Blog‎ > ‎

Challenges to Forest Landscape Modeling Under Climate Change: Model Formulation

posted Jun 24, 2013, 1:04 PM by

Determining the appropriate level of resolution for a model is – in theory – given by the objectives of the study (see my earlier post below). However, what exactly to include or exclude in a model, and how to formulate a specific process (e.g., tree growth), is at the discretion of the modeler. For example, for tree growth options range from simple differential equations describing biomass development of the whole tree (or forest stand) to a detailed description of photosynthesis, respiration, and carbon allocation. While it is clear that we need to adapt our conceptual models of forest change to accommodate climate change, there are very few a priori rules at hand (dos and don'ts) in the context of global change.

However, it is widely acknowledged that we should try to move from a description of phenomena (e.g., the realized niche of a species) to more mechanistic approaches (e.g., how climate affects tree demography) that more robustly represent how forests will respond to climate change. Statistical approaches typically are based on regression (of many different flavors) that correlate current phenomena (e.g., species distributions) to current climate.  These correlations can subsequently be mapped onto future projections of climate change and incorporated into a model (e.g., regeneration can follow a parabolic envelope of climatic tolerance).  Such statistical approaches are fraught with assumptions that become less robust under climate change.  In particular, we are entering an era of “no-analog” forest ecosystems (Williams et al. 2007) that will significantly alter competitive dynamics and will therefore reduce the accuracy of statistical relationships derived from current forests. Ultimately the question for modelers is:  Which representation of a process provides that best compromise between current day accuracy (as demonstrated via validation) and the mechanistic flexibility to accommodate climate effects?  

Regardless of our best intentions, there is no landscape model that is built entirely from "mechanistic" relationships – we do not have the requisite knowledge or computational power to simulate any ecosystem based on first-principle relationships.  All models incorporate statistical relationships at some level. The key question is whether it is reasonable to assume that these relationships will retain their accuracy under a changing climate or whether they contain hidden assumptions about a constant climate.  Unfortunately, there are no rules of thumb and each relationship must be individually reexamined.  

For example, the relationships between climate and ecosystems may be ‘lagged’ or non-equilibrium.  Many models assume that the opposite - that climate and ecosystem are at equilibrium.  Because trees have long life-spans, there are long lag times associated with forest growth and mortality.  In a nutshell, the forests we see today do not necessarily represent the climate of today.  They represent the climate when they were established.  Thus, making statistical assumptions derived from current relationships may either under- or overestimate the sensitivity of forest properties to climate change.

A shift away from statistical relationships and towards a mechanistic approach that acknowledges shifting competitive dynamics will have large consequences for our understanding of succession and silviculture.  New species associations and new competitive dynamics (including the effects of introduced species), in combination with novel disturbance regimes, may reduce our ability to direct succession (aka silviculture).  How do we plan for the future given the magnitude of uncertainty about the future?  How do we develop more flexible and nimble strategies for managing forests?  Our lab is actively working on these issues in collaboration with forest managers.  As part of this effort, forest models need to quickly evolve in order to make a substantial contribution towards keeping silviculture relevant.

Finally, we must consider process interactions and how they will change in the future.  Process interactions will be particularly sensitive to the breadth of scales at which different processes operate.  If we are capturing only one or two scales of process interactions - fine scale interactions (e.g., neighboring tree interactions) or meso-scale processes (disturbance regimes) or very broad scale climate effects (shifting biomes) - we will miss a majority of critical process interactions.  For example, if we were to focus only on broad scale biome shifts (migration) and fine scale neighborhood interactions, we would miss changes in disturbance severity that may ultimately determine species composition.  Likewise, when we focus on meso-scale processes and larger, it is critical that we adequately capture sub-pixel heterogeneity or we may underestimate the ability of rare species to colonize locally.

- Robert Scheller

Williams, J. W., S. T. Jackson, and J. E. Kutzbach. 2007. Projected distributions of novel and disappearing climates by 2100 AD. Proceedings of the National Academy of Sciences 104:5738-5742.