The Problem of Pattern and Scale in Ecology - Summary

1. What did this paper do that made it a citation classic? It summarized a large body of work on spatial aspects of ecology placing it in a unifying conceptual framework based on scaling issues. It illustrated how spatial aspects provide a means to make connections between the previously rather divergent fields of evolutionary/population ecology and ecosystem ecology. It focused attention on multiscale methods to connect different levels of biological hierarchies. The Problem of Pattern and Scale in Ecology - Summary

2. Scale as a central problem in ecology What is scale? Geography considers this in 3 ways: (1) Cartographic scale - depicted size on a map relative to actual size; (2) Analysis scale - size of a unit at which some problem is analyzed (e.g. county, state); (3) Phenomenon scale - size at which some human or physical process occurs. In this paper, the emphasis is making connections between (2) and (3) - deciding on what the appropriate domain sizes (extent in ecology), resolution (grain size or level of detail) to consider. A key point is that there is no single appropriate scale and that part of the complexity of ecology is that it involves processes operating on multiple scales (e.g. extents and grain).

3. Dealing with multiple scales This includes dealing with how information is transferred between scales; how to aggregate and simplify processes operating within one scale so that this can be effectively transferred to another scale. An objective is to see how much detail can be ignored without giving results that are contradictory to observations on particular scales of interest. This is still not really a science - it is part of the art of modeling. This paper focused interest on how one might go about defining appropriate scales, how these relate to properties of ecological components (e.g. individuals, populations, watersheds, etc.), and how these might be constrained by observation.

6. Scale and uncertainty Levin uses dormancy/dispersal trade-offs in an evolutionary context for plants as an illustration of the use of larger regions, or longer time periods to “average out” uncertainties arising at finer scales - broader-scale statistical behaviors are more regular. In probability theory this arises from ergodic properties of the underlying stochastic processes (space- or time- averaged behavior arises by looking across sample paths or equivalently along a single trajectory). He calls this the principle scheme of scientific inquiry - moving from unpredictable individual cases to collections with regular behavior at which generalizations may be derived. Detail is lost, generality is gained through rescaling.

7. Pattern formation Pattern is essentially variation in some quantity, and is inextricably linked to scale since we need to define a scale in order to quantify the level of variation. Key questions in ecology and evolution arise from determining patterns, quantifying them and comparing them across systems. So there has been a lot of work on detecting and describing patterns. The next step is considering mechanisms for pattern development (activator/inhibitor systems, extrinsic forcing, behavioral responses driven by direct interactions), noting however that there may be many mechanisms capable of producing observed patterns (e.g. true in phyllotaxis).

9. Patterns of spread He argues that diffusion-based models have been highly successful at providing a basis for understanding dispersal, invasion, patchiness, pest spread, etc. as they have in pop genetics for analyzing interactions of selection, mutation and drift on population-scale characteristics arising from individual reproduction. In all cases the objective is to understand macro-level properties through the actions of components operating at finer scales. This is arguably the raison d’etre for individual-based models and other approaches that are not diffusion-based.

10. Levels of detail and scale Analogous to determining the properties of populations from considering individuals there is a need to construct ecosystem properties from that of component populations (or patches). This takes a statistical mechanics view of local variation, disturbance and colonization to produce a landscape-level view of system dynamics. The patch dynamics situation is an example of two-time scale methods: demography of patches and within-patch rapid succession. This is a powerful method of moving across scales, but sometimes patchiness can be found at all scales in which case the goal is to understand how the system description changes across scales (windowing) – which some view as the hope for fractal scaling laws.

12. Food webs - similar to the spatial scaling issues, describing food web properties depends upon the level of aggregation used (a scale of observation) - how do the topological regularities depend upon the taxonomic or trophic aggregation used? This led to a huge amount of work on food web structure. Global change impacts through integrated GCMs and ecological models are only recently moving as far along as this paper suggested they should and the move towards community-level analysis of evolutionary processes discussed in the paper might be thought of as a lead-in to the current work on community genomics. But we are still far from the convergence of evolutionary theory and ecosystem processes this paper suggests.