Abstract
The decline of wild bee populations causes the decline of bee-pollinated plant populations through the deterioration of pollination services. Since high bee species richness generally involves high functional group diversity, protecting areas of high bee species richness will help to maintain pollination services for plants. However, those areas do not always include the habitats of bee species with specialized functions that expand the range of plants being pollinated. To map important areas for protecting native bee species and their functions, we estimated the distributions and functional range of 13 bumble bee species and 1 honey bee species in Japan. The distributions were estimated from an ensemble of six species distribution models using bee occurrence data and environmental data. The functional range of bee species was estimated by combining the estimated distributions and proboscis length, which frequently corresponds to the floral shape of the plant species they pollinate. The estimated species richness was high in western Hokkaido and the estimated functional range was wide in central Honshu. Our method is useful to see whether areas important for high species richness of pollinators differ from those for rare species or their functions.
Similar content being viewed by others
Introduction
Identifying important areas for biodiversity conservation is one of the key issues for humans as we face ongoing species extinction around the world. There are three major methods to identify important conservation areas: identify areas inhabited by many species, identify areas inhabited by rare species, and identify areas inhabited by species with unique functional traits. Identifying areas based on the number of species (i.e., species richness) is effective if species richness can predict future contributions to biodiversity conservation and ecological function across different sites1. Areas with rare species help to maintain global biodiversity, and areas with species with unique functional traits can help to maintain biodiversity in multiple taxa if their functions affect species across a wide range of taxa.
Bee species are both conservation targets and pollinators whose functions affect many plant species. A worldwide decline in wild bee populations has been repeatedly reported2,3,4,5,6, raising concerns about the deterioration of pollination services. The deterioration of pollination services causes a consequent decline in bee-pollinated wild plant populations, and the decline in wild plant populations in turn causes a shortage of floral resources and the further decline of wild bee populations. To protect bee species and their pollination services, it is necessary to identify areas with high species richness, rare species, and unique functions for pollination services. Pollinator species richness is often correlated with the efficiency of pollination7,8,9,10,11,12. This relationship is considered as the positive effect of functional group diversity on pollination services9,10,12. The increase in pollinator species richness tends to increase the functional group diversity of pollinators, which enhances the efficiency of pollination. However, areas with high bee species richness may not always match areas with rare species or species with unique functional traits specialized to unique plant species.
Functional trait studies suggest an effective approach to estimating the diversity and uniqueness of pollinator functions. Typical functional traits of bees are morphological characteristics such as proboscis length9,13, body length and hairiness12, life histories such as solitariness and sociality14, time and height of visiting flowers10, and activities dependent on climate conditions15. Among these, we focused on morphological characteristics, especially proboscis length. Floral shapes frequently correspond to the proboscis lengths of their pollinators13,16,17,18. Even when the flowers of a plant species are visited by multiple pollinator species, their floral shape corresponds to the proboscis length of the primary pollinator19. Therefore, the diversity of pollinator functions will be represented by the range of proboscis length. On the other hand, the uniqueness of pollinator functions will be represented by the maximum proboscis length. Plant species with long floral tubes depend solely on bee species with a long proboscis for pollination because bee species with a short proboscis do not visit such flowers or rob their nectar without pollination. The range and maximum value of bee proboscis length will be useful for estimating important areas of pollination services for maintaining the diversity of bee-pollinated plant species.
To map important areas for protecting native bee species and their functions, we estimated the distributions and functional range of 13 bumble bee species and 1 honey bee species in the Japanese archipelago. Bumble bees and honey bees are major social bees in Japan and target species for research on the relationship between their proboscis length and floral tube length. Using their occurrence data20,21 and environmental data, we estimated bee distributions from an ensemble of six species distribution models (SDMs). From the estimated distributions, we mapped the areas of high species richness, those of rare species, and those of high uniqueness (as the sum of the inverse of each species’ habitat area)22. We also estimated the minimum, maximum, and range of proboscis lengths from the estimated distributions and the minimum and maximum proboscis lengths extracted from previous studies23,24,25,26 and Yokoyama’s unpublished data. Based on these results, we discuss important conservation areas for bee species and their functions.
Materials and methods
Bees
Sixteen species of bumble bees and two species of honey bees, including one exotic species of each, inhabit the Japanese archipelago except for the Kuril Islands. Among them, we selected 13 native bumble bee species and 1 native honey bee species as study species: Bombus diversus Smith, B. ardens Smith, B. hypocrita Perez, B. ignitus Smith, B. honshuensis Tkalcu, B. beaticola Tkalcu, B. consobrinus Dahlbom, B. deuteronymus Schulz, B. ussurensis Radoszkowski, B. hypnorum Linnaeus, B. pseudobaicalensis Vogt, B. yezoensis Matsumura, B. schrencki Morawitz, and Apis cerana Fabricius. Two native bumble bee species (B. cryptarum Fabricius and B. norvegicus Sparre-Schneider) were not selected because there was not enough occurrence data.
