Climate scenarios for the mid- to late twenty-first century project not only continued global warming but also an increased frequency and intensity of extreme climatic events (ECEs), which could strongly affect biogeochemical feedbacks to the climate system (IPCC, 2012; Reichstein et al., 2013). The effect strength of an ECE depends on both its intensity and its duration. Strong events can lead to recoverable changes in ecosystem functioning, or, in some cases, to new ecosystem states (i.e. irreversible or long-lasting changes, Fig. 1; Smith, 2011b). Potentially irreversible effects include, for example, high rates of plant mortality or soil loss, and associated functional shifts in the ecosystem. Large mortality events can release considerable quantities of carbon over decades, and often increase the susceptibility of a system to subsequent events and accelerate shifts in vegetation composition (Allen et al., 2010; Reichstein et al., 2013).
Recent studies have advanced our understanding of ecosystem responses to climate extremes (Reichstein et al., 2013; Smith, 2013), but much remains unknown about how ecosystems will respond to a climate in which extremes will be much more pronounced than today. It has been suggested that, by the end of the century, events that are considered extreme today will be in the range of typical interannual climate variability; concurrently, the severity and frequency of events of a given return interval will increase (Fig. 1a; Smith, 2011a; IPCC, 2012; Williams et al., 2013). This may not only result in more frequent transient declines in ecosystem functions, such as photosynthetic carbon uptake or the turnover of soil organic matter, as a result of crossing recoverable response thresholds, but may also cause long-lasting changes in ecosystem functioning as irreversible thresholds are exceeded because of widespread mortality events, changes in the age structure of populations and alterations of the species composition of communities (Smith, 2011b; Reichstein et al., 2013; Fig. 1b). For a given ecosystem, we cannot yet predict the level of climate severity at which thresholds for changes in ecosystem functioning occur. Our understanding is hindered in part by a lack of common metrics making experiments comparable (Vicca et al., 2012), and also by the tendency for experiments to explore likely climate scenarios rather than ecosystem response functions and their critical thresholds (as depicted in Fig. 1b). When these thresholds are passed, not only will biogeochemical process rates respond, but also ecosystem resistance and resilience to disturbances such as plant stress-associated pest outbreaks (McDowell et al., 2011), biological invasions (Diez et al., 2012) and subsequent ECEs (Fig. 1b). Critical transitions in ecosystem states triggered by ECEs may involve an immediate or a lagged degradation of an ecosystem, the latter occurring progressively even when environmental conditions have become less severe (Fig. 1b). However, selection and adaptation of species in response to ECEs can favor long-term ecosystem recovery and modify thresholds that result in impacts on ecosystem functioning, for example by altering rooting depth, flammability, water use efficiency, defense mechanisms and regeneration behavior (Gutschick & BassiriRad, 2003).
We currently lack an understanding of how the climate thresholds of ecosystem functions are modified by recurrent ECEs overlaid upon the ongoing chronic changes in climate. Shifts in ecosystem functioning caused by ECEs or plant acclimation will likely also cause shifts in thresholds (Fig. 1b), although the direction and magnitude of these shifts are largely unknown. Recurrence of stressful conditions has the potential to decrease or increase resistance and resilience of plant and ecosystem functioning (Larcher, 2003; Walter et al., 2013). Gradual climate change can alter ecosystem responses to climate extremes by modifying for example temperature response functions, stoichiometric relationships, and water use. For example, elevated CO2 can reduce water loss via transpiration, which could buffer the effects of extreme droughts on soil water availability and plant growth (e.g. Morgan et al., 2011). However, drought-induced water stress typically reduces stomatal conductance, negating the benefits of elevated CO2 (Franks et al., 2013).
Land-cover changes and shifts in plant functional types, for example, caused by conversions from forests to pastures or by shrub encroachment, may buffer or amplify effects of climate changes with increasing severity of ECEs, and may thereby induce very different response trajectories. For example, grasslands follow a less conservative water use strategy than forests, and have access to less ground water; grasslands are thus more rapidly affected by heat waves (Teuling et al., 2010), while the speed and degree of their recovery from such an ECE could be higher. However, this has not yet been studied in a generalized manner, and meta-analyses are constrained by the dependence of ecosystem responses to climate extremes on climate context and on the timing of extreme events relative to species phenology (Knapp et al., 2008).
