A new analysis of the validity of SNAP climate models is now available.

How can we best judge the reliability of SNAP's climate models? Obviously, we have no future weather data with which to compare model projections. However, we do have historical weather data. By generating model runs for past time periods and then analyzing the statistical relationship between real weather patterns and model outputs, we can assess the validity of the model.

SNAP climate projections are based on outputs from five General Circulation Models (GCMs) from the 2007 Intergovernmental Panel on Climate Change. We compared downscaled GCM output to historical weather station data based on four different metrics: monthly mean values, seasonal (month-to-month) variability, annual (year-to-year) variability, and long-term climate change trends. We performed each of these comparisons with both temperature data and precipitation data.

Overall, SNAP models performed well when analyzed for concurrence with measured climate data with respect to monthly mean values, seasonal variability, annual variability, and long-term climate change trends. Each model exhibited strengths and weaknesses. By comparing the performance of the different models on each statistical test region by region, we can fine-tune how we use these models for predictive purposes. In cases where generating mean values are more important than capturing climate variability, using a composite of all five models is likely to yield the most robust results.

The full document is available for download here or via the documents page listed under "quick links" on our home page.