For the RESONATE project, my task aims at developing high resolution future forest trajectories and disturbance maps for the European continent. Continental scale modelling always comes along with trade-offs regarding the detailed processes. Taking this into account, we follow a bottom-up approach, where we use detailed information from local process-based forest simulations to train deep neural networks. For this, we collected forest simulations under different climate scenarios from hundreds of locations across Europe, covering large gradients of environmental and climatic conditions. By combining simulations from different regions, we can explore the relationship between forest dynamics and climate signals using deep neural networks. These neural networks learn to represent forest dynamics depending on environmental and climate conditions, allowing us to upscale the forest dynamics to continental scale. We believe that with this approach we will make a step towards better capturing local scale dynamics at the macroscale.
But guess what, forest modeling means we spend most of the time in front of our screen, working on code and data that eventually allow a glimpse into the future of forest ecosystems. Although I spend a lot of my leisure time hiking, cycling and sometimes ski touring in the mountains, professionally I spend very little time in the field. Therefore, I was really happy to join the excursion as part of a conference we organized in Berchtesgaden some months ago. The occasion to go to the field with colleagues who spend a lot of time there and visit the system that I am currently modelling is very special and of course informative. And for me, coming from a macroecology background, it is also particularly important to see gradients in the mountain landscape and discuss their impact on vegetation processes as well as disturbances.
Leave a Comment