Bee occurrence data
Bee occurrence data were obtained from the “Hanamaru-Maruhana national census” (Bumble bee national census in English)20. This census is a community science program using photographs taken by volunteers. Most of the latitude and longitude were extracted from GPS data in the Exif information of digital photographs. We used 3863 records from 2006 to 2017 (Table 1). The error range of latitude–longitude was within 500 m. The number and geographic range of bumble bee occurrence data collected in the census were greater than in the Global Biodiversity Information Facility (GBIF) as of April 201620. The data are available in GBIF and the Japan Long-Term Ecological Research Network (JaLTER) databases21, although the accuracy of the latitude and longitude was reduced before publishing to protect rare species and the privacy of volunteers in the community science program.
Bombus deuteronymus and B. pseudobaicalensis were indistinguishable in photographs (their difference is only a tuft of black hair on the second abdominal segment); the locations of these photographs were published as the occurrence data of both species and were used for estimating the distribution of both species. Previous studies reported that B. deuteronymus deuteronymus inhabit Hokkaido (the northern island of the Japanese archipelago), B. deuteronymus maruhanabachi inhabit the central region of Honshu (the main island), and B. pseudobaicalensis inhabit both Hokkaido and the northern region of Honshu27. Bombus deuteronymus and B. pseudobaicalensis belong to the same subgenus, and B. deuteronymus deuteronymus and B. pseudobaicalensis have common features such as the altitude range of their habitats and the period of their activity.
Bombus yezoensis is sometimes treated as a synonym for B. consobrinus 28. Since there was little occurrence data for B. yezoensis, we combined the data with those for B. consobrinus to estimate the distribution of both species. Previous studies have reported that B. yezoensis inhabit Hokkaido, and B. consobrinus inhabit central Honshu27.
In addition to the occurrence data of the 14 study species, we also used those of B. terrestris and Xylocopa appendiculata collected in the census to construct a joint SDM (JSDM) and to make the background and the pseudo-absence data (see “Six species distribution models” and “Estimation of species distributions and its accuracy”).
Environmental data
Daily total precipitation, mean temperature, and mean solar radiation at 1-km resolution from 1978 to 2013 were obtained from the National Institute for Agro-Environmental Sciences (NIAES)29. The means of each variable for 10 years (2004–2013) were calculated to minimize the effect of annual fluctuations (Table 2).
Elevation data at 1-km resolution for 2009 and land cover data at 1-km resolution for 2014 were obtained from the National Land Numerical Information (NLNI) (https://nlftp.mlit.go.jp/ksj/index.html). The NLNI provides elevation data for 2009 based on topographic maps at a 1:25,000 scale from the Geospatial Information Authority of Japan (GSI). The elevation category used for the estimates was the mean. The NLNI provides land cover data for 2014 based on digital maps of the GSI and satellite images from SPOT and RapidEye. The land cover categories used for the estimates were areas of paddy fields, other agricultural land, forests, golf courses, wasteland, other land, buildings, lakes and rivers, beaches, and sea (Table 2). The “other land” is an artificial place without a large forest but is not a residential area such as a ski resort.
To avoid multi-collinearity, we tested correlation coefficients and variance inflation factors (VIFs) among the environmental variables. All correlation coefficients were less than 0.8. VIF for forest area was relatively high, 10.1, slightly above the usual cutoff to remove a variable (VIF = 10;30), but because forest area is an important habitat for some bumblebee species, we kept this environmental variable in the analyses.
Proboscis length data of bumble bees and honey bees
The lengths of the prementum and glossa were combined as the proboscis length. We used the proboscis lengths of 454 bumble bee workers from Inoue and Yokoyama24, 770 workers from Inoue et al.25, and 209 workers from Yokoyama’s unpublished data (Table A1 in Appendix A in Supplementary Information (SI)). We also used the proboscis lengths of honey bee workers from Okada et al.23 and Fujiwara et al.26 (Appendix A in SI).
Six species distribution models
We used an ensemble of six SDMs: Generalized linear model (GLM), Generalized additive model (GAM), Random Forest (RF), Generalized boosted model (GBM), Maximum entropy model (MaxEnt), and Hierarchical Multivariate Probit Regression model (HMPRM)31. To improve the robustness of ensemble model estimation, it is desirable to include modeling algorithms of different classes32. GLM and GAM are conventional regression models with a firm statistical base. RF, GBM, and MaxEnt are machine learning algorithms. HMPRM is a JSDM and estimates the distribution of multiple species considering the correlations between species that are not explained by fixed effects included in the model31. In addition to the occurrence data of the 14 study species, we used the occurrence data of B. terrestris and X. appendiculata to construct a JSDM, because these species might affect the distribution of the study species through species interactions. The estimated distributions of B. terrestris and X. appendiculata were not used to evaluate species richness and functional range.
