Graduate Student University of California Santa Monica, California
Insect pests pose a major threat to humans by jeopardizing food security in agricultural systems, acting as vectors for infectious diseases, and damaging forests and other ecosystems. Despite decades of research aimed at controlling pest populations to mitigate their harmful impacts, effective pest management remains challenging in many systems. This stems, in part, from incomplete knowledge of the mechanisms that drive population dynamics, making it difficult to develop accurate models that predict insect outbreaks. Due to the challenges of mechanistic modeling and historical tendencies of pest managers, many theoretical developments in this space have yet to meet practice. Pest management is often reactive in practice, meaning control actions are taken once outbreaks have already begun, allowing for damage to occur. It is possible to improve pest management, however, by acting in anticipation of an outbreak. We show that a data-driven model can effectively predict outbreaks, thereby, circumventing the need to understand the underlying network driving population dynamics. This allows us to target pests before outbreaks occur. We also show that optimal control can be used with our data-driven model to optimize pest management strategies. In particular, we explore the use of empirical dynamic modeling and Gaussian process regression paired with stochastic dynamic programming to keep insect populations within acceptable bounds of tolerance. We show that this framework effectively prevents outbreaks on simulated and empirical data in a variety of scenarios. Our work provides a management framework that has potential to reduce losses from pests.