Future Forest: A Decision Support System for Smart and Sustainable Forest Management.

Abstract

Future Forest is an “AI Lighthouse” project funded by the German Ministry of the Environment that has two main objectives: develop a decision support system for forest management and build the foundations for a forest transformation data space. The Future Forest decision support system is based on a chain of AI/numerical models. The information used to analyse the best alternatives in an area of interest comes from state-of-the-art process-based forest simulations of specific forest management scenarios, AI-based upscaling techniques, and remotely sensed data on current forest composition and health. This data will cover Germany’s forests wall-to-wall with an unprecedented resolution of 100m for the management scenarios and climate data, and up to 10m for other variables. Creating such a system is impossible without having an accessible pool of data. Since much of the needed information is not freely available, data is collected and organized as an IDSA-compliant data space. Such a data space serves as a platform where various data holders and users converge, exchanging information and analytical applications within a structured data governance framework. This arrangement empowers platform users to retain comprehensive control over their data and enables them to share information with third parties in a controlled and secure environment.   Future Forest is one year away from completion, and we can now present the first results on our way towards a forest management 2.0 system. This system is designed to offer a spectrum of alternatives for effectively managing local forest stands in response to climate change. Considering the forest owner's management objectives, such as timber production or biodiversity, the system proposes alternatives using various ecosystem indicators, encompassing wood production, carbon storage, and biodiversity considerations. The final ranking of the alternatives is based on a multi-criteria decision analysis algorithm, which incorporates also a comprehensive robustness and sensitivity analysis. In this contribution, we outline the tools utilized to make informed decisions, from the neuronal networks for forest classification to the forest dynamic simulations, and the decision support system. We discuss the constraints encountered and highlight the innovations incorporated in each of these tools. We will discuss the attempt made to offer an explainable or even interpretable model, as far as this was possible.