All models were calculated in R v. 4.1.1 software33. We assumed that the response variable follows a Bernoulli distribution and used the logistic link function for GLM and GAM. We performed model selection for GLM and reduced explanatory variables using the stepAIC function of the ‘MASS’ package based on Akaike information criterion (AIC). We included the interaction terms for all the possible combinations of the explanatory variables in the GLM. To fit GAM, we selected smoothing parameters by the ‘REML’ method in the ‘mgcv’ package. For RF, we used the tuneRF function of the ‘randomForest’ package to tune the number of variables. We used the ‘brt’ package to estimate GBM and tuned three parameters, namely, the number of iterations, the complexity of the tree, and the learning rate by tenfold cross-validations. To estimate MaxEnt model, we used MaxEnt v. 3.3.3 k software with linear and quadratic features and the default value of the regularization multiplier according to Suzuki-Ohno et al.20,34. To estimate the JSDM model, we used the R scripts of Pollock et al.31. Each distribution estimated by each model is shown in Appendix B in SI.
Species distribution estimation and its accuracy
We used presence-only data in this study. Such data are prone to survey bias, and correction of bias significantly affects the appropriateness of model estimation. To correct survey bias, we used target-group background35,36, which uses the set of occurrences of all the bee species in our survey as background in the MaxEnt model. In the other models, we used pseudo-absences, which use the set of occurrences of all the bee species in our survey other than the focal species to construct a model as absences. In addition to the occurrence data of the 14 study species, we included the occurrence data of B. terrestris and X. appendiculata in the background and the pseudo-absence data because it is appropriate to use all observed occurrences in a survey to correctly represent the distribution of survey effort.
We combined the occurrence data of B. consobrinus wittenburgi and B. yezoensis as those of both species. The estimated combined distribution was divided into B. consobrinus wittenburgi in Honshu and B. yezoensis in Hokkaido. We also combined the occurrence data of B. deuteronymus and B. pseudobaicalensis as those of both species because they were not discriminated in the occurrence data obtained from the census. The estimated combined distribution was divided into B. deuteronymus maruhanabachi in central Honshu and B. deuteronymus deuteronymus and B. pseudobaicalensis in the other regions, because the former one inhabits only a limited area far from those of B. deuteronymus deuteronymus and B. pseudobaicalensis.
To evaluate the prediction accuracy of the five single-species distribution models (GLM, GAM, RF, GBM, and MaxEnt), we performed tenfold cross-validation and calculated test-AUC (area under the curve) for each. Because that of JSDM requires intensive calculation, we calculated training-AUC for JSDM.
Species richness
The distributions of species richness were calculated from the distributions estimated by the six SDMs. The probability of distributions estimated by SDMs was converted into presence/absence (0 or 1) data by using thresholds, and the presence data of all species were summed as species richness. The threshold value above which we converted the probability to presence was determined by maximizing sensitivity plus specificity method37,38. The presence data of B. ignitus and A. cerana in Hokkaido were not included in this calculation because those species do not inhabit Hokkaido. After the conversion, we averaged the presence/absence (0 or 1) among the six models for each species to obtain ensemble estimates for each species.
Rare species and uniqueness
In the IUCN Red List-like categories set by the Ministry of the Environment in Japan (https://www.env.go.jp/content/900515314.pdf: almost the same categories but applied only within the Japanese territory), B. ignitus and B. cryptarum are classified as Near Threatened (NT), and B. consobrinus wittenburgi and B. ussurensis as Data Deficient (DD) in Japan. Although B. ignitus originally inhabited widely, B. ignitus is classified as NT because declines in B. ignitus have been reported in some local areas in Japan. Bombus deuteronymus is not classified as these groups; however, B. deuteronymus maruhanabachi is a symbolic subspecies inhabiting a limited area in Japan. Therefore, B. ignitus, B. consobrinus wittenburgi, B. ussurensis, and B. deuteronymus maruhanabachi were defined as rare species/subspecies here. Their distributions were calculated from the distributions estimated by the six SDMs.
We estimated the distributions of uniqueness, giving higher weight to the presence of species that have smaller habitat areas even if they are not considered rare. To calculate uniqueness, we used weights proportional to the inverse of the total distribution area of each species. We scaled the weights so that the sum of the weights of all species was 1. Thus, the value of uniqueness has the same scale as the species richness (i.e., when all species are present at a site, uniqueness takes the same value as richness).
The minimum, maximum, and range of bee function
The minimum, maximum, and range of bee function were estimated from the estimated distributions and proboscis length data. We used the minimum and maximum of proboscis length for each species obtained from Okada et al.23, Inoue and Yokoyama24, Inoue et al.25, Fujiwara et al.26, and Yokoyama (unpublished data) (Table A1 in Appendix A in SI). We distinguished between the proboscis length of B. consobrinus wittenburgi and that of B. yezoensis, or between those of subspecies (Table A1). When estimating bee function in an area, the proboscis length of subspecies inhabiting the area was selected.
