Modelling and analysis of carbon-climate interactions in northern forest ecosystems

Summary
This project focuses on the carbon-climate interactions of boreal forests, combining multiple data sources with process models to understand and predict the evolution of these ecosystems. The project is joint between the Universities of Edinburgh and Helsinki.

Research questions
The overarching objective of the project is to advance our understanding of climate-C coupling in forest ecosystems and the representation of this in a range of process-based ecosystem models. To this end, we will address the following questions (i) how do alternate representations for internal processing of C in forest ecosystem models compare?; (ii) how does process uncertainty affects forecasts of C sink/source strength?; (iii) what is the climate and management sensitivity of different C processes and their interactions?

Methodology
The project builds on existing modelling resources, field observation data, and model-data fusion expertise, at Edinburgh and Helsinki. The project will use two forest C models, one developed in Edinburgh (DALEC, Williams et al. 2005) and the other in Helsinki (PREBASSO, Valentine and Mäkelä 2005, Minunno et al. 2019). These models differ in their representation of key C cycle processes - photosynthesis, allocation, turnover, phenology, respiration, decomposition, mineralisation. They also differ in the climate sensitivity of these processes. Further, there are >15 different versions of DALEC, varying in the number of C pools, and process representation. This diversity of model structures, complexity, and climate sensitivity can be explored, quantified and tested across boreal forest systems. The project will determine the strengths and weaknesses of the process representation and their coupling, providing insights to advancing theory and improve model structure.

Models and observations can be combined, so that the observations calibrate and validate the models, while the models provide a test of theorised links between observed processes. Bayesian approaches are revolutionising the science of model-data fusion (Williams et al, 2005; Van Oijen 2017). Bayesian approaches recognise that models should be calibrated probabilistically, based on the uncertainty in observational data. Bayesian analysis also allows for model comparisons and using the most probable combinations of models for making future predictions. Bayesian approaches ascertain that uncertainty is attached to parameter estimates, and the interactions between processes are clarified. These properties make Bayesian analysis an efficient tool for model development and testing.