Report

Anthropogenic environmental changes affect ecosystem stability via biodiversity

See allHide authors and affiliations

Science  17 Apr 2015:
Vol. 348, Issue 6232, pp. 336-340
DOI: 10.1126/science.aaa1788

Biodiversity protects grassland stability

How biodiversity interacts with ecosystem stability and productivity is key to understanding the impacts of environmental changes on ecosystem functions. In a series of decade-long experiments in temperate grassland, Hautier et al. manipulated nitrogen, water, carbon dioxide, herbivory, and fire. In all cases, plant species diversity was important for preserving ecosystem function during environmental change. Hence, the preservation and restoration of biodiversity buffer ecosystems against anthropogenic assault.

Science, this issue p. 336

Abstract

Human-driven environmental changes may simultaneously affect the biodiversity, productivity, and stability of Earth’s ecosystems, but there is no consensus on the causal relationships linking these variables. Data from 12 multiyear experiments that manipulate important anthropogenic drivers, including plant diversity, nitrogen, carbon dioxide, fire, herbivory, and water, show that each driver influences ecosystem productivity. However, the stability of ecosystem productivity is only changed by those drivers that alter biodiversity, with a given decrease in plant species numbers leading to a quantitatively similar decrease in ecosystem stability regardless of which driver caused the biodiversity loss. These results suggest that changes in biodiversity caused by drivers of environmental change may be a major factor determining how global environmental changes affect ecosystem stability.

Human domination of Earth’s ecosystems, especially conversion of about half of the Earth’s ice-free terrestrial ecosystems into cropland and pasture, is simplifying ecosystems via the local loss of biodiversity (1, 2). Other major global anthropogenic changes include nutrient eutrophication, fire suppression and elevated fire frequencies, predator decimation, climate warming, and drought, which likely affect many aspects of ecosystem functioning, especially ecosystem productivity, stability, and biodiversity (1, 37). However, to date there has been little evidence showing whether or how these three ecosystem responses may be mechanistically linked. Rather, at present each anthropogenic driver of environmental change has been considered to have its own idiosyncratic syndrome of impacts on ecosystem productivity, stability, and biodiversity (1, 510).

This perspective was recently called into question by a study showing that the initial impacts of nutrient addition on grassland productivity were reduced through time in proportion to the extent to which nutrient addition led to the loss of plant diversity (11). In essence, that study suggested that the positive dependence of productivity on plant diversity (1217), in combination with the negative effect of eutrophication on diversity (8, 18), caused the initial increase in productivity with nitrogen enrichment to diminish over time due to the loss of plant diversity caused by chronic nitrogen fertilization (11). This suggests the hypothesis that other drivers of global environmental change may have biodiversity-mediated effects on ecosystem functioning (19)—that changes in biodiversity resulting from anthropogenic drivers may be an intermediate cause of subsequent changes in ecosystem functioning. Here we test this hypothesis. Numerous biodiversity experiments have shown that reduced plant diversity leads to decreased temporal stability of productivity because of reductions in compensatory dynamics or in asynchronous responses to environmental fluctuations (12, 16, 20, 21). Here, our test determines how experimental manipulations of nitrogen (N), carbon dioxide (CO2), fire, herbivory, and water affect biodiversity and productivity; and if changes in ecosystem stability associated with each environmental driver have the same dependence on biodiversity as observed in biodiversity experiments, or if each driver has an individualistic impact on stability (5, 6).

We perform this particular test because, whereas effects of anthropogenic drivers on biodiversity and productivity have been widely investigated (5, 6, 11), their long-term impacts on the temporal stability of productivity have received less attention, and the few published studies examining a single driver report mixed results (7, 9, 10, 2225). A commonly used measure of stability among many proposed in the ecological literature (26, 27) defines the temporal stability of productivity (S) as the ratio of the temporal mean of productivity to its temporal variability as measured by its standard deviation (SD) (28). This measure of stability is the inverse of the coefficient of variation. Under this definition, a driver could increase stability by increasing the mean productivity relative to the SD, by decreasing the SD relative to the mean productivity, or both. Drivers that increase the SD may also increase stability if there is a correspondingly larger proportional increase in mean productivity (or vice versa) (7, 20, 29). Importantly, given that the temporal mean and SD of productivity can depend on biodiversity (7, 21, 29), drivers might influence stability through their long-term effects on biodiversity. The simultaneous impacts of various drivers on ecosystem biodiversity, productivity, and stability have not previously been explored, thus limiting our current understanding.