We calculated the functional range in each 1 km cell from the estimated presence/absence. We compared the minimum proboscis length of species estimated to be present in the cell and used the smallest value as the minimum proboscis length in the cell. Similarly, we defined the maximum proboscis length in a cell as the largest of the maximum values of each species present in the cell. The range of proboscis lengths in a cell is the difference between the minimum and maximum values. To check whether there is a range of non-existing proboscis lengths in the range, we also calculated the difference between the minimum and maximum proboscis lengths minus the range of non-existing proboscis lengths.
Results
Accuracies of estimated distributions
The average AUCs of the estimations of all species were ≥ 0.7 (Table 3). The average AUC of B. beaticola was the highest among these bee species (Table 3). Since B. beaticola prefers cool climates in high-altitude regions, their distribution could be estimated with high accuracy regardless of model. The average AUC of B. diversus was the lowest, even though this species had the most occurrence data (Tables 1, 3). This result was mainly attributed to the ability of B. diversus to inhabit a wide range of environments ranging from natural to urban areas.
AUCs depended mainly on the combination of models and species. The machine learning models (RF, GBM, and MaxEnt) had high AUCs of the estimation for species with smaller distribution areas such as B. deuteronymus, B. hypnorum, B. schrencki, and B. ussurensis. The classical regression models (GLM and GAM) had high AUCs of the estimation for species with wider distribution areas such as B. ardens, B. hypocrita, B. ignitus, and A. cerana. The ensemble estimations showed stable estimation accuracy, because the characteristics of each model complemented each other. Estimated distributions were discussed in Appendix B in SI.
Species richness
Estimated species richness was high in western Hokkaido (Fig. 1). Estimated species in Hokkaido mainly consists of B. ardens, B. beaticola, B. yezoensis, B. deuteronymus deuteronymus or B. pseudobaicalensis, B. diversus, B. honshuensis, B. hypnorum, B. hypocrita, and B. schrencki.
Estimated species richness of bee species. Blue and red represent low and high values, respectively. This map was created by R v. 4.1.1 software33.
Estimated species richness was also high in high-altitude areas in central Honshu, but it was slightly lower than that in Hokkaido. Estimated species in the areas mainly consist of B. ardens, B. beaticola, B. consobrinus wittenburgi, B. deuteronymus maruhanabachi, B. diversus, B. honshuensis, B. hypocrita, and B. ussurensis.
Rare species and uniqueness
The occurrence data of the rare species B. ignitus were found over a wide area (black points in Fig. 2a), and their distribution was estimated in low-altitude areas such as near the coast (Fig. 2a). On the other hand, those of the rare species B. consobrinus wittenburgi, B. deuteronymus maruhanabachi, and B. ussurensis were estimated in higher-altitude areas in central Honshu (Fig. 2b).
Estimated distributions of rare bee species. In the IUCN Red List-like categories set by the Ministry of the Environment in Japan, B. ignitus is Near Threatened (NT), and B. consobrinus and B. ussurensis are Data Deficient (DD). Bombus deuteronymus maruhanabachi is a symbolic subspecies in Japan. (a) Mean of probabilities of B. ignitus distribution estimated by six models. Black points represent the locations of occurrence data. (b) Estimated species richness of B. consobrinus, B. ussurensis, and B. deuteronymus maruhanabachi. Blue and red represent low and high values, respectively. This map was created by R v. 4.1.1 software 33.
Estimated uniqueness was high in central Honshu and Hokkaido (Fig. 3). High uniqueness in central Honshu was attributed to the estimated distributions of the rare species/subspecies B. consobrinus wittenburgi, B. deuteronymus maruhanabachi, and B. ussurensis (Fig. 2b). Uniqueness was also high in Hokkaido because of high species richness (Fig. 1).
Estimated uniqueness. Blue and red represent low and high values, respectively. This map was created by R v. 4.1.1 software33.
The minimum, maximum, and range of the function
The minimum proboscis length was lower in low-altitude areas and the maximum proboscis length was higher in high-altitude areas (Fig. 4a and b). Estimated functional range between the minimum and maximum was wide in central Honshu (Fig. 4c). Wide functional range in central Honshu was closely related to the maximum value of proboscis length (Fig. 4b). The rare species B. consobrinus wittenburgi has the longest proboscis (Table A1), and it inhabits only central Honshu. Estimated functional range minus the non-existing range was also high in central Honshu (Fig. 4d). Although the values for eastern Hokkaido are lower in Fig. 4d compared to Fig. 4c, they showed a similar pattern except for it.
Estimated functional range. (a) the minimum, (b) the maximum, (c) the range between the minimum and maximum, (d) the range between the minimum and maximum minus the non-existing range of proboscis length. Blue and red represent low and high values, respectively. This map was created by R v. 4.1.1 software33.
