As the Earth’s climate continues to change, natural resource managers are faced with the challenge of forecasting how populations will respond to these changes. There are a variety of different forecasting approaches, ranging from complex process-based models that incorporate biological mechanisms to machine learning techniques that do not consider underlying biological processes. In a recent study published in the Proceedings of the National Academy of Sciences, a team of researchers that includes Ecology and Evolutionary Biology Assistant Professor Celia Symons has come up with a new approach to forecasting.
Their approach, called MTE-EDM, uses the metabolic theory of ecology (MTE) to constrain an equation-free machine learning approach called empirical dynamic modeling (EDM). MTE is a theory that posits that biological rates, such as metabolism and growth, scale with temperature. By rescaling time with temperature and modeling dynamics on a “metabolic time step,” the researchers were able to use the MTE to improve forecast accuracy in 18 of 19 populations. On average, the improved accuracy was 19%, with the largest gains observed in more seasonal environments.
This study highlights the potential for combining modern data-driven forecasting techniques with ecological theory and mechanistic understanding to predict the response of complex ecosystems to temperature variability and trends. The MTE-EDM methodology provides a new tool for forecasting abundances in environments with seasonal and/or interannual temperature change.
The study provides an important step toward improving our ability to forecast the response of ecological systems to environmental change, which is critical for sustainable management of natural resources.
Learn more about the research of Professor Symons here.