Sources of vegetation model uncertainty: Differences in soil and climate data inputs affect future projections

Background/Question/Methods

The MC1 vegetation model has been used to inform land managers about potential future dynamics of fire, carbon, stream flows, and habitat shifts. Uncertainties are associated with the input datasets, the models that generate climate projections, the downscaling process, and the structure of the vegetation model. We quantified some of these uncertainties. We compared input datasets from two past projects with the same spatial resolution (half degree lat x lon) but different domains (conterminous USA and global). We wanted to see how the two input datasets differed and how much these differences affected resulting plant carbon pools. NATSGO soil was used for conterminous USA (VEMAP) and FAO soils data were used for the global domain. Climate data for the conterminous USA were provided by the PRISM group while global data were provided by the Climatic Research Unit. Except for temperature, ratios of USA / Global data were calculated for each grid-cell. Differences were used for temperature data. For all time-sensitive variables relative values were calculated for each time step and the median, 10^th percentile, and 90^th percentile values were calculated for each grid-cell. Soil attributes examined included bulk density, mineral depth, and percent clay.

Results/Conclusions

The two input data sources produced comparable precipitation values generally, but over the Northern Rockies and the Cascades VEMAP data were almost always higher. The only place in the conterminous USA where precipitation was almost always lower was along some of the eastern slopes of the Cascades. Temperature values from the two sources were comparable, but VEMAP was almost always lower along the crest of the Northern Rockies and higher in the Colorado Plateau. With the exception of precipitation differences in input levels did not translate into comparable differences in outputs. Higher precipitation did generally yield higher vegetation carbon levels. Plant carbon pool sizes differed between the two different sources to a far greater extent than was observed in any of the input variables. VEMAP plant carbon pools were consistently higher over 60 percent of the map and lower over 20 percent. The apparent lack of correlation between inputs and outputs may results from the fact that only one output variable was chosen. The values in it reflect a cumulative impact. In the future we will examine in detail carbon fluxes and fire emissions.