Discussion
Our method shows whether areas important for high species richness differ from those for rare species or their functional range. Estimated species richness was high in western Hokkaido (Fig. 1). The distribution of the rare species B. ignitus was estimated in low-altitude areas in Honshu (Fig. 2a), and the distributions of the rare species B. consobrinus wittenburgi, B. deuteronymus maruhanabachi, and B. ussurensis were estimated in high-altitude areas in central Honshu (Fig. 2b). Estimated uniqueness in central Honshu was approximately the same as that of Hokkaido (Fig. 3). Estimated functional range was wide in central Honshu (Fig. 4), and the wide functional range was mainly attributed to the longest proboscis length of the rare species B. consobrinus wittenburgi (Fig. 4b and Table A1).
Conservation areas for pollinators can be selected based on the distributions of estimated species richness, rare species, and functional range with models that prioritize conservation areas, such as Marxan39 and Zonation40. Which of species richness, rare species, and functional range should be weighted for conservation areas will change depending on their relationships, the time, and the country. In this study, there was not much difference in estimated uniqueness between Hokkaido and central Honshu. The areas of estimated wide functional range corresponded to the estimated distributions of rare species in central Honshu. Therefore, it is efficient to decide on conservation areas by focusing on species richness and functional range. If you increase the weight for species richness, western Hokkaido will become larger in conservation areas, and if you increase the weight for functional range, central Honshu will become larger. In the future study, a prioritization analysis including such weighting and future conservation plans for bees considering the effects of global warming will be important in identifying specific areas for bee conservation.
The locations of existing protected areas also affect the future conservation plans for bees. The Ministry of the Environment designates national and quasi-national parks to preserve the environment and biodiversity in Japan. There are six national parks in Hokkaido and five in central Honshu (Nagano Prefecture and surrounding areas), covering areas of 24,166 to 226,764 hectares. There are six quasi-national parks in Hokkaido and four in central Honshu (Nagano Prefecture and surrounding areas), covering areas of 9083 to 103,447 hectares. These parks partially overlap with the important areas for bumble bees estimated in this study. However, western Hokkaido, which is estimated to have the highest species richness, has few national and quasi-national parks. Suzuki-Ohno et al.34 estimated that past global warming would have a large impact on the distribution of bumble bees in Hokkaido. We will have to increase conservation areas for bumble bees in western Hokkaido.
Hokkaido had the highest species richness, followed by central Honshu. High species richness in both regions is based on the biotic factors of the bumble bee itself and the biogeographic factors of the Japanese archipelago. In contrast to the other eusocial bee tribes in the family Apidae (e.g. centers of both Apini and Meliponini are in the tropical regions), the majority of species diversity of bumble bees (tribe Bombini) is in the temperate Palaearctic and Oriental regions (https://www.nhm.ac.uk/research-curation/research/projects/bombus/introduction.html) and tend to prefer relatively cool climates. Because of the preferences, the species richness of bumble bees is positively correlated with latitude in the United States of America41. In Europe, the highest predicted species richness of bumble bees appeared between 40° N and 50° N latitude42. Since the latitudinal range of Hokkaido is from ca. 41.4° N and 45.5° N, it is natural that Hokkaido holds the highest species richness. Central Honshu is a special place in Honshu because this place involves many high mountain complexes (highest altitude > 3000 m) and the average elevation is the highest in Japan (1132 m in Nagano Prefecture, the main prefecture of central Honshu). The temperature condition of Hokkaido is almost the same as that at the average altitude of Nagano Prefecture. This explain why central Honshu also has a high species richness.
Most of the bumble bee species in Japan are common in continental Asia27. Divergence times of these species estimated from molecular phylogenies43,44 indicate that they originated after the separation of the Japanese archipelago from Eurasia (ca. 20 Ma45,46). These facts suggest that most of the species diversity of Japanese bumble bees was brought by migration from the continent. There are two main routes of migration to the major Japanese islands: the northern (from Sakhalin or Kamchatka to Hokkaido) and southern (from the Korean Peninsula) routes47,48. Hokkaido has a high species richness of bumble bees because it is close to the migration route from the continent, which has a higher diversity of bumble bee fauna (see Appendix C in SI for the details of the history of migration). Our study suggests that Hokkaido has rich bumble bee fauna while central Honshu has bumble bee fauna with a wide functional range. Central Honshu has richer flora, including many endemic species, than other regions in the Japanese archipelago49,50. Central Honshu is considered to have provided refugia in terms of both long-term in the east Asia and glacial- and inter-glacial-period cycles in the Quaternary51,52. It is possible that the floral richness, the refugial function, and the relationship between bumble bees and flowering plants maintain the diversity of proboscis lengths of bumble bees in central Honshu.
