Taking action for biodiversity when multiple dimensions of forests are measured at once

Scientific monitoring of biodiversity changes requires simultaneously measuring many dimensions of forests, but aggregating over indicators to assess overall condition is a persistent challenge in practice. For example, which monitored changes should constitute “degradation” under the EU’s Regulation on Deforestation-Free Products? There are three general options: (1) choose one indicator to measure, such as surface reflectance (giving incomplete information and obscuring trade-offs), (2) report on many indicators at once, as in the Montréal Process (leaving users to assess overall condition, which is frequently contested), or (3) combine measurements of many indicators in a pre-defined way. Optimal aggregation methods to combine measurements have received attention in ecology, economics, and operations research. We assess common elements of leading solutions across these different fields, focusing on aggregation methods, assumptions, data requirements, and weaknesses in the specific case of monitoring changes in forest condition. Our results highlight weight selection problems when indicators lack some common basis (for example, when both habitat composition and species abundance matter for managing biodiversity, rather than extinction risk alone). We then propose a topological approach, based on estimating the state space of forests using the Natural Range of Variation (NRV) concept. Our conceptual approach treats each observation of forest condition as an n-dimensional vector; the resulting point cloud defines the NRV, and various methods of characterizing this cloud yield distance and volume measures that describe changes in overall condition, and planes which describe trade-offs. We operationalize our approach using Data Envelopment Analysis, Hypervolumes, and Finite Mixture Models, assessing the merits of these related methods when applied to assessments of forest condition in practice (e.g. probabilistic nature of statements, treatment of non-convex topologies, and data and processing requirements). Although our subject matter is technical, our work is motivated by Canadian efforts to develop practical tools that can assess overall forest condition in an objective way. Our contribution lies in synthesis and application of existing methods in the specific policy context of forest degradation monitoring to manage biodiversity.