Here, we determine if ecologically or societally relevant magnitudes of change in six important anthropogenic drivers influence the stability of ecosystem productivity and whether changes in stability correspond with changes in biodiversity. In particular, we test the hypothesis that changes in biodiversity, regardless of the causal factor, consistently affect the stability of ecosystem productivity.

We used data from 12 experiments that manipulated one or more anthropogenic drivers over a period of 4 to 28 years (table S1). We examine both long-term stability (temporal stability determined using all 4 to 28 years of data collected on aboveground biomass in each experiment) and short-term stability (the temporal stability of each 3-year period of each experiment) and the dependence of these metrics of stability on the concurrent measures of plant species numbers.

We begin by evaluating the extent to which changes in grassland plant diversity, whether experimentally manipulated or in response to other anthropogenic drivers, including N, CO2, fire, herbivory, and water, predict changes in the long-term temporal stability of productivity. Our analyses control for what otherwise might be potentially confounding variables by including only experiments at the Cedar Creek Ecosystem Science Reserve on well-drained sandy soils of east-central Minnesota, USA, that used perennial grassland ecosystems of similar plant species compositions (5). We determined long-term temporal stability, S, as μ/σ, where μ is the average productivity of a plot across all years and σ is the temporal standard deviation in the productivity of that plot across all years. We calculated long-term stability responses as the natural logarithm of the ratio (log response ratio or lrr) of the long-term stability within each treatment plot divided by the average long-term stability in the reference plots (lrr.S). Similarly, we calculated the associated plant species richness responses as the natural logarithm of the ratio of the average richness across all years within each treatment plot divided by the average richness across all years in the reference plots (lrr.rich). Log response ratios quantify the proportional change in treatment plots relative to reference plots. Because lrr.S is the difference between the log response ratio of the temporal mean (lrr.mean) and the log response ratio of the temporal standard deviation (lrr.SD), it separates the effects of anthropogenic drivers on stability into their simultaneous effects on the mean and variance of productivity.

Reference plots were unmanipulated or otherwise had more historically typical conditions, such as high diversity or ambient N, CO2, herbivory, and water conditions or presettlement fire conditions. In particular, we compared biodiversity from plots planted with one, two, and four species to reference plots planted with 16 species, a level representative of a high-diversity (16.3 species m−2) natural grassland community in this area (5). N additions of 270, 170, 95, 54, 34, 20, and 10 kg ha−1 were compared to plots receiving no N, and addition of CO2 and water, fire suppression, and herbivore exclusion were compared to grassland plots with ambient or presettlement conditions. These treatments (except 270, 170, and 95 kg N ha−1 and perhaps the monocultures of biodiversity experiments) also fall within the ranges occurring in natural grassland ecosystems of this region (5).

We found that changes in plant diversity in response to anthropogenic drivers, including N, CO2, fire, herbivory, and water, were positively associated with changes in temporal stability of productivity (black line in Fig. 1; Fig. 2, C and D). This positive association was independent of the nature of the driver, resulting in parallel relationships (all colored lines except red in Fig. 1; table S2). This suggests that biodiversity-mediated effects on stability are independent of the factor driving changes in biodiversity. Moreover, the positive association between changes in plant diversity and changes in stability in response to anthropogenic drivers was similar to that observed in two neighboring experiments that directly manipulated plant diversity (compare the black and red lines in Fig. 1) (21). Thus, changes in biodiversity resulting from anthropogenic environmental changes have similar effects on stability as observed in biodiversity experiments, suggesting that changes in biodiversity may be an intermediary factor influencing how anthropogenic environmental changes affect ecosystem stability. For example, whether a 30% change in plant diversity (lrr.rich = –0.357) resulted from elevated N, CO2, or water or from herbivore exclusion, fire suppression, or direct manipulation of plant diversity, stability tended to decrease in parallel by 8% (lrr.S = –0.082). This conclusion is supported by analyses showing that there was no remaining effect of anthropogenic drivers on changes in stability after biodiversity-mediated effects were taken into account (table S3) and that changes in stability based on biodiversity manipulations predict changes in stability in response to anthropogenic drivers (fig. S1).

