Trends in ecosystem recovery from drought

An analysis suggests that the time taken for ecosystems to recover from drought increased during the twentieth century. If the frequency of drought events rises, some ecosystems might never have the chance to fully recover. See Letter p.202

On page 202, Schwalm et al.¹ assess historical trends in the recovery of ecosystems from drought, using three state-of-the-art data sets of gross primary productivity (GPP) — the amount of atmospheric carbon dioxide that is fixed in ecosystems by photosynthesis. Intriguingly, the analysis suggests that the total area of ecosystems that are in a post-drought recovery state increased during the twentieth century. Moreover, the authors identify areas where the drought-recovery time is particularly long, and which might, therefore, be particularly vulnerable to more-frequent droughts.

Changes in the recurrence of climate events, and the associated times taken for affected ecosystems to recover, are rarely considered in evaluations of the global impact of climate change. For instance, the most recent assessments of the Intergovernmental Panel on Climate Change (IPCC) generally focus on mean climate changes, providing only a few analyses of the frequency with which extreme events occur^2,3, and little or no information on the statistics of time lapses between recurring extreme events. Climate scientists increasingly consider the timescale of recovery processes to be an integral part of impact assessments, but such considerations have generally been confined to local-scale experiments, regional assessments, literature reviews and theoretical studies^4,5,6,7.

Schwalm et al. now offer an interesting global and historical perspective on this issue by considering how photosynthesis recovers from drought across different ecosystems, using data sets that are based at least partly on observations. They report that the total ecosystem area affected by drought recovery increased during the twentieth century, and that recovery times have also tended to increase. This finding might have major implications under increasing global warming: if an ecosystem has not fully recovered from a drought, the effects of a subsequent drought might increase in a nonlinear, unpredictable way (Fig. 1).

Figure 1: Effects of changes to recurrent environmental stressors.

a, A recurring environmental stressor such as drought will have recurring effects on ecosystems, characterized by an impact phase and a recovery phase. b, If the stressor pattern remains the same (red line) but the time taken for the ecosystem to recover to pre-drought conditions increases, or if the time period over which the stressor occurs increases (yellow line), then the impact phase of a new event might overlap with the longer recovery phase of a past event, potentially leading to unknown impact levels. Schwalm et al.¹ report that the time taken for various ecosystems to recover from drought increased during the twentieth century.

The authors also note that regions at high northern latitudes and in the tropics have particularly long recovery times — highlighting these areas as being especially vulnerable to any increases in the recurrence of drought. This might have important ramifications for the carbon cycle, given that tropical regions in particular act as a substantial land carbon sink.

One should nonetheless note some limitations of the study, which may require the results to be interpreted cautiously. First, the main conclusions are based on GPP data sets generated from land-surface models⁸. The simulations produced by these models are constrained by observational data regarding precipitation, temperature and the amount of sunlight available for GPP in given time periods and regions. However, the simulated carbon fixation by GPP is not constrained by observations — at least, not in a way that goes beyond the basic, standard procedures generally used for model validation. Furthermore, the land-surface models do not explicitly represent many of the dynamic processes relevant to drought impacts and recovery^6,7, such as plant mortality caused by hydraulic failure (a drought-induced physiological problem that affects water uptake); changes in vegetation that occur after fires; and pest outbreaks. The results are therefore likely to mainly reflect changes in climate features, rather than the full response from vegetation.

Schwalm et al. validated the analysed simulations by comparing them with two other GPP estimates^9,10, but both of these are also based only partly on observations and depend strongly on model-like assumptions (that is, on the algorithms developed to derive the data sets). Most notably, the comparator data sets were calibrated using carbon-flux data, which are scarce, and therefore include estimates for some areas, such as tropical regions, that rely on only a handful of observations^9,10. In addition, the algorithms underlying the development of these data sets are relatively simple, and do not, for instance, account for the effects of increased atmospheric CO₂ levels on photosynthesis and GPP¹¹, nor explicitly represent the effects of soil-moisture limitation on CO₂ assimilation by ecosystems^6,12. It would therefore be helpful to include alternative estimates of GPP in future analyses.

Another limitation is that the authors use a single metric¹³ to quantify and characterize droughts. But drought is difficult to assess properly^3,14,15 — particularly when assessing changes relevant to ecosystems, which relate more to changes in soil-moisture storage¹² than to changes in precipitation. The metric used by Schwalm et al. partly accounts for evapotranspiration (the sum of evaporation and plant transpiration from Earth's surface, a factor that affects soil-moisture storage), in addition to precipitation deficits. But a lack of global-scale measurements means that the effects of evapotranspiration factored into the metric are based mainly on estimates of a quantity known as potential evaporation¹³, which is the maximum possible evaporation and thus tends to overestimate drought¹⁶ when used instead of evapotranspiration. Moreover, potential evaporation is highly sensitive to air temperature. But under strong drought conditions, air temperature is increased by soil-moisture limitation (which reduces evaporative cooling¹²), and doesn't lead to further drying because plants reduce evapotranspiration during drought. Using a different drought metric in the analysis could produce smaller trends in drought recovery. It would thus be beneficial to expand the authors' analysis to include other data sets that describe drought.

Limitations aside, Schwalm and colleagues' study is highly valuable because it points to an under-appreciated dimension of drought impacts: the timescale of recovery and its relationship to the occurrence of drought events. The work offers crucial perspectives that might help in the development of new indices¹⁷ of extreme climate that are more directly relevant to ecosystem impacts than existing metrics. Given that current models of the Earth system and climate do not simulate the complex processes required to properly address ecosystem drought recovery⁵, projected changes in land carbon uptake may be biased, potentially affecting climate-change projections. Schwalm and co-workers bring attention to that issue, thus allowing better plans to be made to adapt to or mitigate the effects of climate change.¹