Applying complex network metrics to individual-tree diameter growth modeling

Mathematical approaches that explain growth increments, while accounting for inter-tree competition, are advancing. As interactions between trees represent a network structure within an ecological system, they can be described by topological metrics. These metrics may support individual-tree growth modeling in line with ecological processes. Our objective was to compare the performance of traditional indices and complex network metrics in modeling diameter growth. The study area is a semi-deciduous seasonal montane forest in Brazil, where Copaifera langsdorffii was naturally dominant from 2010 to 2017. We selected this species as the subject tree in our study. Tree competitors were identified using the Bitterlich procedure (basal area factor = 4). The periodic annual diameter increment (PAId) was modeled using four strategies, including a genetic algorithm and the random forest method, which involved different competition metrics: distance-dependent, distance-independent, and semi-independent indices, as well as topological metrics from complex networks. We assessed modeling performance based on the analysis of error metrics. The approach developed using topological metrics shows strong potential to explain the PAId through a competition network within an ecological context. The innovative approach used in this study offers robust modeling to support forest growth analysis. Therefore, we encourage the application of this interdisciplinary tool to generate insights into forest science.