Goal 2, Priority 4: Supporting and Conducting Regional-Scale Investigations of Climate and Weather
As climate change and societal vulnerability to severe weather become more apparent, decision makers want to know what changes are likely to occur in particular places. Advances in computational power, modeling and analysis techniques offer opportunities for regional-scale research in the atmospheric sciences, and in the study of environmental and societal impacts. NCAR intends to move ahead rapidly in this area, conducting not only the important fundamental research that must be done but also applying that work through creation of new decision support capabilities.
FY2007 Accomplishments
Click to enlarge. Quantifying uncertainty. The left plot is the .025 quantile of the posterior distribution and could be used as a lower bound for the 25-year return level estimate. Similarly, the center plot is the .975 quantile and can be used as an upper bound. The right plot is the difference of the two, and thus shows the range of the 95% credible interval. The right panel also has black dots at the locations of the stations providing data for the analysis. The units are centimeters of precipitation per day (cm/day).
The output of multi-model climate experiments such as those for the IPCC or NARCCAP pose unique statistical challenges for synthesizing results and quantifying uncertainty. As a result, researchers from IMAGe’s Geophysical Statistics Project (GSP) and ISSE, in collaboration with university researchers, are adapting new statistical methodology for summarizing this unique kind of data. Among the FY2007 achievements:
- Regional climate advances in this area included development of a multivariate Markov random field approach to compare differences in mean temperature and precipitation across different seasons, changes in extreme precipitation, and temperature for gridded regional model output.
- Also, a functional ANOVA approach has been explored (using a European Prediction of Regional scenarios and Uncertainties for Defining European Climate change risks and Effects project subset) to understand interactions and consistent effects from specific models.
- To address climate model biases and uncertainty, the Bayesian approach for interpreting multi-model ensembles was extended to combine results across different regions, and also for a bivariate response (temperature and precipitation).
- Another project using spatial statistics models for processes on the sphere analyzed the similarity of model biases for mean seasonal temperature and temperature trend for the models submitted to the IPCC archive. The models show substantial correlations indicating that biases are not independent across the ensemble.
FY2008 Plans for Strategic Priority 4
FY2008 plans include extending the functional ANOVA approach for regional model output to extreme statistics, including a quantification of how extremes scale in space, from point locations to global model grid cells. The analysis of model bias will be adapted to the NARCCAP experiments.
Related Lab Annual Report Sections:
Goal 2, Priority 4