Experiments have provided powerful tests of the effects of climate variation and extremes on ecosystem structure and functioning, but have rarely tested for critical thresholds (Smith, 2011b; Beier et al., 2012). A future generation of experiments should take simulated ECEs further to identify both thresholds and trajectories of ecosystem response and recovery, and to elucidate the underlying mechanisms. Such mechanisms are potentially related to species physiology (stress response, including repair mechanisms), species composition (biota and functional groups) and interactions, as well as physico-chemical changes in the soil. Experiments should explore how recurrent ECEs and interactions with other climatic changes and land-use changes modify threshold responses of ecosystem functioning.
In addition to experiments, observational tools should be improved, including the application of remote sensing for detecting long-term trends in ECEs on larger spatial scales (McDowell et al., 2011; Reichstein et al., 2013; Zscheischler et al., 2013). Long-term monitoring programs (e.g. Fluxnet, NEON, ICOS, LTER) are not only essential for ground-truthing and calibrating remotely sensed information, but can provide important insights on baseline ecosystem processes before the occurrence of an ECE and during recovery. Mobile laboratories have the potential for a rapid in-depth analysis of ecosystem recovery responses after ECEs, and could be used for comparative observation of recently disturbed vs undisturbed ecosystems. Searching for analogue conditions in paleo-records is another valuable yet challenging strategy (Williams et al., 2013) that, when compared to process models, can provide confidence in both our understanding of process and our predictions of future conditions.
Extreme climatic events are becoming an increasingly important driver of changes to the Earth's terrestrial systems, with implications for humans and the biosphere. By altering the carbon cycle they have substantial potential to feedback to the climate system, potentially accelerating climate change (Reichstein et al., 2013). The current generation of Earth system models (ESMs) cannot simulate adequately the impacts of climate extremes on biogeochemical cycles and related climate feedbacks. There are many areas in which ESMs could be considerably improved with respect to climate–biosphere interactions in a more extreme world; we highlight five:
- Current ESMs operate with non-adapting ‘average’ individuals, which makes them prone to be oversensitive to ECEs. An implementation of trait- and individual-based dynamic vegetation models (e.g. Scheiter et al., 2013) in ESMs would permit accounting for biological adaptation, plasticity and diversity and their potential to shift thresholds and to dampen ecosystem responses to ECEs.
- Future ESMs should better represent climate-dependent demographic processes. They should thereby not only account for mortality and establishment patterns, but also consider that plant sensitivity to ECEs and respective thresholds are often age-dependent (Niinemets, 2010). The resulting long-term effects on forest structure and carbon sequestration potential (Bond-Lamberty et al., 2014) need to be captured by ESMs.
- Dynamic vegetation change due to drought, fire, wind, or other disturbances are critical components of ECEs, but not all ESMs incorporate it (Arora et al., 2013) and most have not been tested adequately (McDowell et al., 2013; Powell et al., 2013). Furthermore, legacy effects of ECEs on the susceptibility of ecosystems to pests or subsequent ECEs should be included in future ESMs.
- ESMs should implement more explicitly ecosystem-internal feedbacks, such as two-way carbon–water cycle interactions, by which changes in soil carbon or erosion may alter soil hydrology through effects on water holding capacity. Such effects may have strong implications for the long-term response of ecosystems (Heimann & Reichstein, 2008) and therefore need to be better represented in ESMs.
- Finally, the capacity to simulate climate extremes needs to be improved in the climate model components. Higher-resolution modelling and improved biophysical coupling between the atmosphere and the biosphere will improve the representation of these events (Fischer et al., 2007).
To obtain realistic boundaries to our expectations for future terrestrial function in a more extreme climate we clearly need to improve our understanding of the patterns and mechanisms of threshold responses of ecosystems (Smith, 2011b). If researchers can identify consistent patterns of biogeochemical responses as ecosystems are pushed to thresholds for shifts in structure and function, these patterns can be used to detect and potentially respond to early warning signs of critical transitions (Lenton, 2011). In this letter, we have highlighted that future research efforts should address emerging questions such as: (1) identification of the severity of climate conditions that may lead to recoverable or irreversible/long-lasting changes in ecosystem processes, and how these different thresholds may be affected by future climate change and interactions with other global change drivers; (2) legacy effects of ECEs on the resistance and resilience of plant and ecosystem functioning to disturbances, including subsequent ECEs; (3) effects of acclimation responses of species and functional diversity, and causes and consequences of changes in demography caused by ECEs; (4) interactions between ECEs and land-cover changes and shifts in plant functional types; and (5) improvement of ESMs to explicitly account for the discussed effects.