The Influence of Ambiguity on Optimal Forest Management: An Approach with Multi-Model Markov Decision Processes

In the context of climate change, it has become difficult for private forest owners to project themselves into the future and to optimize their decisions in order to meet the objectives they have set for the management of their forests. Indeed, climate change is not perfectly predictable, and the risks of forest fires, strongly linked to climate change, will also evolve. Forest owners must therefore act amid ambiguity. In addition, private forest owners are very heterogeneous in terms of objectives, attitudes, and their behavior in forest management, but also in their perception and defiance of ambiguity. It is therefore essential in any given framework to take this heterogeneity into account. In such an environment, providing recommendations to private forest owners that include their preferences regarding ambiguity is a fundamental societal challenge in the face of climate change. This study aims to provide some answers in this context of ambiguity to optimize forest owners' decisions and identify optimal policies. We define a maximization approach to obtain optimal forest management policies that account for ambiguity and ambiguity aversion, explicitly considering climate change and the various possible fire risks. We adopt an infinite-horizon, stationary Multi-Model Markov Decision Process (MMDP) framework to model this problem. The main contribution of this work is to design an MMDP model to evaluate forest management policies in an ambiguous context due to climate change, and to generate forest management policies evaluated under several parameters of the forest owner's ambiguity preferences, according to two decision criteria, the -MEU model or the smooth ambiguity model. This enables forest owners to develop insights into economic forest management under climate change. The MMDP framework is applied to a non-industrial private forest owner located in southwestern France facing a fire risk. The optimal management policy is unaffected by ambiguity aversion, whereas risk aversion reduces the optimal cutting age.