Simulating climate change effects at various scales: can usable guidance emerge from an ensemble of outcomes?

Background/Question/Methods

Climate change impact assessments require climate change scenarios at a relevant scale for land managers and conservation practitioners. While summaries of strengths and weaknesses of the various climate models have been published, there is still a lack of clear guidance for model/scale relevance to ecophysiological processes and disturbance regime shifts that impacts models simulate. While ensemble means of climate projections are generally in better agreement with observations than any individual climate model, impacts model that simulate ecosystem dynamics and disturbance-climate interactions cannot use mean climate. Furthermore, downscaling techniques have proliferated in the climate change impacts community making comparison of climate projections difficult and providing unclear choices to assess future climate risks.

Results/Conclusions

We have run our vegetation models at difference spatial scales (50km, 12km, 10km, 8km, 4km, 800m) using historical climate and future climate scenarios from a variety of sources including both downscaled general circulation models (GCM) and regional climate models (RCM) projections. We present a commentary of the effect of using these different climate drivers on the results of our simulations: 1) RCMs are often available for limited time periods because of the large size of the result files; these products are inadequate for dynamic vegetation models that require uninterrupted climate time series to simulate trajectories, ie future shifts in vegetation are caused by earlier vegetation and fire dynamics; 2) the RCM regional tiles have different uncertainties based on their boundary conditions and proximity to ocean; for example we have observed differences as large as 40% in precipitation levels for an area where two tiles (including one with ocean boundary) overlapped; 3) RCM historical climate seem to be in general "wetter" than the baseline used in the past (PRISM): we have not been able to use "raw" NCEP-driven RCM-generated current climate. Besides climate data we have compared gridded soil datasets at the various scales and have found lack of agreement between available soil data sources, which greatly affect vegetation model results: for example, some fine -scale soils dataset derived from STATGO and SSURGO show soil depth of ~1m at high elevation in the California Sierras which causes the vegetation model to simulate forested areas which could mitigate the effects of warming but observations show rocky outcrops devoid of soil. We discuss the strengths and weaknesses of this ensemble of results across spatial scales for three vulnerability assessment projects that used vegetation model results to inform management decisions.