11 Combined Stresses in Forests 239
(Lloret et al. 2004 ). Bansal et al. ( 2013 ), as well as others (Jacquet et al. 2013 ), have
shown that recovery is hampered to a greater extent by the presence of a primary
stressor such as drought, rather than the addition of a secondary agent such as her-
bivores.
11.5 How Do We Predict Responses and Impacts from
Multiple Stressors?
As highlighted in the preceding sections, forest responses to stresses are complex
and extrapolation of experimental results to situations outside experimental condi-
tions may be inappropriate. Models can play an important role in quantifying the
impacts of stress under a range of conditions or scenarios. Some progress has been
achieved in simulating the impacts of single stress-related disturbances such as fire
and harvesting on net primary productivity in Canadian boreal forests (Li et al.
2003 ), spruce bark beetle outbreaks in Norway (Jönsson et al. 2012 ) and drought
impacts on biomass stocks in the Amazon (Rammig et al. 2010 ). However, even
single stresses are, in general, poorly represented in models, reflecting the difficulty
in representing complex, nonlinear responses to stress, a lack of mechanistic un-
derstanding of response processes and a lack of data for model validation (Jönsson
et al. 2012 ; Pinkard et al. 2011 ). The importance of incorporating physiological re-
sponses into modeling the impacts of biotic attack is demonstrated by a recent study
into carbon exchange in a conifer forest. The magnitude of change in ecosystem
carbon fluxes during spruce beetle infestation was influenced not only by the final
rates of mortality in the two dominant species but also by carbon losses incurred
during the period in which tree growth and gas exchange were declining (Frank
et al. 2014 ). Thus, plant physiological responses are foundational to the prediction
of broader ecological outcomes during stress.
The challenges in modeling stress are exacerbated when considering multiple
stress dynamics arising from interplay between primary, secondary, conditioning,
and anthropogenic factors. A comparison of the capacity of six models, ranging in
scale from tree to globe, to simulate drought mortality in pinyon pine-juniper wood-
lands, found that none of the tested models dealt well with multiple stress interac-
tions such as biotic agents and drought. The authors concluded that the models were
useful for defining key processes rather than quantifying impacts (McDowell et al.
2013 ). There is always a trade-off in models between the need to represent complex
systems appropriately and making the models so complex it is difficult to param-
eterize them. McDowell et al’s study illustrated that models of varying complexity
can be effective in determining the likely direction of change in a system. How-
ever, quantifying the level of change at the ecosystem-level will require modeling
frameworks that consider a hierarchy of stress responses and interactions distrib-
uted across the forest landscape. While there are few if any examples of models that
can achieve this, synthesizing data from a diversity of sources, such as controlled
experiments, environmental drivers (climatic, pest dynamics, soils), stand-level