps affects the mortality of native bee and domestic honeybee (Monceau et al. 2013; Hanna and Eason 2013). Although the European beewolf is currently a minor predator of bees, Cresswell (2017) estimated that rare predators with low prey-capture rates can significantly affect pollinator populations. He projected that 47% of bumblebee workers will be killed by spiders at a 1% chance of being caught. If the European beewolf increases its distribution range, it could exert strong top-down control on bee populations in the future. Until now, high densities of the European beewolf (Simonthomas and Simonthomas 1980; Else 1995a, b) were considered to be rather a local phenomenon. However, attention should be paid to the plasticity of nesting preferences of the European beewolf. Colonies of this species have been documented on moors (Falk 2010)⁠, dunes (Saure 2020), in coal ash and dust, in dumps from coal mines (Smit and de Boer 2008), and even in post-industrial sodium carbonate landfills (Twerd et al. 2017). The species has also become more frequently found in city centers (Burger 2007). Our study, predicting range expansion of European beewolf, is in line with studies predicting range expansion of other honeybee pests (Cornelissen et al. 2019; Giliba et al. 2020). Together with models predicting the decline of suitable niche for bee species (Potts et al. 2010; Goulson et al. 2015; Imbah et al. 2017), climate change can therefore adversely affect both future global bee diversity and the beekeeping industry.

The factors determining the the threat posed by the presence of the European beewolf may be the local abundance of its population in a given area (Petrov 1996) and possible disturbances (e.g., increasing pressure on A. mellifera) in the relative proportions of predator and prey populations. Our results showed that climate change might lead to expansion of the European beewolf in Europe. Therefore, further studies on the phenology of the European beewolf (e.g., the number of generations per season) can better help understand the potential impact of this predator on honeybee survival and the beekeeping industry.

Conclusions

In summary, our findings highlight that a honeybee predator, the European beewolf, may be affected by climate warming to expand its range to a potential new niche in the north and north-east of Europe. Therefore, there is a risk that the species will become widespread there and may put an additional burden on honeybee colonies and endanger the productivity of apiaries. By causing shifts in species niches and their phenology, climate change could lead to significant and, as yet, underestimated changes in community structure and interspecies interactions. Hence, it cannot be ruled out that the European beewolf, whose substantial negative effects on honeybees to date have only been described as local, may become a more significant pest of apiaries in the future. Therefore, we believe that further research is needed to better understand the ecology and population dynamics of this species, and to estimate whether a monitoring of its occurrence and abundance is needed.