Climate Change, Keystone Predation, and Biodiversity Loss

See allHide authors and affiliations

Science  25 Nov 2011:
Vol. 334, Issue 6059, pp. 1124-1127
DOI: 10.1126/science.1210199


Climate change can affect organisms both directly via physiological stress and indirectly via changing relationships among species. However, we do not fully understand how changing interspecific relationships contribute to community- and ecosystem-level responses to environmental forcing. I used experiments and spatial and temporal comparisons to demonstrate that warming substantially reduces predator-free space on rocky shores. The vertical extent of mussel beds decreased by 51% in 52 years, and reproductive populations of mussels disappeared at several sites. Prey species were able to occupy a hot, extralimital site if predation pressure was experimentally reduced, and local species richness more than doubled as a result. These results suggest that anthropogenic climate change can alter interspecific interactions and produce unexpected changes in species distributions, community structure, and diversity.

Predictions concerning biological responses to climate change are largely based on the environmental tolerances of individual species and the assumption that these species will remain within their bioclimatic envelope as conditions change (1). At coarse scales, these predictions generally match observed changes in distribution and abundance across gradients such as latitude, elevation, and depth (2, 3). However, changing climatic conditions also lead to altered community composition (4) and shifts in the strength and sign of interspecific interactions (5, 6)—changes that may greatly affect community dynamics and ecosystem function (7). Because species interactions can accelerate, retard, prevent, or even reverse predicted biotic changes based solely on simplistic models, interspecific relationships must be incorporated into the predictive framework of climate change (1, 8). Although distributional shifts forced by interspecific relationships have been demonstrated in the lab (9) and predicted by data-driven models (10), appropriate field tests are largely absent.

Rocky intertidal communities are ideal test-beds for studying the effects of climatic warming because many intertidal organisms already live very close to their thermal tolerance limits (11). Intertidal species’ distributional limits are correlated with their upper thermal tolerance (12), and changes in their distribution and abundance over time are associated with warming temperatures (1315). However, species distributions are also strongly influenced by interspecific interactions (16), and these interactions are temperature-sensitive (6). Observed ecological patterns will therefore depend on both environmental stress and interspecific interactions (10, 17, 18).

In this study, I examined the roles of temperature and predation on the intertidal community in the Salish Sea, which spans a regional-scale thermal gradient from west to east (Fig. 1 and table S1). The oceanic terminus of the Salish Sea (the western end of the Strait of Juan de Fuca) is exposed to cool maritime weather, frequent fog and cloud cover, and early morning low tides, resulting in minimal intertidal thermal stress. The more eastern portions of the Salish Sea are warmer and sunnier, receive less cooling wave swash and spray, and feature summer low tides near midday. As a result, mid-intertidal rocks and organismal body temperatures become progressively hotter from west to east (Fig. 1) (19, 20).

Fig. 1

Study region and thermal gradient. The Salish Sea includes the Strait of Juan de Fuca and the interconnected water bodies to the north and south of the Strait’s eastern terminus. The study sites indicated by gray squares were used in the spatial comparison of temperature and zonation. Gray circles represent sites used for the temporal comparison. The graph shows the increase in mid-intertidal rock temperatures from west to east.

To determine the ecological consequences of this spatial gradient in thermal stress, I first surveyed vertical zonation patterns of sessile invertebrate species and their principle predator. The upper limits of the mussels Mytilus californianus and M. trossulus and the barnacles Semibalanus cariosus and Balanus glandula were all negatively correlated with mid-intertidal rock temperatures (Fig. 2, A to D). In contrast, the upper foraging limit of the predatory sea star Pisaster ochraceus was independent of the thermal stress gradient (Fig. 2E). The slope of upper limits versus temperature was significantly steeper for prey species than for the predators [analysis of covariance (ANCOVA) using site means, species nested within trophic level; trophic level × temperature interaction F = 9.37, P = 0.006] (Fig. 2F).

Fig. 2

Relation between zonation patterns and temperature. The upper limits of the barnacles Balanus glandula (A) and Semibalanus cariosus (B) and mussels Mytilus trossulus (C) and M. californianus (D) were negatively correlated with mid-intertidal substratum temperature (ANCOVA, temperature effect: P ≤ 0.001 in all cases, see table S3 for details), whereas the upper foraging limit of the predatory sea star Pisaster ochraceus (E) was not (ANCOVA, temperature effect P = 0.292). Regression lines in (A) to (E) represent data pooled across sites. These five regression lines are superimposed in (F); note that the regression lines for S. cariosus and M. trossulus overlap. The y axes represent scaled intertidal heights, where 0 = extreme low water and 1 = extreme high water for the year in which the data were collected (2006).

