Extrapolation in ecology and the problem of spatiotemporal variability

Description

In current philosophy of science, extrapolation is conceptualized as an inference from a study- to a distinct target population of interest, and the reliability of such an inference is thought to depend on the extent to which study and target are similar in relevant respects. In this talk, I will argue that this understanding needs to be extended when dealing with extrapolation in ecological systems. My argument rests on two premises: First, extrapolation in ecology is only sometimes adequately characterized as an inference from a model to a distinct target because it often includes inferences from a small-scale, local section of an ecosystem to the large-scale system at the regional, landscape, or even biome level. Ecologists extrapolate across spatiotemporal scales when independent variables of interest cannot be
directly manipulated at the extensive spatial and temporal scales at which their effect needs to be understood.
Second, extrapolation across scales adds another dimension of heterogeneity of study and target systems because ecological systems have frequent causal discontinuities. Factors that drive processes at one scale may be irrelevant at another scale. For example, interspecific competition is an important factor driving species distributions at small scales. However, it is nearly irrelevant at biome or continental levels where climatic factors are the primary driver. This implies that extrapolating ecological relationships across spatiotemporal scales can be erroneous even when study and target system appear similar in terms of causal factors and covariates. While extrapolation in philosophy has so far mainly been understood as an inference from a model to a different target, without explicit consideration of scale, the case of ecology alerts us to the fact that considering differences in scale between study and target system can be an important factor in justifying extrapolative inferences.
For this reason, I introduce the concept of spatiotemporal variability in this talk and contrast it with compositional variability. Compositional variability describes our current understanding of heterogeneity in the philosophical literature on extrapolation, which focuses on differences in the distributions of causal factors and covariates or qualitative features of mechanisms. It explains our difficulties when extrapolating from a model to a distinct target at roughly similar scales. However, when extrapolating across scales, the relevant heterogeneity to consider is not only
about the internal composition of study and target but also about the variability of a system over spatial and temporal scales. Spatiotemporal variability thus describes how heterogeneity manifests across spatial or temporal extent and explains the problems of extrapolating causal relationships from small to large-scale systems.
In this talk, I will introduce and further explain the concept of spatiotemporal variability and its implications for extrapolation in ecology. Throughout, I will use examples from biodiversity-ecosystem function research to illustrate my arguments. I will also show that clarifying the problem of extrapolation in ecology can broaden the philosophical debate and advance our general understanding of the multifaceted problems that arise when transferring knowledge across different systems of interest.

Period	25 Mar 2025
Event title	International conference of the German Society for Philosophy of Science (Gesellschaft für Wissenschaftsphilosophie – GWP)
Event type	Conference
Location	Erlangen, Germany, BavariaShow on map
Degree of Recognition	International