Describing a recent drought-related forest dieback event in south-western Australia in 2011 (Matusick 2012), a colleague remarked upon the distinctive sounds of wood boring beetles feeding on weakened eucalypt trees during one of the most severe droughts on record (K. Ruthrof, personal communication). For this insect population, normally scarce and benign, drought stress had most likely triggered a surge in their abundance, thereby amplifying declines in forest health during an already stressful event. Observations of an apparent coincidence of stressors such as drought and pests are of course common across many ecosystems (Mattson and Haack 1987, Ayres and Lombardero 2000). The subsequent impacts on forest function and structure from stressors such as drought and herbivory represent complex interactions between abiotic and biotic factors (Raffa et al. 2008).
During its lifetime, a tree faces a diverse range of conditions arising from the combination of climatic, soil and pest dynamics and disturbances that may induce periods of stress, culminating in reduced growth, vigour, canopy dieback or mortality. Based on the sequence, timing and intensity of these stressors, their associated impacts may be persistent or acute and manifest over different temporal and spatial scales. For example, recent observations of episodic mortality from acute drought stress across the USA, Europe and Australia highlight the role of single extreme events (Martínez-Vilalta and Piñol 2002, Breshears et al. 2005, Matusick 2012), while longer term forest declines result from persistent and subtle stress interactions that make identifying specific causes more difficult (Manion 1981, Close and Davidson 2004). It is increasingly suggested that forests in the future will experience increased levels of stress as a result of the direct impacts of shifting climatic patterns, increasing atmospheric [CO2], land use change and associated feedbacks on processes such as pest demographics, invasive species and host susceptibility (Intergovernmental Panel on Climate Change 2007). Any assessment of future forest vulnerability must consider the differential contributions from single and multiple stress factors and how their timing, intensity and frequency may interact to affect individuals, stands and ecosystems.
Climate is arguably the most influential factor affecting vulnerability in managed and unmanaged forests due to its role as a primary or underlying stress, that precedes or enables the arrival and expansion of associated biotic stressors (Figure 1a) (Kurz et al. 2008). Drought itself is not a proximal stress but the cause of multiple stresses that impact on trees (Figure 1a) and is often characterized by multiple stressors: i.e., large water deficits and high temperatures/heat waves (Allen et al. 2010). In turn, these can interact with carbon and water dynamics differentially (McDowell et al. 2008). In many cases the primary stress such as drought promotes the presence of secondary stressors such as insects or pathogens via a weakening of plant defences (Boyer 1995), changes in pest distributions and disruption of food webs that would otherwise be kept in check (Figure 1a) (Carnicer et al. 2011).
The paper by Bansal et al. (2013) in this issue directly addresses the role of singular and combined impacts of drought and herbivory on plant functioning. The paper demonstrates that the impacts of multiple stressors on leaf gas exchange and growth traits are not always equal to the sum of the parts and are instead ‘antagonistic’ (less than the sum of the impacts from the singular stressor). Similarly, the interactive effects of elevated [CO2] and temperature on herbivory are generally opposing; increasing [CO2] diminishes leaf quality for herbivores in host plant species independent of temperature, while higher temperature may enhance herbivore activity through changes in metabolism and partially compensate for these reductions in plant quality (Zvereva and Kozlov 2006, Murray et al. 2012). These types of emergent responses show that secondary stressors such as bark herbivory can act to both dampen and amplify the effects of a primary stressor, such as drought, presumably via changes in carbon balance and feedbacks on growth and photosynthesis (Pinkard et al. 2011).
The intensity of single stress events such as drought has important consequences for the impacts of subsequent stressors (Jactel et al. 2012). As drought stress progresses, cell expansion and growth are often the first casualties of water deficit, followed by a cessation of photosynthesis (Hsiao et al. 1976), and the eventual breakdown of water and sugar transport (Hölttä et al. 2009). The addition of a secondary stressor for a given drought intensity can act to preserve or further disrupt these processes. The study by Bansal et al. (2013) considered a range of both drought and herbivory intensities and found that the impacts of multiple stressors were dependent on the stress intensity of one or both of the stressors. For example, the combined impacts of drought and herbivory at moderate intensities reduced radial growth to a larger extent than when either of these stressors were severe (Bansal et al. 2013). Figure 1b provides a hypothetical representation of why predicting the impacts from multiple stressors is notoriously complex and difficult. Primary and secondary stressors such as drought and herbivory presented in this example have differing levels of impacts individually at any given intensity. However, their interactive impact is not always additive, as demonstrated by Bansal et al. (2013), and may be a function of intensity of one or either of the stressors (Figure 1b). The importance of intensity of one or both stressors underscores the role of threshold dependence on emergent relationships between the stress and physiological responses across the whole plant. The response of photosynthesis to combined effects of elevated [CO2] and drought stress is a clear example of this; at high soil water contents the enhancement of photosynthesis from high [CO2] is realized, yet as soil water declines below a threshold value the impacts of drought on stomatal closure overwhelm [CO2]-mediated increases in photosynthesis (Centritto et al. 1999). The temporal dynamics of stressors are also critical to assessing stress interactions particularly where the primary stress is characterized by low to moderate intensities and longer durations or higher frequencies. In the case of drought, the duration of the event can promote carbon depletion through prolonged periods of low to zero assimilation (Mitchell et al. 2013), while enhancing the period of exposure to biotic agents that can further reduce vigour and the ability to recover (Galiano et al. 2011).
