THREE-DIMENSIONAL DEEP LEARNING FOR LEAF-WOOD SEGMENTATION OF TROPICAL TREE POINT CLOUDS

We present ForEdgeClim, a process-based model that simulates forest microclimate temperatures from edges to forest interiors. The model combines high-resolution forest structure, meteorological data, and a physically based energy balance that includes vertical and lateral radiation and heat exchange. Validation with field measurements shows that ForEdgeClim captures observed edge-to-core temperature gradients, supporting its use for studying forest fragmentation and climate impacts.

We integrated state-of-the-art observations of the structure of the vegetation in a temperate forest to constrain a vegetation model that aims to reproduce such an ecosystem in silico. We showed that the use of this information helps to constrain the model structure, its critical parameters, as well as its initial state. This research confirms the critical importance of the representation of the vegetation structure in vegetation models and proposes a method to overcome this challenge.