Climate change gives bumblebees a tongue twister

Bombus sylvicola

Editor's Introduction

Functional mismatch in a bumble bee pollination mutualism under climate change

Climate change can force change in many living organisms. In this paper, authors investigate the role of climate change on bumble bee tongue length. Over the years, alpine bumble bee tongues have gotten shorter! The authors investigate which factors played a role in the change in tongue length and if this evolutionary change will play a role in future alpine bumble bee survival.

NOTE: This article is part of a Collection of student-annotated papers that are the product of the SitC team’s research into best practices for using primary literature to support STEM education. For this reason, these papers have undergone an alternate review process and may lack educator guides. To learn more, visit the main Collection page: SitC Lab.

Paper Details

Original title
Functional mismatch in a bumble bee pollination mutualism under climate change
Original publication date
Vol. 349, Issue 6255, pp. 1541-1544
Issue name


Ecological partnerships, or mutualisms, are globally widespread, sustaining agriculture and biodiversity. Mutualisms evolve through the matching of functional traits between partners, such as tongue length of pollinators and flower tube depth of plants. Long-tongued pollinators specialize on flowers with deep corolla tubes, whereas shorter-tongued pollinators generalize across tube lengths. Losses of functional guilds because of shifts in global climate may disrupt mutualisms and threaten partner species. We found that in two alpine bumble bee species, decreases in tongue length have evolved over 40 years. Co-occurring flowers have not become shallower, nor are small-flowered plants more prolific. We argue that declining floral resources because of warmer summers have favored generalist foraging, leading to a mismatch between shorter-tongued bees and the longer-tubed plants they once pollinated.


Long-tongued bumble bees have coevolved to pollinate plants that possess elongated corolla tubes in a mutualistic relationship. Recent declines in such long-tongued bee populations suggest that historical selection regimes in these systems are changing (13), yet the mechanisms driving these declines are unclear. Spatial and temporal discrepancies with food plants, habitat destruction, and pressure from invasive competitors have been implicated (36), but the details of these declines and their causes remain unresolved.

Matching of functional morphology between partners increases benefits and reduces costs in mutualisms (78). The match between flower tube depth and pollinator tongue length influences resource acquisition and pollination effectiveness (910). For bees, hummingbirds, bats, moths, and flies, morphological matching increases handling efficiency on flowers (914). Thus, changes that disrupt such matching can alter plant species recruitment and the trajectory of coevolution. Although the climate change impacts on phenological and spatial overlap of mutualists are well known, the role of climate change in generating functional discrepancies between them is less understood. Using historical data, we show that reduced flower abundance in bumble bee host-plants at the landscape scale has accompanied recent warming, leading to evolutionary shifts in foraging traits of two alpine bumble bee species (Bombus balteatus and B. sylvicola). Rapid evolution of shorter tongues in these species may inform our understanding of widespread declines in long-tongued Bombus specialists.

We measured the change in tongue length of B. balteatus and B. sylvicola using specimens collected from 1966–1980 and 2012–2014 in the central Rocky Mountains (15). These two species historically comprised 95 to 99% of bumble bees at our high-altitude field sites (1618). B. balteatus workers were collected from three geographically isolated locations: Mount Evans (39°35.033′N, 105°38.307′W), Niwot Ridge (40°3.567′N, 105°37.000′W), and Pennsylvania Mountain (39°15.803′N, 106°8.564′W).

B. sylvicola workers were collected from Niwot Ridge and Pennsylvania Mountain. Mean tongue length has decreased 0.61% annually and 24.4% cumulatively in these taxa (F1,23 = 17.02, P = 0.0004 and F1,67 = 46.14, P < 0.0001) (Fig. 1 and table S1). Using archived specimens and field surveys of bumble bees and host plants, we explored four potential mechanisms for this change in tongue length: (i) decreasing body size, (ii) coevolution with floral traits, (iii) competition from subalpine invaders, and (iv) diminishing floral resources.

figure 1

Fig. 1. Change in tongue length for B. balteatus and B. sylvicola on Mount Evans, Niwot Ridge, and Pennsylvania Mountain. (AB. balteatus. (BB. sylvicola. Bars represent least squares means >± SE. (15). Asterisks denote significant differences (P < 0.05) between means. Dagger denotes a trend (P < 0.06).

