Spatiotemporal Data and Models in Ecology: Current Applications and Emerging Methods

Marine ecosystems are changing rapidly under the combined pressures of climate warming, deoxygenation, and human exploitation. Understanding how these pressures alter the growth, size structure, distribution, and interactions of marine organisms is a key challenge in ecology today.

A central tenet of my research program is that ecological processes unfold continuously across space and time, and that accounting for spatiotemporal structure and patterns is essential for understanding ecosystem dynamics in a changing ocean. The spatiotemporal data we use to understand these processes are rich in ecological information. However, they require adequate modelling approaches to account for the first law of geography, namely that locations close to one another tend to be more similar than those farther apart. Such spatial patterns can help us learn about the system we are studying, and ignoring this spatial dependence can invalidate statistical inference.

I start the presentation by briefly reviewing key methodological advances of the last decade, which have greatly facilitated the analysis of spatiotemporal data. Specifically, by combining ecological covariates with latent spatial fields in hierarchical models, we can both improve predictions and partition variance into process and observation error. These tools also allow us to address ecological questions relevant to management and climate assessment more efficiently, with spatial models of biomass density serving as a common currency.

I next illustrate this with examples from my work on Baltic cod, using it as a case study to show how spatiotemporal models can be applied to investigate climate-related changes in distribution, identify drivers of changes in body condition, and to quantify the strength of species interactions in marine food webs. These findings demonstrate that latent processes, often spatially structured, often explain considerable variability. This forces us to reconsider how confidently we can attribute ecological change to the drivers commonly assumed.

In the final part, I focus on more ongoing work and recent methodological advances. In particular, I highlight new statistical approaches for estimating non-local and time-lagged environmental effects, and multivariate spatiotemporal models, which can be used to explicitly link range shifts and trait changes within a unified framework, and to jointly model species range shifts across multiple dimensions. Expanding the use of spatiotemporal models is essential for improving predictions, propagating uncertainty, and to address the next generation of questions in climate change ecology.