When conservation targets are plants that need bee pollinators to reproduce, it is desirable to estimate the distribution of floral tube lengths and the distribution of proboscis lengths and to identify any mismatches between them. If a mismatch is estimated, the actual floral tube lengths and proboscis lengths in the area can be investigated to validate the mismatch. Conservation plans for the plants and their effective pollinators can be considered from such results of field surveys. Our method will be also useful for comprehensive conservation planning for plants and pollinators.
When we estimate species distributions and their functional range, we should always consider the risks of overestimation/underestimation. Overestimation is costly for conservation plans, whereas underestimation may exclude important conservation areas. Our method has risks of overestimating or underestimating (1) species richness, (2) the distribution area of rare species, and (3) functional range. Risks (1) and (2) could be caused by limited or spatially biased occurrence data. For example, the number of occurrence data in Hokkaido was smaller than that of Honshu, and the estimated distributions by GBM (Figs. B2, B4, B6, B8, B10, B12, B14, B16, B18, B20, B22, and B24 in Appendix B in SI) tended to be smaller than those previously reported in field research. In contrast, distributions estimated by MaxEnt and JSDM tended to be larger than those previously reported in field research. The risk of overestimation/underestimation could be mitigated by averaging the estimations of the six SDMs. Risk (3) could be caused by limited data on intraspecific differences in function by local adaptation. We could incorporate intraspecific differences in proboscis length for B. beaticola, B. consobrinus, and B. diversus (Table A1). In Hokkaido, the wide functional range depended on the proboscis length of B. yezoensis, which is sometimes treated as a synonym of B. consobrinus. In central Honshu, it depended on the proboscis length of B. consobrinus wittenburgi, inhabiting only central Honshu. Therefore, in this study, the risk of overestimating/underestimating the functional range was reduced by discriminating the proboscis length of B. yezoensis from that of B. consobrinus. We should check whether the estimated functional range in a certain region depends on the functional range of a specific species and whether the specific species has intraspecific variation in the functional range. The risk of overestimating/underestimating the functional range should be reduced by using regional functional trait data. Using regional functional trait data is also important to develop conservation plans for the plants that they pollinate.
Data availability
The bee occurrence data at coarser coordinate resolution are available on two databases, JaLTER (https://db.cger.nies.go.jp/JaLTER/metacat/metacat/ERDP-2021-04.1/jalter-en) and GBIF (https://www.gbif.org/ja/dataset/5f7aeace-46b9-4516-8b1e-09650aee9024)21. The original bee occurrence data are available from the corresponding authors upon reasonable request.
References
Fleishman, E., Noss, R. F. & Noon, B. R. Utility and limitation of species richness metrics for conservation planning. Ecol. Ind. 6, 543–553 (2006).
Biesmeijer, J. C. et al. Parallel declines in pollinators and insect-pollinated plants in Britain and the Netherlands. Science 313, 351–354 (2006).
Winfree, R., Aguilar, R., Vázquez, D. P., LeBuhn, G. & Aizen, M. A. A meta-analysis of bee’s responses to anthropogenic disturbance. Ecology 90, 2068–2076 (2009).
Potts, S. G. et al. Global pollinator declines: Trends, impacts and drivers. Trends Ecol. Evol. 25, 345–353 (2010).
Ollerton, J., Erenler, H., Edwards, M. & Crockett, R. Extinctions of aculeate pollinators in Britain and the large-scale agricultural changes. Science 346, 1360–1362 (2014).
Ollerton, J. Pollinator diversity: Distribution, ecological function, and conservation. Ann. Rev. Ecol. Evol. Syst. 48, 353–376 (2017).
Klein, A.-M., Steffan-Dewenter, I. & Tscharntke, T. Fruit set of highland coffee increases with the diversity of pollinating bees. Proc. R. Soc. London Ser. B Biol. Sci. 270, 955–961 (2003).
Kremen, C. Managing for ecosystem services: What do we need to know about their ecology?. Ecol. Lett. 8, 468–479 (2005).
Fontaine, C., Dajoz, I., Meriguet, J. & Loreau, M. Functional diversity of plant-pollinator interaction webs enhances the persistence of plant communities. PloS Biol. 4, e1 (2005).
Hoehn, P., Tscharntke, T., Tylianakis, J. M. & Steffan-Dewenter, I. Functional group diversity of bee pollinators increases crop yield. Proc. R. Soc. London Ser. B Biol. Sci. 275, 2283–2291 (2008).
Garibaldi, L. A. et al. Wild pollinators enhance fruit set of crops regardless of honey bee abundance. Science 339, 1608–1611 (2013).
Woodcock, B. A. et al. Meta-analysis reveals that pollinator functional diversity and abundance enhance crop pollination and yield. Nat. Commun. 10, 1481 (2019).
Stang, M., Klinkhamer, P. G. L. & van der Meijden, E. Asymmetric specialization and extinction risk in plant-flower visitor webs: A matter of morphology or abundance?. Oecologia 151, 442–453 (2007).
