Esra Büyüktahtakιn Toy, PhD, Virginia Tech

Abstract

Biological systems have highly uncertain behaviors to predict. The inherent uncertainty in species’ population dynamics and combinatorial resource allocation decisions often result in multi-stage stochastic optimization problems. In this talk, we will present an innovative data-driven multi-stage stochastic mixed-integer optimization modeling and cutting-plane approach to tackle a large-scale optimization problem under decision-dependent uncertainty. We demonstrate our model and the algorithm on one of the most pressing problems of the USDA Forest Service, the Emerald Ash Borer (EAB) infestation killing millions of ash trees in North America. We validate our operations research approach using 7-years of unique ash health data collected by our collaborators from the USDA Forest Service over multiple spatial locations in Ohio. Computational results show that our cutting plane-based method can substantially reduce the solution time for this forest insect infestation problem with binary and continuous resource allocation decisions. Our findings provide critical insight into a long-debated question among foresters: treatment versus removal of the ash trees to save as many trees as possible. This study is a part of the ongoing joint work with collaborators from the U.S. Forest Service (Robert Haight, PhD; Kathleen Knight, PhD; and Charlie Flower, PhD) and former doctoral students (Eyyub Kibis, PhD; Sabah Bushaj, PhD; and Chen Chen, PhD) and a faculty collaborator from NJIT, Wenbo Cai, PhD.