Markov decision processes in non-autonomous socio-ecological systems

Our ability to effectively manage natural resources is founded in an understanding of how our actions and the environment influence populations, communities, and ecosystems. Current practices use monitoring data from the past to determine key ecological relationships and make predictions about the future with the assumption that those relationships will remain constant. However, many natural systems are undergoing rapid changes due to external factors including climate change, urbanization, and energy development, leading to a situation in which our observations of the past are poor predictors of the future. Ignoring such changes could lead to management decisions that are sub-optimal at best or detrimental at worst. This proposal is focused on establishing a theoretical framework for understanding the best management practices under global change, which will depend not only on the current state of the system but also on predictions of how that state will change over time. This work will draw from the fields of decision analysis, optimal control theory, population and community ecology, climate modeling, and natural resource management. Using decision problems related to the management of migratory birds in North America as motivating examples, we seek to develop generalizable strategies for optimal time-dependent natural resource management in a rapidly changing world.  

Principal Investigators:
Michael C. Runge (USGS - Patuxent Wildlife Research Center)
Patrick K. Devers (U.S. Fish and Wildlife Service)
James E. Lyons (USGS - Patuxent Wildlife Research Center)
Anna Tucker (fellow - USGS - Patuxent Wildlife Research Center))


A conceptual diagram showing the links between global change, non-stationary dynamics in natural systems, and time-dependent optimal strategies for natural resource management. In the top left panel, global change is depicted as a map showing difference in average temperature experienced in 2018 relative to 1981-2010 (NOAA NNVL, Data: NCEI). In the bottom left panel is a figure showing the percent change in a management-relevant parameter (e.g., carrying capacity or population growth rate) over time under both stationarity (constant) or non-stationarity (declining). In the right panel, theoretical time-dependent optimal policies are shown for both the stationary and non-stationary scenarios.