Realistic representation of forest harvesting for large-scale models – integrating harvest information from national forest inventories to LPJ-GUESS

Harvesting of wood is one of the key processes of forest management, strongly impacting the structure and dynamics of European forests. This makes accounting for it crucial in any large-scale analysis of forest ecosystems. Yet, the representation of forest harvests in large-scale models is typically far from realistic, as the actual management regimes are not well described by simple rules or even by formal management guidelines.

Here, we show an implementation of national forest inventory (NFI) -based forest harvesting regimes in a demographic vegetation model, LPJ-GUESS. In our approach, the probability of harvest in the model simulation is based on frequency of harvest events in the NFI data in forests with similar structure and geospatial location. Similarly, the characteristics of the harvest event (the percentage of the removed tree basal area and, in case of partial harvests, the tree size targeted in the harvest) are based on the observed harvest events in the data, and depend in the simulation on forest structure and location. This means that model simulations are dynamic, responding to the real state of the forest. We demonstrate this with several countries in Europe, for which we have earlier created NFI-based harvest regimes based on analysis of more than 180 000 forest plots. Forests are simulated with LPJ-GUESS with different forest harvesting set-ups, allowing us to compare the outcome of the suggested NFI-based harvest implementation to other approaches, including simplified clear-cut rules and density-based thinning (based on Reineke’s rule). In addition, the simulation results are compared to observational evaluation data.

Moving from simple rule-based approaches to observed NFI-based harvest regimes can bring the model simulations closer to how forests are actually currently managed. Our approach blending big data and dynamics modelling has potential to both enable improved assessments of continental-scale carbon dynamics and provide a realistic reference to which potential future forest management changes can be compared to.