Albrecht, M., Schmid, B., Hautier, Y. & Müller, C. B. Diverse pollinator communities enhance plant reproductive success. Proc. R. Soc. B Biol. Sci. 279(1748), 4845–4852 (2012).
Fründ, J., Dormann, C. F., Holzschuh, A. & Tscharntke, T. Bee diversity effects on pollination depend on functional complementarity and niche shifts. Ecology 94, 2042–2054 (2013).
Inouye, D. W. The effect of proboscis and corolla tube lengths on patterns and rates of flower visitation by bumblebees. Oecologia 45, 197–201 (1980).
Suzuki, K. Bumblebee pollinators and pollination ecotypes of Isodon umbrosus and I. shikokianus (Lamiaceae). Plant Spec. Biol. 7, 37–47 (1992).
Kobayashi, S., Inoue, K. & Kato, M. Evidence of pollen transfer efficiency as the natural selection factor favoring a large corolla of Campanula punctata pollinated by Bombus diversus. Oecologia 111, 535–542 (1997).
Shimizu, A. et al. Fine-tuned bee-flower coevolutionary state hidden within multiple pollination interactions. Sci. Rep. 4, 3988 (2014).
Suzuki-Ohno, Y., Yokoyama, J., Nakashizuka, T. & Kawata, M. Utilization of photographs taken by citizens for estimating bumblebee distributions. Sci. Rep. 7, 11215. https://doi.org/10.1038/s41598-017-10581-x (2017).
Suzuki-Ohno, Y., Yokoyama, J., Nakashizuka, T. & Kawata, M. Bee occurrence data collected in citizen science program “Hanamaru-Maruhana national census” in Japan. Ecol. Res. https://doi.org/10.1111/1440-1703.12261 (2021).
Hoffmann, S., Beierkuhnlein, C., Field, R., Provenzale, A. & Chiarucci, A. Uniqueness of protected areas for conservation strategies in the European Union. Sci. Rep. 8, 6445 (2018).
Okada, I., Sakai, T. & Hasegawa, M. Notes on some morphological characters of Japanese honeybee. Kontyu 24, 145–153 (1956) (in Japanese).
Inoue, M. N. & Yokoyama, J. Morphological variation in relation to flower use in bumblebees. Entomol. Sci. 9, 147–159 (2006).
Inoue, M. N., Yokoyama, J. & Washitani, I. Displacement of Japanese native bumblebees by the recently introduced Bombus terrestris (L.) (Hymenoptera: Apidae). J. Insect Conserv. 12, 135–146 (2008).
Fujiwara, A., Wada, S. & Washitani, I. Conservation-ecological characterization of Apis cerana japonica on Amami Ōshima Island: body size, nest sites, natural enemies, colony activity, and swarming during the breeding season. Jpn. J. Conserv. Ecol. 20, 131–145 (2015) (in Japanese).
Kinota, K., Takamizawa, K. & Ito, M. The Bumblebees of Japan (Hokkaido University Press, 2013) (in Japanese).
Williams, P. H. An annotated checklist of bumble bees with an analysis of patterns of description (Hymenoptera: Apidae, Bombini). Bull. Nat. Hist. Mus. (Entomol.) 67, 79–152 (1998).
Seino, H. An estimation of distribution of meteorological elements using GIS and AMeDAS data. J. Agric. Meteorol. 48, 379–383 (1993) (in Japanese).
Vittinghoff, E., Glidden, D. V., Shiboski, S. C. & McCulloch, C. E. Regression Methods in Biostatistics: Linear, Logistic, Survival, and Repeated Measures Models (Springer Publishing Co., 2005).
Pollock, L. J. et al. Understanding co-occurrence by modelling species simultaneously with a Joint Species Distribution Model (JSDM). Methods Ecol. Evol. 5, 397–406 (2014).
Marmion, M., Parviainen, M., Luoto, M., Heikkinen, R. K. & Thuiller, W. Evaluation of consensus methods in predictive species distribution modelling. Divers. Distrib. 15, 59–69 (2009).
R Core Team R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2021).
Suzuki-Ohno, Y., Yokoyama, J., Nakashizuka, T. & Kawata, M. Estimating possible bumblebee range shifts in response to climate and land cover changes. Sci. Rep. 10, 19622. https://doi.org/10.1038/s41598-020-76164-5 (2020).
Elith, J. & Leathwick, J. Predicting species distributions from museum and herbarium records using multiresponse models fitted with multivariate adaptive regression splines. Divers. Distrib. 13, 265–275 (2007).
Phillips, S. J. & Dudík, M. Modeling of species distributions with Maxent: New extensions and a comprehensive evaluation. Ecography 31, 161–175 (2008).
Liu, C., Berry, P. M., Dawson, T. P. & Pearson, R. G. Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28, 385–393 (2005).