Tongue length

Results show that in all three locations used in the study, bumble bees had longer tongues in the past then they do in the present. For all three locations the differences between tongue length are statistically significant. (p<.05).

Change in tongue length

The change in tongue length of B. sylvicola (a species of bumble bee) in Niwot Ridge and Pennsylvania Mountain. A decrease in average (mean) tongue length was seen in both locations however only the change in Niwot Ridge is significant (P < .05). The change seen in Pennsylvania Mountain denotes a trend meaning that there is likely an underlying cause leading to a decrease in the length of B. sylvicola tongues.

Temporal changes in bumble bee tongue length are not explained by plasticity in body size. When phenotypic variance in tongue length is partitioned among underlying sources, size accounts for less than 20% of variation (table S1) (15). Size has declined in some populations (B. balteatusF2,96= 8.61, P = 0.0004; B. sylvicolaF1,76 = 29.01, P < 0.0001) (fig. S1 and table S1) and is phenotypically correlated with tongue length [correlation coefficient (r) = 0.50 to 0.60, P < 0.005] (fig. S1) but contributes little to its reduction over time. After removing variance explained by body size, analysis of covariance shows significant temporal changes in tongue length (B. balteatusF1,23 = 17.02, P = 0.0004; B. sylvicolaF1,67 = 46.14, P < 0.0001) (Fig. 1 and table S1).

Selection to track the floral traits of host plants should favor short-tongued pollinators when flowers become shallower or deep flowers less common (910). We tested this hypothesis by comparing flower depth of herbarium specimens collected from 1960–1982 and 2012–2013 near Mount Evans and Niwot Ridge (15). In six species that historically provided 88% of floral resources for B. balteatus and B. sylvicola (18), the change in flower depth over time varied among species (F6,13 = 9.42, P = 0.0004). Species that now have shallower flowers received few (<10%) bumble bee visits historically (fig. S2A). On Pennsylvania Mountain and Niwot Ridge, short-tubed flowers show no systematic increase in abundance [coefficient of determination (R2) = 0.227, t1,4 = 1.21, P= 0.294; R2 = 0.0004, t1,9 = –0.62, P = 0.952, respectively) (fig. S2, B and C), suggesting that recent changes in floral trait distributions are insufficient to drive tongue length adaptation in bumble bees.

Alternatively, shifts in bumble bee tongue length may reflect competition from subalpine congeners moving upward with climate change. Comparisons of past (1960s and 1970s) (1617) and present (2008 and 2011–2014) bumble bee communities on Pennsylvania Mountain and in the Front Range [Mount Evans and Niwot Ridge combined (16)] indicate increased species diversity (respectively, χ2 = 293.4, df = 7, and χ2 = 579.4, df = 12, P < 0.001) (Fig. 2, A and B), immigration of short-tongued species from lower altitudes, and a 24.1% decrease in the frequency of long-tongued bees (Front Range: F1,1998 = 94,618, P < 0.0001; Pennsylvania Mountain: F1,1988 = 85.6, P < 0.0001) (Fig. 2, C and D, and table S2) (15). With increasing competition from immigrant species, foraging breadth of resident bees should contract (1920). Yet alpine bumble bee host choice shows the opposite trend. In 2012–2014, we resurveyed bumble bee visitation on Mount Evans and Niwot Ridge in accordance with historical observations (18). Despite a 10-fold difference between past (n = 4099 visits observed) and present (n = 519 visits observed) collection effort, surveys indicate that resident bumble bees have broadened their diet. Resampling historical visitation data to match present collection effort reveals that foraging breadth (Levin’s niche breadth) (15) increased from 2.61 to 7.01 for B. balteatus [z score (Z) = 28.48, P < 0.0001] and 2.09 to 5.07 for B. sylvicola (Z = 19.78, P < 0.0001). Bumble bees have added flowers with shorter and more variable tube depth to their diet (B. balteatusF1,1997 = 7554, P < 0.0001; B. sylvicolaF1,1997 = 64,851, P < 0.0001) (Fig. 2, E and F, and table S3).

