Challenges and opportunities for Artificial Intelligence for predicting changes in ecological communities

Individual-based Ecology, vol. 2

Abstract

Ecology should be capable of convincingly demonstrating the consequences of changes in advance to become the strong predictive science that it needs to be. Predictive ecology could greatly assist conservation policy-makers by enabling them to understand and assess alternative management scenarios. Predicting changes in ecological communities require complex models and large empirical datasets. Deep Learning (DL), in combination with Organism-based Ecology (OE), could be used to construct models capable of predicting how organisms interact with each other and with their environment and how these interactions drive community change. At present, traditional ecological data are insufficient to train such models. A completely new type of data is needed, data that captures the information organisms use to guide their actions and responses to changes within their communities. To address this, we propose developing sensory instruments that simulate organisms’ sensory organs to collect these data. We also call for the explicit study and publication of the limits of applicability of DL models and we propose establishing a set of ethically responsible rules for applying DL models in ecology.

Authors 2

1. C.J.M. Musters Aachen Institute of Environmental Sciences (CML)

   Leiden University

   Affiliation as printed

   Leiden University, Institute of Environmental Sciences, Leiden, Netherlands

2. G.R. de Snoo Aachen Institute of Environmental Sciences (CML)

   Leiden University · Royal Netherlands Academy of Arts and Sciences

   Affiliation as printed

   Leiden University, Institute of Environmental Sciences, Leiden, Netherlands

   Royal Netherlands Academy of Arts and Sciences, Amsterdam, Netherlands