Jiménez-Valverde, A. & Lobo, J. M. Threshold criteria for conversion of probability of species presence to either- or presence-absence. Acta Oecol. 31, 361–369 (2007).
Watts, et al. Marxan with Zones: Software for optimal conservation based land- and sea-use zoning. Environ. Modell. Softw. 24, 1513–1521 (2009).
Moilanen, A. et al. Balancing alternative land uses in conservation prioritization. Ecol. Appl. 21, 1419–1426 (2011).
Strange, J. P. & Tripodi, A. D. Characterizing bumble bee (Bombus) communities in the United States and assessing a conservation monitoring method. Ecol. Evol. 9, 1061–1069 (2019).
Polce, C. et al. Distribution of bumblebees across Europe. One Ecosyst. 3, e28143. https://doi.org/10.3897/oneeco.3.e28143 (2018).
Hines, H. M. Historical biogeography, divergence times, and diversification patterns of bumble bees (Hymenoptera: Apidae: Bombus). Syst. Biol. 57, 58–75 (2008).
Santos Júnior, J. E. et al. Biogeography and diversification of bumblebees (Hymenoptera: Apidae), with emphasis on Neotropical species. Diversity 14, 238. https://doi.org/10.3390/d14040238 (2022).
Otofuji, Y. I., Matsuda, T. & Nohda, S. Opening mode of the Japan Sea inferred from the palaeomagnetism of the Japan Arc. Nature 317, 603–604 (1985).
Isozaki, Y., Aoki, K., Nakama, T. & Yanai, S. New insight into a subduction-related orogen: A reappraisal of the geotectonic framework and evolution of the Japanese Islands. Gondwana Res. 18, 82–105 (2010).
Dobson, M. Patterns of distribution in Japanese land mammals. Mammal Rev. 24, 91–111 (1994).
Millien-Parra, V. & Jaeger, J.-J. Island biogeography of the Japanese terrestrial mammal assemblages: An example of a relict fauna. J. Biogeogr. 26, 959–972 (1999).
Ohsawa, T. & Ide, Y. Phylogeographic patterns of highland and lowland plant species in Japan. Alpine Botany 121, 49–61 (2011).
Kubota, Y., Shiono, T. & Kusumoto, B. Role of climate and geohistorical factors in driving plant richness patterns and endemicity on the east Asian continental islands. Ecography 38, 639–648 (2015).
Tang, C. Q. et al. Identifying long-term stable refugia for relict plant species in East Asia. Nat. Commun. 9, 4488. https://doi.org/10.1038/s41467-018-06837-3 (2018).
Ikeda, H. Decades-long phylogeographic issues: complex historical processes and ecological factors on genetic structure of alpine plants in the Japanese Archipelago. J. Plant Res. 135, 191–201 (2022).
Acknowledgements
We greatly thank the many volunteers participating in the Hanamaru-Maruhana National Census, especially K. Minemura, H. Kawabata, Y. Ishimori, Y. Sakai, H. Morishima, “Myano”, H. Matsuda, K. Tuchita, “Hakiri-ari”, Y. Asou, Y. Horimoto, N. Nakamoto, and D. Yarimizu for their help in taking bee photographs. We thank Y. Hatakeyama for his help in supporting the census; and Y. Ampo (Hokkaido Environment Foundation), H. Abe (Hokkaido Government), W. Ohnishi (Kanagawa Prefectural Museum of Natural History), U. **bo (National Museum of Nature and Science), M. Yamazaki (Sapporo Museum Activities Center), Y. Shirosaka (Bihoro Museum), S. Kariyama (Kurashiki Museum of Natural History), and K. Nakagawa (Minamisoma City Museum) for their help in advertising the census. JY thank Y. Uyama and Y. Igusa for their help in field works. YSO was supported by a Grant-in-Aid for JSPS Fellows (JP16J40194). FI was supported by the Environment Research and Technology Development Fund (JPMEERF23S12115) of the Environmental Restoration and Conservation Agency provided by Ministry of the Environment of Japan. TN was supported by the Environment Research and Technology Development Fund (S-15 Predicting and Assessing Natural Capital and Ecosystem Services) of the Ministry of the Environment, Japan, and a JSPS KAKENHI Grant (17H03835).
Author information
Authors and Affiliations
Contributions
Y.S.O., F.I., J.Y., T.N., and M.K. conceived the ideas and designed the methodology; Y.S.O. and J.Y. arranged the data; F.I. estimated bee distributions and functional range; Y.S.O. led the writing of the manuscript; Y.S.O., F.I., J.Y., M.N.I., T.N., and M.K. contributed critically to the drafts and gave final approval for publication.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Suzuki-Ohno, Y., Ishihama, F., Yokoyama, J. et al. Estimating bee distributions and their functional range to map important areas for protecting bee species and their functions. Sci Rep 14, 12842 (2024). https://doi.org/10.1038/s41598-024-61848-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-024-61848-z
- Springer Nature Limited