Crossing the threshold: the power of multi-level experiments in identifying global change responses


Predicting how plants will respond to multiple global changes and how these responses will in turn influence key ecosystem processes has become one of the most compelling topics in modern ecology. Single-factor experiments have provided preliminary information on community and ecosystem responses to individual global change drivers. However, manipulations of multiple global change factors have revealed that responses of communities and ecosystems to one factor are often modified by a second, and further highlight the need to consider nonlinear responses to global changes (Templer & Reinmann, 2011). Moreover, variation in the magnitude of global change factors across space and time can strongly influence their effects on plant community patterns of dominance, productivity and potential carbon sink–source dynamics (Ostle et al., 2009; Phoenix et al., 2011). In this issue of New Phytologist, Bradford et al. (pp. 462–471), build upon these findings by testing ecosystem and plant community responses to multiple levels of two key global change factors: atmospheric [CO2] and nitrogen (N) enrichment (Fig. 1). Using grassland microcosms, they show strong interactive effects of [CO2] and N enrichment on ecosystem responses; notably, [CO2] effects on aboveground net primary productivity (ANPP) depended on the level of N addition (Fig. 1a). These results provide powerful information on how ecosystem processes can respond interactively, rather than additively, to multiple global change factors.

‘… their findings suggest that changes in abiotic soil properties in response to one global change factor may influence plant productivity responses to changes in a second factor.’

Figure 1.

Conceptual flow diagram showing multi-level global change effects on ecosystem processes (i.e. aboveground net primary productivity, ANPP) and plant community composition as tested in grassland microcosms by Bradford et al. (this issue of New Phytologist, pp. 462–471). Effects of elevated atmospheric [CO2] and nitrogen (N) enrichment on ANPP are mediated by the functional leaf and root traits of the component plant species. ANPP responses to elevated [CO2] were nonlinear and were contingent on the level of N addition (a). The two dominant plant species (x and y) varied in leaf trait plasticity in response to N addition (b), which would have altered their competitive balance, and hence, resulted in a shift in plant community composition (c). Bradford et al. found that the plant community responses were unaffected by [CO2]; however, in other cases individual plant and community responses could be expected to depend on interaction effects of [CO2] and N level (Luo et al., 2004).

Nonlinear responses of plants and ecosystems to global change are often likely to be common, dynamic, and ecologically relevant. Resource co-limitation could be a major driver of nonlinear plant and ecosystem responses to global changes; physiological and biogeochemical constraints and dynamic trophic interactions may further contribute to nonlinearity (Tylianakis et al., 2008). While two-level experiments are unable to capture such nonlinear responses (Fig. 2), the use of multi-level experiments allow partitioning of global change effects into linear, quadratic, and other shaped responses. Identifying the shape of so-called ‘response functions’ for plants and ecosystem processes to global change factors is an important challenge in global change research (Fig. 2). In recent years, it has become increasingly clear that human-induced global changes have the potential to push ecosystems past critical thresholds, or ‘tipping points’, at which an ecosystem shifts abruptly from one state to another (Lenton et al., 2008; Scheffer et al., 2009). While predicting such tipping points before they are reached is difficult, multi-level experiments are integral to identifying when an ecosystem may cross a tipping point that could shift it to an alternative state. In the context of multiple global changes, it is therefore of particular interest to test if and how response functions for one global change factor (e.g. [CO2]) depend on another (e.g. N enrichment). Interdependence of global change factors can reshape response functions and may shift thresholds and tipping points (Fig. 2).

Figure 2.

Potential scenarios of plant and ecosystem responses to single and two-factor global changes. Other scenarios are possible. (a, e) Positive linear response. (b, f) Logistic response, indicating a ‘tipping point’ (indicated by white dots) at which a small shift in environmental conditions qualitatively alters the state of the response variable. (c, g) Inverse exponential response. (d, h) Quadratic, ‘hump-shaped’ response. Two-level experimental manipulations of a single global change factor (e.g. ambient atmospheric [CO2] vs elevated [CO2]) would not capture nonlinear responses to a global change factor, as indicated by the black dots in (a–d). Instead, multilevel experiments are needed to identify the shape of response functions. Moreover, other global change factors may alter the position or shape of response curves. (a–d) Additive effects of a second global change factor would shift the response curve along the y-axis, but would not alter the shape of the curve. (e–h) Non-additive, non-linear interaction effects of a second global change factor would reshape the response curve. In the case of a logistic response, this would imply a shift of the tipping point.

The availability of key nutrients such as N and phosphorus (P) can limit plant responses to elevated [CO2] (Reich et al., 2006; Lewis et al., 2010), and therefore influence the potential carbon (C) sink strength of vegetation. Bradford et al. support these previous findings by demonstrating co-limitation of plant productivity by [CO2] and N. The shape of ANPP response across varying [CO2] levels depended upon N addition rate, indicating that ANPP responses to elevated [CO2] are N-limited. Their findings further indicate how ecosystems that differ in nutrient status, as a result of either human activity (e.g. fertilization) or natural disturbance (e.g. wildfire), and subsequent successional development in these ecosystems, may respond to increasing [CO2]. The highest ANPP was reached at the highest N addition rate under intermediate [CO2] levels (between 450 and 550 ppm). Additionally, increasing [CO2] levels have been shown to increase water use efficiency (Morgan et al., 2011), a result supported by Bradford et al. A decreased impact of acidification caused by N additions in soils exposed to high [CO2] levels may have further contributed to differences in plant responses across treatments. Although the authors were unable to disentangle the relative role of N additions, soil moisture and pH on ANPP responses under different [CO2] levels, their findings suggest that changes in abiotic soil properties in response to one global change factor may influence plant productivity responses to changes in a second factor. Further investigation is needed to tease apart the relative contribution of such individual factors to future plant community and ecosystem responses, including carbon sink–source dynamics.

