Spatial and physiological detail in crown representation matters when simulating tree growth

Climate change is projected to expose nearly 70 % of tree species to novel temperature and moisture regimes. Process-based models offer a powerful approach to predict how forests might respond to these unprecedented climates. However, process-based forest models commonly assume spatial homogeneity along one or more spatial axes, which limits their ability to fully capture the interplay between forest structure and tree functioning. Here, we present ForSTEM, a novel process-based, individual-based, spatially explicit modelling approach that simulates inter-annual variation in tree growth by capturing interactions between forest structure, microclimatic conditions, and tree physiology. Our first aim was to validate ForSTEM on dendrochronological data using five output metrics that span a range of time scales and crown dominances (dominant, co-dominant and suppressed) in three tree species (Pseudotsuga menziesii, Pinus sylvestris, and Fagus sylvatica) growing in the Netherlands. The model made robust predictions of long-term (30 year) tree growth at both the plot (R2=0.95) and individual tree levels (R2=0.76), as well as short-term (intra-annual) growth patterns (R2=0.23-0.66), but not at a yearly time scale (R2=0.02-0.2). Our second aim was to test whether model predictions improve with an increase in spatial detail in crown representation and leaf plasticity to micro-climatic conditions within the crown. Our findings showed that predictions of long-term forest growth can be improved by a detailed representation of the dynamic link between canopy structure and leaf functioning at small spatial scales, but also that major knowledge gaps remain in predicting variation in inter-annual growth.