figure 2

Fig. 2. Changing bumble bee community composition, bumble bee tongue length distributions, and tube depth distributions of visited flowers over time. (A and B) Bumble bee community composition. (C and D) Bumble bee tongue length. (E and F) Flower tube depth distribution. Bombus species abundance in alpine communities is indicated by the proportion of total foragers (15). Species are ordered by increasing tongue length [in (A), species’ names follow (18)]. Bimodality of the density functions (15) indicates that bumble bee communities contain two predominant phenotypes, short-tongued and long-tongued [(C) and (D)]. (E) and (F) show the tube depth density functions for flowers visited by, respectively, B. balteatus and B. sylvicola in the Front Range [Mount Evans and Niwot Ridge (15)]. For tongue length [(C) and (D)] and tube depth [(E) and (F)], representative density functions for simulated communities (15) are shown.

Mt. Evans & Niwot Ridge: Total population

The proportion that various types of bumble bee contribute to the total amount of bumble bees in the areas of Mt. Evans and Niwot Ridge. Gray bars represent past communities (1966-1969) where white bars represent "present" communities (2012-2014).

Pennsylvania Mountain: Total population

The proportion that various types of bumble bee contribute to the total amount of bumble bees in the area of Pennsylvania Mountain. Gray bars represent past communities (1977-1979) where white bars represent "present" communities (2008, 2011, 2013).

Mt. Evans & Niwot Ridge: Tongue density

The density in tongue length of bumble bees around Mt. Evans and Niwot Ridge. The blue line indicates past tongue lengths (1966-1969) whereas the red indicates present tongue lengths (2012-2014).

Pennsylvania Mountain: Tongue density

The density in tongue length of bumble bees around Pennsylvania Mountain. The blue line indicates past tongue lengths (1966-1969) whereas the red indicates present tongue lengths (2012-2014).

B. balteatus

The density of flowers visited by B. balteatus in relation to the depth of a given flower. The blue line shows past trends (1966-1969) whereas the red line shows present trends (2012-2104).

B. sylvicola

The density of flowers visited by B. sylvicola in relation to the depth of a given flower. The blue line shows past trends (1966-1969) whereas the red line shows present trends (2012-2104).

In response to warmer temperatures and drying soils, flowering has declined in alpine and arctic habitats worldwide (2124). Optimal foraging theory predicts that foragers will expand their niche in response to such resource scarcity (2526). When bumble bees (B. balteatus) encounter low densities of preferred host plants, they incorporate shallower flowers into their diet (F1,194 = 29.39, P < 0.0001) (table S4) (15). The expansion of foraging breadth over time in Front Range bumble bees is congruent with this behavior (Fig. 2, E and F, and table S3). Climate records from Niwot Ridge show warming summer minimum temperatures over the past 56 years (27).We see similar changes on Mount Evans (R2 = 0.383, t1,52 = 5.68, P < 0.0001) and Pennsylvania Mountain (R2 = 0.341, t1,52 = 5.20, P < 0.0001) (fig. S3, A and B), where summer minimums have increased ~2°C since 1960. We used a nonlinear model to characterize the relationship between peak flower density (PFD; flowers per square meter) and summer minimum temperature. For four bumble bee host species monitored in 6 years between 1977 and 2014 (15), average PFD on Pennsylvania Mountain rose as minimum summer temperature increased from 1.8°C to 3.25°C but declined above this value (quadratic R2 = 0.19, t1,17 = –2.18, P = 0.040) (fig. S3C) (15). Temperatures that are associated with reduced flowering (greater than 3.25°C) have become more frequent, occurring on Pennsylvania Mountain in only 12% of years from 1960 to 1985 but 48% of years since 1985 (χ2 = 8.19, df = 1, P < 0.0041) (fig. S3B).