Fig. 1 Human-driven environmental changes affect ecosystem stability via biodiversity.

Effect of anthropogenic drivers of environmental change on the stability of productivity, as mediated by experimentally imposed changes in biodiversity [red line; slope and 95% confidence intervals (CIs): 0.14 (0.08 to 0.20)], or from biodiversity changes arising from anthropogenic drivers including N, CO2, water, fire, and herbivory [black and other colored lines; slopes and 95% CIs: 0.22 (0.15 to 0.31)]. Black and red lines are based on separate fits; their similar slopes show that changes in biodiversity caused by anthropogenic drivers have effects on stability similar to those resulting from experimentally imposed changes in plant biodiversity (F1,561 = 3.29, P = 0.07). Relative changes were calculated as the natural logarithm of the ratio (lrr) of the variable within each treatment plot divided by the average of the variable in the reference plots. Black line is the fixed-effect linear regression slope across all anthropogenic drivers in the mixed-effects model; colored lines show trends for each driver. Colors for the points correspond to treatments in Fig. 2.

Fig. 2 Simultaneous effect of human-driven environmental changes on ecosystem productivity, stability, and biodiversity.

Effect of anthropogenic drivers of environmental change on relative changes in the (A) mean, (B) standard deviation (SD), (C) stability of ecosystem productivity, and (D) plant diversity. Treatment effects are shown with their 95% CI such that treatments with intervals overlapping zero are not significantly different from zero (table S4).

We next evaluated the extent to which changes in temporal stability of productivity in response to anthropogenic drivers were caused by changing the temporal mean of productivity or the temporal variance of productivity. We found that when a driver of environmental change caused mean productivity to change, it did not consistently lead to higher or lower stability of productivity (Fig. 2 and table S4). For example, decreases in biodiversity from 16 species to one, two, and four species decreased both the temporal mean and stability of productivity (Fig. 2, A and C). By contrast, addition of N, CO2, and water; fire suppression; and herbivore exclusion generally increased the temporal mean of productivity, although not always significantly (Fig. 2A), but either increased (N addition of 10 kg ha−1, fire suppression, and water addition), reduced (N addition of 270, 170, 95, and 54 kg ha−1), or had no detectable effects (N addition of 34 and 20 kg ha−1, addition of CO2, and herbivore exclusion) on stability (Fig. 2C). These differing effects on stability (Fig. 2C) were due to differences in the direction and magnitude of drivers’ impact on mean productivity (Fig. 2A) compared to their variance (Fig. 2B). For example, experimental decreases in biodiversity caused a larger decrease in mean productivity than in its variance, resulting in decreased stability; whereas N addition of 10 kg ha−1, fire suppression, and water addition each caused a larger increase in mean productivity than in its variance, resulting in increased stability. By contrast, N addition of 270, 170, 95, and 54 kg ha−1 caused a larger increase in the variance than the mean, resulting in reduced stability. We do not expect the direction and magnitude of changes in the numerator or denominator of the stability ratio to be universal. For example, in other biodiversity experiments, decreases in biodiversity caused a larger decrease in the variance of productivity than the mean (29). Our results, however, do indicate that drivers consistently reduce stability when they reduce biodiversity.

Together, these results suggest that changes in biodiversity, whether experimentally manipulated or in response to other anthropogenic drivers, caused consistent changes in ecosystem stability of productivity (Figs. 1 and 2, C and D) not because of consistent effects of a driver or biodiversity on either the temporal mean of productivity or on its temporal variance (Fig. 2, A and B) but rather because of consistent effects on their ratio, which is stability (Figs. 1 and 2, C and D). The repeatedly observed quantitative effects of changes in biodiversity on ecosystem stability in this study are consistent with predictions of ecosystem stability by models of interactions among species that coexist because of interspecific trade-offs (30). They are also consistent with results of numerous biodiversity experiments (29).

We found no evidence that biodiversity-mediated effects on stability were caused by similar shifts in the abundances of functional groups or species (fig. S2). For example, although diversity and stability declined, native perennial C4 grasses increased under herbivory exclusion (e.g., Sorghastrum nutans) and declined under high levels of chronic nitrogen enrichment (e.g., Schizachyrium scoparium), while non-native perennial C3 grasses declined under herbivory exclusion (e.g., Koeleria cristata) and increased under high levels of chronic nitrogen enrichment (e.g., Agropyron repens). Thus, various drivers led to similar changes in stability by causing changes in biodiversity, even though the various drivers had different effects on functional groups and particular species.

