Predictive analytics of tree growth based on complex networks of tree competition

Competition between individual trees is a major factor influencing the development of forests. However, due to the complexity of such interactions, that span over vast geographic areas, systematic analysis of competition has only recently become possible through the concepts of so-called predictive analytics. The rationale behind the utilised approach is that a prediction model, which is capable of forecasting future increments of tree development parameters accurately, contains knowledge about the underlying relationships that govern them. The analysis of such model, therefore, holds the potential to reveal new insights into the critical factors that influence forest developments. Within this study, we utilise an Evolutionary Algorithm in order to enable predictive analytics based on a complex-network representation of competition. This allowed us to study the patterns related to spatial distribution of individual trees. We discovered that triplets of competing trees, and their betweenness centralities, have significantly greater influence on the development of each individual tree than traditionally observed parameters like the number of a tree's competitors and distances between them. While this indicates preferable spatial patterns for optimal forest development, the introduced methodology proved to be an efficient predictive analytics tool that allows for their discovery.