Because upper distributional limits of predator and prey do not shift in unison, there exists a thermally forced reduction in enemy-free space that could result in the exclusion of some sessile species from thermally stressful sites, as has been shown for an intertidal alga (17). The reduction and/or disappearance of sessile invertebrates from west to east is related to their vertical proximity to the foraging range of P. ochraceus; mid-intertidal mussels (M. californianus) do not occur at the two easternmost study sites, and higher-intertidal M. trossulus and S. cariosus are exceedingly rare and do not typically reach adulthood at the most eastern sites (table S1). To test the importance of predation in a thermally stressful environment, I experimentally excluded P. ochraceus from plots in the high-intertidal barnacle zone on Saddlebag Island, Washington, which was the easternmost and hottest site along the gradient. M. trossulus and S. cariosus, which were common at cool sites but virtually absent on Saddlebag Island, came to dominate free space in Pisaster-exclusion cages (Fig. 3A). Because M. trossulus and S. cariosus are ecosystem engineers and provide cool, moist microhabitats for a diverse suite of other species, total species richness increased by a factor of 2.5 after the exclusion of P. ochraceus (Fig. 3B and tables S4 and S5).

Fig. 3

Effects of predator removals at a “hot” site (Saddlebag Island). (A) P. ochraceus exclusion altered sessile invertebrate cover in the stressful high-intertidal barnacle zone. Data are means ± SE. Treatment differences (nonparametric Kruskal-Wallis tests were used to overcome departures from normality) were significant for M. trossulus2 = 18.0, ***P < 0.0001), S. cariosus2 = 12.3, **P = 0.002), and C. dalli2 = 20.8, ***P < 0.0001), but not for B. glandula2 = 0.74, P = 0.69). The few M. trossulus and S. cariosus recorded outside of predator exclusion cages were recent recruits that would typically be consumed well before adulthood. (B) Species richness (all taxa) in P. ochraceus exclusion experiments. Data are means ± SE. P. ochraceus predation had a strong negative effect on overall species richness (Kruskal-Wallis test, χ2 = 22.4, ***P < 0.0001).

These spatial patterns and experimental results suggest that the upper limits of sessile species are largely controlled by thermal stress, lower limits are determined by predation, and the intertidal (vertical) variation in predation pressure is not sensitive to among-site variation in temperature. This leads to three testable predictions with regard to climate change: (i) Climatic warming through time should result in downshore shifts in the upper limits of sessile invertebrates; (ii) lower limits should remain invariant through time; and (iii) the vertical zones of sessile species should become compressed, potentially resulting in local extinction at some sites. To test these predictions, I compared Salish Sea zonation patterns in acorn barnacles (Chthamalus dalli and B. glandula combined) and mussels (M. californianus) in 2009 to 2010 to a historical data set from 1957 to 1958 (21). Sampling sites were relocated with the aid of the original investigator and, in most cases, the 2009 to 2010 surveys are believed to be within ~30 m of the original survey sites (22).

The 52 years separating the two sampling intervals span a period of climatic warming. During the latter half of the 20th century, maximum air temperatures near the eastern and western ends of the Strait of Juan de Fuca increased by ~0.2° and ~0.13°C per decade, respectively, and mean annual water temperatures along the southern and western Vancouver Island coast have warmed by ~0.08° to 0.11°C per decade (14, 23). Average daily maximum air temperatures during the summer, which are particularly relevant to thermal stress experienced in the intertidal zone, have warmed even more rapidly. At Victoria, summer average daily maxima have risen approximately linearly at a rate of 0.654°C per decade since 1950, which corresponds to an increase of 3.40°C over 52 years (Fig. 4A). As predicted, invertebrate upper limits have shifted significantly downward during this period of warming, whereas lower limits did not change (Fig. 4B). As a result, the vertical range of M. californianus was reduced by 50.9 ± 10.0 cm (mean ± SE; one-tailed paired t test: N = 15 sites, t = –5.08, P < 0.0001), which corresponds to a 50.6% reduction of the M. californianus zone. Associated with this collapse in vertical range were M. californianus local extinctions at 13.3% (2 of 15) of the sites resurveyed in 2009 to 2010. Historical data for M. trossulus were only available for one site (former vertical range = 49 cm). M. trossulus had been completely eliminated from that site by 2010, with the exception of three small juveniles found under a single rock.