Evaluating recovery often provides a better gauge of the combined severity of the stress than observations made during the stress event. A seemingly severe drought event for a species capable of resprouting may merely manifest as a transient loss of above-ground tissues that can be rapidly regrown and represent an adaptive strategy that helps a species avoid dehydration. Broadly speaking, both symptomatology and interpretation of event severity must be considered from a ‘plants-eye’ view in the context of the species adaptive life strategies. Tracking the trajectory of physiological responses such as growth beyond the climatic event may also reveal the extent to which the stress was transient, delayed or sustained and elucidate the contributions of resistance and resilience-based strategies for plant through to ecosystem responses (Mitchell et al. 2013). Bansal et al. (2013) as well as others (Jacquet et al. 2013) have shown that recovery is hampered by high-intensity drought in the previous year and not herbivory. Patterns in stress response and recovery reflect the history of stressors on growth, carbon and water status and instil some level of ‘physiological memory’ that may affect the severity of future stressors on survivorship (Loehle and LeBlanc 1996; Niinemets 2010). The state and conditioning of the system at the time of an event impacts on the event severity from a plant or community perspective—the event severity or physiological stress is not independent of the system state (Figure 1a). For example, an analysis of radial growth patterns in pinyon-pine stands after severe drought showed that regardless of site condition, patterns of tree mortality were associated with reduced and more variable ring widths in the preceding 10–15-year period (Ogle et al. 2000). The mechanisms of recovery can impose serious metabolic costs that delay a return to pre-stress levels as demonstrated by the impact of cavitation on water transport during severe drought. Brodribb et al. (2010) found that recovery of gas exchange via the restoration of hydraulic conductance tracks the growth of new xylem tissues, suggesting that recovery imposes significant carbon costs after drought.
Stress impacts at the stand level are rarely uniform and emphasize the view that the observed responses of individual trees are a function of conditioning factors such as stress history and tolerance, ontogeny, competition and acclimation (Figure 1a). In the real world, no two events are identical—the preconditions and the progression of the stress event are always unique. Capturing some of the important drivers of stand-level variation involves a shift from deterministic thinking that invokes one-to-one causality to a more probabilistic way of thinking. Within a probabilistic context, processes such as forest stand mortality are viewed as a manifestation of the individual responses and performance of the constituent trees. Modelling frameworks appropriate for this task will need to consider a hierarchy of multiple stressors and interactions with their environmental drivers and distribute them within the forest landscape using probabilistic approaches that can provide an assessment of future risk and uncertainty in key forest processes. Hierarchical Bayesian modelling approaches are an ideal candidate in that they facilitate the synthesis of diverse data sources such as controlled experiments, environmental drivers (climatic, pest dynamics, soils), stand-level response data and mechanistic/process-based models (Ogle and Barber 2008, Cable et al. 2009, Metcalf et al. 2009).
Studies such as Bansal et al. (2013) elucidate the direction and severity of multiple stress interactions that would not be predicted by considering the singular stress impact in isolation. Extrapolating this growing body of experimental studies on multiple environmental stressors into the real world is one of the key challenges in forest and ecosystem science. Within the framework that we have proposed, multiple stress interactions need to be considered across a range of intensities, durations (that include the recovery period) and frequencies within the state and conditioning of the system in which they are occurring. Obtaining such data is unlikely and we emphasize the role of merging experimental data, field observational data and process models within an appropriate hierarchical and probabilistic framework. A suitable approach will be when one captures the impacts of stressors across heterogeneous stands that reflect underlying responses of a collection of individuals and provide a means of propagating parameter uncertainty across scales from trees through stands to landscapes.