On Pennsylvania Mountain, alpine bumble bees forage over hundreds of meters to provision their nests (28). To ask how warming has affected floral resources at this scale, we measured PFD of six bumble bee host plants from 1977–1980 and 2012–2014 in five habitats along a 400-m altitudinal span (table S5). Land surface area decreases with altitude above tree line in the Rocky Mountains (29), declining by more than an order of magnitude on Pennsylvania Mountain, where 58% of habitable terrain is found below 3800 m and only 4% above 3938 m on the summit (Fig. 3Aand table S5). Because bumble bees forage across the 400-m altitudinal range (28), we evaluated the temporal change in flower production at this landscape scale. For each habitat, we multiplied PFD (flowers per square meter) within sampling plots by surface area (square meters) to estimate of total flower production (15). PFD fell by 73 to 80% within krummholz and slope habitats below 3800 m, which occupy 1.95 km2. Conversely, PFD increased by 75% in 0.10 km2 of summit habitat (F4,385 = 5.55, P = 0.0002) (fig. 3B and table S6). Because declines in flowering occurred at low altitude, they affected the majority of the mountain landscape; in these extensive habitats, millions of flowers were lost. Thus, even with gains of a few thousand flowers on the summit, total food resources for alpine bumble bees on Pennsylvania Mountain have fallen by 60% since the 1970s (Fig. 3C).

figure 3

Fig. 3. Change in flower abundance at landscape and local scales along a 400-m altitudinal gradient on Pennsylvania Mountain. (A) Map showing areas where PFD decreased (1.95 km2), is stable (1.29 km2), and increased (0.10 km2). Unshaded (excluded) areas contain cliff, talus, mining disturbance, and subalpine forest. (B) PFD (mean >± SE) for plots in krummholz (KRUM); tundra slopes (SLOPE); wet meadow (SWALE), false summit (FSUMMIT); and summit (SUMMIT) habitats (N = 6 species; F4,385 = 5.55, P = 0.0002). Asterisks indicate significant differences at P < 0.05. (C) Total flower production (in millions) is the product of total surface area for (A) each habitat (table S5) (15) and (B) mean PFD.

Changing peak flower density

The trend in changing peak flower density (PFD) seen across an area of Pennsylvania Mountain. Areas in red denote a decreasing PFD, areas in blue denote an increasing PFD, areas in gray denote a stable PFD and white areas are excluded.

Mean peak flower density

The mean PFD of the habitats of; krummholz, tundra slopes, wet meadows, false summits, and summits. The gray bars depict the PFD from 1977-1980 where the white bars depict the PFD from 2012-2013.

Total flower production

The total flower production of each noted habitat from 1977-1980 (gray bars) and from 2012-2013 (white bars).

Alpine regions are considered “canaries in the coal mine” for their sensitivity to global warming (29). Using a simple model adapted from (26), we tested whether reduced flowering in other ecosystems could drive the evolution of pollinator foraging traits as indicated for alpine bumble bees (15). The model predicts changes in the energetic advantage of generalization with floral density. Long-tongued bumble bees exhibit greater specialization than that of short-tongued bees (1630). Across a range of flight speed and plant community composition (15), the advantage of generalizing increases as flower density declines (Fig. 4). Theoretical and empirical studies alike suggest that with lower floral resources, fitness advantages of long-tongued specialist phenotypes have diminished, potentially driving the rapid evolution of shorter-tongued bees. We have documented decreases in bumble bee tongue length within species and communities on three peaks in the Rocky Mountains. Our analyses suggest that reduced flower density at the landscape scale is driving this shift in tongue length. Although populations of long-tongued bees are undergoing widespread decline (13), shifts in foraging strategies may allow alpine bumble bees to cope with environmental change. We see broader bumble bee foraging niches, immigration by short-tongued bumble bees, and shorter tongue length within resident bee populations as floral resources have dwindled. In remote mountain habitats—largely isolated from habitat destruction, toxins, and pathogens (31)—evolution is helping wild bees keep pace with climate change.

figure 4

Fig. 4. Relative advantage of generalizing as a function of flower density and the proportion of deep flowers in the community. Outcomes with flight speed of 0.5 m s−1 are shown (15). The generalist is favored when its relative advantage is >1 (pink shading).


Flower Density

Results show with fewer flowers, the evolutionary advantages of having a long-tongued have diminished, potentially driving the rapid evolution of shorter-tongued bees. The analysis from this suggest that reduced flower density at the landscape scale is driving this shift in tongue length.

Changes in foraging techniques

The environmental changes occurring in the alpine have caused shifts in foraging strategies allowing alpine bumble bees to cope with environmental change. The scientists have see broader bumble bee foraging techniques, immigration by short-tongued bumble bees, and shorter tongue length within resident bee populations as floral resources have dwindled.