Fig. 4

Changes in temperature and sessile invertebrate zonation through time. (A) Long-term trend in average daily maximum air temperature for the summer months (June to August) in Victoria, British Columbia. Data are second-generation homogenized temperatures, which account for nonclimatic shifts such as changes in sensors or observing practices, obtained from Environment Canada. The warming trend is highly significant (F = 98.4, P < 0.0001, R2 = 0.625). (B) Upper limits of acorn barnacles (open triangles) and mussels (open squares) were significantly lower in 2009 to 2010 for both taxa (one-tailed paired t tests, acorn barnacles: N = 18, t = –3.30, P = 0.002; mussels N = 13, t = –4.23, P < 0.001). Mussel lower limits (black squares) were not significantly different (two-tailed paired t tests, mussels N = 13, t = 1.15, P = 0.27; historical lower limits for acorn barnacles were not available for comparison). Data are means ± SE. The y axis is height above Canadian chart datum, which approximates the lowest astronomical tide.

Large-scale patterns of intertidal species’ distributions in the Salish Sea result from the increased overlap between the foraging range of the dominant predator and the habitable zone of its prey. At cool, wave exposed sites, sessile invertebrates can occupy shore levels well above the foraging range of P. ochraceus. At hotter, wave-protected sites, this high shore refuge becomes unavailable as thermal and desiccation stresses restrict sessile invertebrates to lower shore levels (19). In contrast, mobile P. ochraceus are capable of avoiding or adjusting to periods of thermal stress (24, 25); thus, it is not surprising that the upper limit of P. ochraceus foraging is independent of the regional-scale stress gradient. As enemy-free space is reduced or eliminated at hotter sites, certain sessile species are excluded from these sites [see (26) for a wave-exposure analog]. A similar phenomenon pertains to climatically forced change through time. Upper limits of sessile species have shifted downward, presumably due to warming, whereas their lower limits have remained stationary, presumably due to constant predation pressure. Just as M. californianus is excluded from eastern sites by a lack of predator-free space, it has also suffered local extinctions through time as a result of climatic warming. Because species such as mussels provide critical habitat for an array of mesofaunal species (27), the loss of mussel beds over time has probably resulted in declines in species richness similar to those I have demonstrated experimentally on the stressful shores of Saddlebag Island.

These results are applicable to a variety of systems through both space and time. At local scales, experimental warming increased the vertical (plant-stem scale) overlap of two species of spiders in New England grasslands (28). This increased overlap resulted in the elimination of one spider from the system via intraguild predation and had considerable direct and indirect consequences for both herbivores and primary producers (28). At broader scales, observations and predictions regarding the effects of climate change in strongly zoned systems (such as montane forests or the coastal marine benthos) invoked changing patterns of zonation (8, 29, 30). Indeed, many species are predicted to go, or have already gone, locally extinct as their vertical zones collapse in response to biotic, abiotic, and anthropogenic changes (30, 31). More broadly, these results highlight the importance of incorporating interspecific interactions into predictions regarding ecological responses to climate change. Ecological change can only be accurately anticipated if we are able to understand the ways in which biotic and abiotic factors interact to determine the distribution and abundance of species in space and time.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1

Tables S1 to S6

References (3243)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: T. Widdowson was instrumental in locating the historical data collection sites and accurately repeating the original surveys; his help is gratefully acknowledged. I also thank P. Martone, B. Matthews, M. O’Connor, R. T. Paine, T. Widdowson, and two anonymous referees for thoughtful comments on the manuscript. The Huu-ay-aht First Nation, Beecher Bay First Nation, Makah Tribal Nation, Correctional Service Canada, the Pfaff family, Point No Point Resort, and Friday Harbor Laboratories provided access to field sites. Historical temperature data for Victoria, British Columbia, are reproduced and distributed with the permission of the government of Canada. Funding was provided by the National Science and Engineering Research Council (Canada) and the Padilla Bay National Estuarine Research Reserve. Data are archived in the supporting online material.
View Abstract

Stay Connected to Science

Navigate This Article