The productivity of global forests will have distinct responses to potential climate scenarios, which are important to capture in order to place an economic valuation on forested land. However, it would greatly increase the run-time and complexity of a multisectoral global model to include a processed-based model of forest productivity. Therefore we propose statistically emulating global forest productivity using a large ensemble approach for training data. We will use the Terrestrial Ecosystem Model (TEM), which is a process-based biogeochemistry model that uses spatially referenced information on climate, atmospheric chemistry, elevation, soils, and land use and land cover change to estimate fluxes and pool sizes of C, N and water in vegetation and soils. These pools are influenced by factors like CO2 fertilization, climate change and variability, land-use change, ozone pollution and atmospheric N deposition. We have conducted the large ensemble using scenarios from the Climate Model Intercomparison Project (CMIP5 and CMIP6), and their associated land-use scenarios from the Land Use Harmonization and atmospheric chemistry from the Atmospheric Chemistry and Climate Model Intercomparison Project.

The key questions we are trying to answer are (1) what are the most important predictors of forest productivity and its response to environmental and land-use changes and (2) how can we use TEM to emulate these responses and examine land productivity and land dynamics in the Global Change Analysis Model (GCAM)? We hypothesize that temperature and vapor pressure deficit will be key climate variables and that the initial nitrogen pool of the forested area will have lasting effects on forest productivity. We will initially predict the response of above and below-ground carbon to forest stand age, plant functional type, and climate variables and then develop a spatially referenced model. We will test different statistical methods, for example fixed-effects regression and generalized additive models and determine which approach best emulates TEM’s nonlinear response of ecosystem function to climate.