Supplementary Materials

Materials and Methods

Figs. S1 to S4

Tables S1 to S9

References (3259)

References and Notes

  1. S. A. Cameron et al., Proc. Natl. Acad. Sci. U.S.A. 108, 662–667 (2011).
  2. E. F. Ploquin, J. M. Herrera, J. R. Obeso, Oecologia 173, 1649–1660 (2013).
  3. R. Bommarco, O. Lundin, H. G. Smith, M. Rundlöf, Proc. Biol. Sci. 279, 309–315 (2012).
  4. J. C. Biesmeijer et al., Science 313, 351–354 (2006).
  5. J. C. Grixti, L. T. Wong, S. A. Cameron, C. Favret, Biol. Conserv. 142, 75–84 (2009).
  6. C. Matsumura, J. Yokoyama, I. Washitani, Glob. Environ. Res. 8, 51–66 (2004).
  7. M. Stang, P. G. L. Klinkhamer, N. M. Waser, I. Stang, E. van der Meijden, Ann. Bot. (Lond.) 103, 1459–1469 (2009).
  8. D. P. Vázquez, N. Blüthgen, L. Cagnolo, N. P. Chacoff, Ann. Bot. (Lond.) 103, 1445–1457 (2009).
  9. M. A. Rodríguez-Gironés, A. L. Llandres, PLOS ONE 3, e2992 (2008).
  10. L. D. Harder, Oecologia 57, 274–280 (1983).
  11. N. Muchhala, J. D. Thomson, Proc. R. Soc. Biol. Sci. 276, 2147–2152 (2009).
  12. V. Grant, E. J. Temeles, Proc. Natl. Acad. Sci. U.S.A. 89, 9400–9404 (1992).
  13. R. B. Miller, Evolution 35, 763–774 (1981).
  14. F. S. Gilbert, Ecol. Entomol. 6, 245–262 (1981).
  15. Materials and methods are available as supplementary materials on Science Online.
  16. N. E. Miller-Struttmann, C. Galen, Oecologia 176, 1033–1045 (2014).
  17. P. A. Byron, thesis, University of Colorado, Boulder (1980).
  18. L. W. Macior, Melanderia 15, 1–59 (1974).
  19. R. MacArthur, R. Levins, Am. Nat. 101, 377–385 (1967).
  20. D. W. Inouye, Ecology 59, 672–678 (1978).
  21. C. W. Kopp, E. E. Cleland, J. Veg. Sci. 25, 135–146 (2014).
  22. A. J. Miller-Rushing, D. W. Inouye, Am. J. Bot. 96, 1821–1829 (2009).
  23. T. T. Høye, E. Post, N. M. Schmidt, K. Trojelsgaard, M. C. Forchhammer, Nat. Clim. Change 3, 759–763 (2013).
  24. D. W. Inouye, Ecology 89, 353–362 (2008).
  25. C. Fontaine, C. L. Collin, I. Dajoz, J. Ecol. 96, 1002–1010 (2008).
  26. C. J. Essenberg, Am. Nat. 180, 153–166 (2012).
  27. C. R. McGuire, C. R. Nufio, M. D. Bowers, R. P. Guralnick, PLOS ONE 7, e44370 (2012).
  28. J. C. Geib, J. P. Strange, C. Galen, Ecol. Appl. 25, 768–778 (2015).
  29. P. R. Elsen, M. W. Tingley, Nat. Clim. Change. 5, 772–776 (2015).
  30. G. H. Pyke, D. W. Inouye, J. D. Thomson, Environ. Entomol. 41, 1332–1349 (2012).
  31. D. Goulson, E. Nicholls, C. Botías, E. L. Rotheray, Science 347, 1255957 (2015).
  32. J. H. Cane, J. Kans. Entomol. Soc. 60, 145–147 (1987).
  33. R Development Core Team, R: A language and environment for statistical computing v. 3.1.2 (2014); available at
  34. J. Fox et al., An R Companion to Applied Regression (SAGE Publications, Thousand Oaks, CA, 2013).
  35. J. T. E. Richardson , Educ. Res. Rev. 6, 135–147 (2011).
  36. D. J. Navarro, lsr: Companion to “Learning statistics with R: A tutorial for psychology students and other beginners.” University of Adelaide. Adelaide, Australia (2015); available
  37. W. A. Webber, R. C. Wittmann, Colorado Flora: Eastern Slope, A Field Guide to the Vascular Plants (University Press of Colorado, Boulder, CO, ed. 4, 2012).