Bradford et al. also tested for effects of N addition regime and found greater plant N acquisition under acute (fertilization-proxy) N vs chronic (deposition-proxy) N addition. Further, ANPP responses to acute N additions increasingly diverged from chronic-N additions as the rate of addition increased. With increasing soil inorganic N concentrations under acute N addition, plants may acquire N for growth more rapidly (Kraiser et al., 2011). As such, these results suggest a clear difference in ecosystem responses and the potential future C sink strength between fertilized agricultural ecosystems and unmanaged, less fertile ecosystems to rising atmospheric [CO2]. Increased understanding of plant productivity responses to elevated [CO2] across agricultural and natural ecosystems could contribute significantly to the accuracy of global C cycling models.

Global change effects on plant community productivity are expected to drive major shifts in the relative dominance of the component species, with potential ecosystem-level consequences (Kardol et al., 2010a; Langley & Megonigal, 2010). Bradford et al. found a shift in the relative dominance of two common grass species. The species that dominated under high levels of N addition was also the one that displayed the greatest plasticity in functional leaf traits associated with light interception and photosynthetic rates (Fig. 1b,c). Surprisingly, the constructed grassland ecosystems by Bradford et al. did not reveal any interactions between the [CO2] and N treatments on plant community dominance. While the authors interpret this as an indication that plant community response to global change might be easier to predict than ecosystem responses, this will not necessarily hold true for other constructed or natural ecosystems. Bradford et al. used the shift in proportional abundance of the two dominant plant species as a measure of community responses. While this approach may inform on the ecosystem consequences of plant community composition shifts in response to elevated [CO2] and N addition, it is a simplification of the complex community dynamics that are expected as a consequence of global change under natural conditions. The competitive balance between dominant and subordinate species may change (Kardol et al., 2010b) and lead to species turnover. As such, some species may become locally extinct if global change pushes them outside of their environmental envelopes, while new species may enter the community as their ranges shift. A future focus explicitly on plant functional response and effect traits (sensu Suding et al., 2008) could assist in better prediction of plant community responses to global changes and the resultant ecosystem consequences.

Aboveground responses to global change are not necessarily reflected belowground (e.g. Klironomos et al., 2005). Empirical evidence provided by Bradford et al., revealing how foliar and root responses to global change factors differ, further calls into question whether aboveground responses to global changes act as a suitable proxy for belowground responses. In addition, because aboveground communities are inherently linked to belowground communities, interactions between aboveground and belowground subsystems may mediate ecosystem responses to global change (Antoninka et al., 2009; Kardol et al., 2010a). Careful consideration must therefore be given when generalizing among the responses of these two subsystems to global change. Whole-ecosystem responses to global change may depend on complex interactions between aboveground and belowground communities and the functions they provide. Although our knowledge in this realm is lacking, the findings from Bradford et al. further highlight the need for coordinated approaches to incorporate plant–soil linkages in predictions of future ecosystem responses to global change (Ostle et al., 2009).

Temperate grasslands have been the most commonly targeted systems in global climate change experiments due to their cosmopolitan distribution, economic importance, and the relative ease with which experiments can be established in such systems (White et al., 2011). Although these studies have been integral in establishing initial parameters for climate modeling, we must be cautious in extrapolating results from one biome to another. Intensifying research in other major biomes which contribute significantly to global C storage, such as boreal, tropical and arctic systems should be a major focus of future studies (e.g. Fig. 3). Establishing a series of long-term experiments globally with a systematic, universally standardized methodology accounting for multi-level global change factors would be an obvious next step. Performing long-term research in less accessible, climatically harsh systems is challenging both logistically and economically, but broadening our scope of study is the only way in which we can hope to achieve a holistic view of ecosystem responses to climate change on a global scale.

Figure 3.

The Spruce and Peatland Responses Under Climatic and Environmental Change (SPRUCE) experiment ( is being set up in an ombrotrophic bog in northern Minnesota (USA), to assess the response of this high-carbon ecosystem to increases in temperature and elevated atmospheric [CO2]. A range of temperature increases will be imposed in a regression design and the series of warming treatments will be repeated under ambient and elevated [CO2]. This experimental design will support questions about the response functions to temperature and whether elevated [CO2] reshapes the response function. The boardwalks (under construction) show the location of some of the experimental plots where a 12-m diameter open-top chamber, coupled with belowground heaters, will be installed (a). (b) A prototype chamber at the Oak Ridge National Laboratory in Tennessee, USA. Photographs courtesy of ORNL.