Climate extremes will elicit responses from the individual to the ecosystem level. However, only recently have ecologists begun to synthetically assess responses to climate extremes across multiple levels of ecological organization. We review the literature to examine how plant responses vary and interact across levels of organization, focusing on how individual, population and community responses may inform ecosystem-level responses in herbaceous and forest plant communities. We report a high degree of variability at the individual level, and a consequential inconsistency in the translation of individual or population responses to directional changes in community- or ecosystem-level processes. The scaling of individual or population responses to community or ecosystem responses is often predicated upon the functional identity of the species in the community, in particular, the dominant species. Furthermore, the reported stability in plant community composition and functioning with respect to extremes is often driven by processes that operate at the community level, such as species niche partitioning and compensatory responses during or after the event. Future research efforts would benefit from assessing ecological responses across multiple levels of organization, as this will provide both a holistic and mechanistic understanding of ecosystem responses to increasing climatic variability. This article is part of the themed issue ‘Behavioural, ecological and evolutionary responses to extreme climatic events’.

As droughts become longer and more intense, impacts on terrestrial primary productivity are expected to increase progressively. Yet, some ecosystems appear to acclimate to multiyear drought, with constant or diminishing reductions in productivity as drought duration increases. We quantified the combined effects of drought duration and intensity on aboveground productivity in 74 grasslands and shrublands distributed globally. Ecosystem acclimation with multiyear drought was observed overall, except when droughts were extreme (i.e., ≤1-in-100-year likelihood of occurrence). Productivity losses after four consecutive years of extreme drought increased by ~2.5-fold compared with those of the first year. These results portend a foundational shift in ecosystem behavior if drought duration and intensity increase, from maintenance of reduced functioning over time to progressive and profound losses of productivity when droughts are extreme.

In semiarid regions, vegetation constraints on plant growth responses to precipitation (PPT) are hypothesized to place an upper limit on net primary productivity (NPP), leading to predictions of future shifts from currently defined linear to saturating NPP–PPT relationships as increases in both dry and wet PPT extremes occur. We experimentally tested this prediction by imposing a replicated gradient of growing season PPT (GSP, n = 11 levels, n = 4 replicates), ranging from the driest to wettest conditions in the 75‐yr climate record, within a semiarid grassland. We focused on responses of two key ecosystem processes: aboveground NPP (ANPP) and soil respiration (R_s). ANPP and R_s both exhibited greater relative responses to wet vs. dry GSP extremes, with a linear relationship consistently best explaining the response of both processes to GSP. However, this responsiveness to GSP peaked at moderate levels of extremity for both processes, and declined at the most extreme GSP levels, suggesting that greater sensitivity of ANPP and R_s to wet vs. dry conditions may diminish under increased magnitudes of GSP extremes. Underlying these responses was rapid plant compositional change driven by increased forb production and cover as GSP transitioned to extreme wet conditions. This compositional shift increased the magnitude of ANPP responses to wet GSP extremes, as well as the slope and variability explained in the ANPP–GSP relationship. Our findings suggest that rapid plant compositional change may act as a mediator of semiarid ecosystem responses to predicted changes in GSP extremes.

Experiments are widely used in ecology, particularly for assessing global change impacts on ecosystem function. However, results from experiments often are inconsistent with observations made under natural conditions, suggesting the need for rigorous comparisons of experimental and observational studies. We conducted such a “reality check” for a grassland ecosystem by compiling results from nine independently conducted climate change experiments. Each experiment manipulated growing season precipitation (GSP) and measured responses in aboveground net primary production (ANPP). We compared results from experiments with long‐term (33‐yr) annual precipitation andANPPrecords to ask if collectively (n = 44 experiment‐years) experiments yielded estimates ofANPP, rain‐use efficiency (RUE, grams per square meter ANPP per mm precipitation), and the relationship betweenGSPandANPPcomparable to observations. We found that meanANPPandRUEfrom experiments did not deviate from observations. Experiments and observational data also yielded similar functional relationships betweenANPPandGSP, but only within the range of historically observedGSP. Fewer experiments imposed extreme levels ofGSP(outside the observed 33‐yr record), but when these were included, they altered theGSP–ANPPrelationship. This result underscores the need for more experiments imposing extreme precipitation levels to resolve how forecast changes in climate regimes will affect ecosystem function in the future.