Sensitivity of a dynamic global vegetation model to input soil data: Trends, mechanisms and needs for better data

Background/Question/Methods

Dynamic global vegetation models (DGVM’s) simulate terrestrial vegetation’s transient response to climate conditions. Multiple DGVM’s given the same climate and soil input data have generated divergent forecasts. The uncertainty represented by this divergence has been the main challenge for advancing DGVM science.

We investigated the sensitivity of BIOMAP, a new DGVM with a focus on the vegetation dynamics in North America, to two widely used soil data sets: one dataset from Food and Agriculture Organization of the United Nation (FAO) and the other from USDA Natural Resources Conservation Service (NRCS). BIOMAP uses soil data to calculate soil water holding capacity, which in turn affects vegetation growth simulation. We performed geographic analysis of the two data sets and the resulting soil water holding capacity maps to identify spatial trends. We simulated vegetation dynamics for conterminous US (CONUS) from 1900 to 2100 to identify temporal trends in the impact of soil data on BIOMAP, and to assess the overall sensitivity of BIOMAP to the quality of soil input data. Based on our analysis, we identify regions in the U.S. where better soil data is needed.

Results/Conclusions

NRCS soil data has greater detail than FAO data. Greatest differences, as reflected by soil water holding capacity values, are in the Great Plains, the Southeast, and California Coast. These regions coincide with areas most difficult for DGVM’s to simulate accurately. BIOMAP has high sensitivity to soil data: 30-year NEP values for CONUS in the historical period can vary 67%, and under future climate scenarios 30-year NEP values can vary more than 100% when poorer quality soil data is used. Sensitivity is greatest in the semi-arid zones of the Southwest and in the Great Plains. Temporally, sensitivity is greater during the next 50 years when BIOMAP forecasts increased GPP on a continental scale, than in the second half of the century when simulated NEP has a decreasing trend. Soil data has a direct impact on vegetation simulation by mediating and extending the availability of water, especially in water-limited systems. Improvement in data quality is most critical where BIOMAP is the most sensitive, in the Great Plains, California and Southeast.