Interpretation and Uncertainty in Climate, Earth System, Integrated Assessment, and Impact Models and Observations I Posters
Wednesday, December 14, 2016 08:00–12:20
Moscone West Poster Hall
Simplified representations of processes driving global forest biomass in Earth system models contribute to large uncertainty and variability among climate predictions, in particular for the simulations of biomass magnitude, allocation, and the responses of biomass to changing climatic conditions. In this study, we evaluated forest biomass from the historical runs of eight coupled Earth system models in the Coupled Model Intercomparison Project Phase 5 (CMIP5) archive, using a recent data product synthesized from remote sensing and ground-based observations across northern extratropical latitudes (30°N–80°N). Compared to this data product, all models excluding two Hadley Centre‘s models overpredicted global forest biomass in wood by 166%±153% whereas biomass in roots was underestimated by −82%∓2% in all models except the IPSL models (133%±46%). In addition, the IPSL models had the largest biases in total forest carbon mass estimates (154%±51%), which was attributed mainly to the overestimated wood component (163%±56%). Nevertheless, the allocation of modeled forest biomass in roots (21%) and in wood (76%–77%) found in the IPSL models was more consistent with observations (22% for roots and 73% for wood). Our results also demonstrated that both observed and modeled forest biomass was positively correlated with precipitation variations in most regions, while surface temperature was as important as precipitation at higher latitudes. Moreover, small differences in forest biomass between the pre-industrial period and the modern time period implied that the biases in forest biomass may have been introduced at the beginning of the simulations. Our work suggests that caution should be exercised for (1) allocating carbon mass to forest components, (2) apportioning vegetation types within modeled gridcells, and (3) reducing the uncertainty in vegetation inputs for Earth system models with correct vegetation parameterizations during the spin-up processes.