We also assessed whether the diversity and stability responses were consistent through time by dividing the 4 to 28 years of annual data into overlapping intervals of three consecutive years and calculating short-term stability and average species richness for each interval. This allows us to account for the effects of the different duration of the experiments (31). Effects of anthropogenic drivers on diversity and short-term stability were consistent through time. Specifically, diversity and stability had a weak tendency to decrease in unison with increasing treatment duration independently of the nature of the driver, resulting in parallel negative relationships (Fig. 3). These results further suggest that the decrease in stability over time was associated with declining plant diversity in response to anthropogenic drivers.

Fig. 3 Temporal trends in effect sizes of ecosystem stability and biodiversity responses to anthropogenic drivers of environmental change.

Effects of anthropogenic drivers on (A stability (F13,220 = 30.6, P < 0.001) and (B) diversity (F10,154 = 103.3, P < 0.001) were consistent through time (Drivers × Time: P > 0.1 in both cases). Stability (F1,154 = 86.5, P < 0.001) and diversity (F1,220 = 24.8, P < 0.001) had a weak tendency to decrease with increasing treatment duration. Data were divided into overlapping intervals of 3 years reported as posttreatment period after initiation of the experiment (31), with diversity and stability determined for each interval. Colors for the points and lines correspond to treatments in Fig. 2.

In total, we found that the loss of plant diversity was associated with decreased stability not only in experiments that manipulate diversity (20, 21) but also when biodiversity changed in response to other anthropogenic drivers. In combination with recent demonstrations that biodiversity is a major determinant of productivity (5, 6, 11), these findings suggest that any drivers of environmental change that affect biodiversity are likely to have long-term ecosystem impacts that result from these changes in biodiversity (19). Furthermore, biodiversity-mediated effects on stability did not qualitatively depend either on the particular factor that caused the change in biodiversity or on shifts in the abundance of particular functional groups or species. Altogether, our multiyear experiments suggest that there may be a universal impact of biodiversity change on ecosystem stability in response to anthropogenic environmental changes, with decreased plant species numbers leading to lower ecosystem stability regardless of the cause of biodiversity loss. Our work suggests that conservation policies should encourage management procedures that restore or maintain natural levels of biodiversity or minimize the negative impacts of anthropogenic global environmental changes on biodiversity loss to ensure the stable provision of ecosystem services.

Supplementary Materials

www.sciencemag.org/content/348/6232/336/suppl/DC1

Materials and Methods

Figs. S1 and S2

Tables S1 to S4

References (3239)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The research leading to these results received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 298935 to Y.H. and has been supported by the U.S. National Science Foundation (NSF) Long-Term Ecological Research (LTER) Program at Cedar Creek (DEB-8811884, DEB-9411972, DEB-0080382, DEB-0620652, and DEB-1234162), Biocomplexity Coupled Biogeochemical Cycles (DEB-0322057), Long-Term Research in Environmental Biology (DEB-0716587, DEB-1242531), and Ecosystem Sciences (NSF DEB-1120064) Programs, as well as the U.S. Department of Energy (DOE) Program for Ecosystem Research (DE-FG02-96ER62291) and the U.S. DOE National Institute for Climatic Change Research (DE-FC02-06ER64158). We thank T. Mielke, D. Bahauddin, K. Worm, S. Barrott, E. Lind, and many summer interns for their assistance with this research. We also thank R. S. L. Veiga, W. S. Harpole, T. Züst, and M. Tanadini for suggestions that improved the manuscript. For detailed methods and original data, see www.cedarcreek.umn.edu/research/data. See table S1 for links to data for each experiment and supplementary materials for guidance on data access. The authors declare no conflict of interests. Author contributions: Y.H. and D.T. developed and framed the research question; D.T., F.I., E.W.S., E.T.B., and P.B.R. designed research; D.T., E.W.S., E.T.B., and P.B.R. performed research; Y.H. and F.I. analyzed data; and Y.H. wrote the paper with inputs from all coauthors.
View Abstract

Navigate This Article