  38. D. R. Johnson, D. Ebert-May, P. J. Webber, C. E. Tweedie, Ambio 40, 693–704 (2011).
  39. A. M. Arft et al., Ecol. Monogr. 69, 491–511 (1999).
  40. J. C. Geib, thesis, University of Missouri, Columbia, MO (2010).
  41. D. Young et al., mixtools: Tools for Analyzing Finite Mixture Models v. 1.0.0 (2013); available
  42. R. Levins, Evolution in Changing Environments: Some Theoretical Explorations (Princeton Univ. Press, Princeton, NJ, 1968).
  43. J. Miller, M. Litvak, S. Kelso, A. Vargo , Arct. Alp. Res. 26, 297–303 (1994).
  44. P. J. Clark, F. C. Evans , Ecology 35, 445–453 (1954).
  45. SAS Institute, The SAS system, Release 9.2. SAS Institute, Cary NC (2013).
  46. C. Daly, M. Halbleib, J. I. Smith, W. P. Gibson, M. K. Doggett, G. H. Taylor, J. Curtis, P. P. Pasteris, Int. J. Climatol. 28, 2031–2064 (2008).
  47. U.S. Geological Survey, USGS NED n40w107 1/3 arc-second 2013 1 x 1 degree IMG (2013); available at
  48. Environmental Systems Resource Institute (ESRI), ArcGIS 10.2 (ESRI, Redlands, CA, 2013).
  49. S. Fretwell, H. Lucas Jr. , Acta Biotheor. 19, 16–36 (1969).
  50. D. W. Inouye , Oecologia 45, 197–201 (1980).
  51. E. J. Temeles, C. R. Koulouris, S. E. Sander, W. J. Kress , Ecology 90, 1147–1161 (2009).
  52. C. P. Ellington , J. Exp. Biol. 202,3439–3448 (1999). pmid:1056252
  53. J. Spaethe, J. Tautz, L. Chittka , Proc. Natl. Acad. Sci. U.S.A. 98, 3898–3903 (2001).
  54. B. Heinrich, Bumblebee Economics (Harvard Univ. Press, Cambridge, MA, 2004).
  55. B. Heinrich , J. Comp. Physiol96, 155–166 (1975).
  56. J. M. Pleasants , Ecology 62, 1648–1661 (1981).
  57. J. F. Ornelas, M. Ordano, A. J. De-Nova, M. E. Quintero, T. Garland Jr., J. Evol. Biol. 20, 1904–1917 (2007).
  58. J. Stout, D. Goulson , The influence of nectar secretion rates on the responses of bumblebees (Bombusspp.) to previously visited flowers. Behav. Ecol. Sociobiol. 52, 239–246 (2002).
  59. O. Langsrud , Stat. Comput. 13, 163–167 (2003).
  60. Acknowledgments: We acknowledge L. W. Macior and P. A. Byron for their meticulous work on Rocky Mountain bumble bees; J. Myrick, J. Guinnup, A. Drew, L. Rimmer, J. Stoehr, M. Pallo, L. Hesh, and B. Lubinski for laboratory and fieldwork; and the Mountain Research Station, University of Colorado and Mount Evans Field Station, Denver University for research facilities. The Arapaho National Forest, Niwot Ridge Long-Term Ecological Research (NSF grant DEB-1027341) and Mountain Area Land Trust (Pennsylvania Mountain) provided access to research sites. The Canadian National Collection of Insects; Rocky Mountain Herbarium, University of Wyoming; Kathryn Kalmbach Herbarium, Denver Botanic Garden; University of Colorado Herbarium; Colorado State University Herbarium; and the Missouri Botanical Garden loaned specimens. Research was supported by NSF (grants DEB-79-10786 and 1045322). Data and specific code are archived at DOI: 10.5061/dryad.10278 PRISM Climate Group data for Mount Evans and Pennsylvania Mountain are from