Sandia Model Comparison Activities

After several discussions and much thought, we have outlined a plan of attack for comparing the model simulations. Thanks goes to Jessie for putting up with our hour-long list of questions!

The idea is to start simple and work towards higher complexity. The first step will be comparing model output directly to the observation data across the study area. We intend to look at the similarity (or lack thereof) of the statistical model of the different data against the corresponding output variable for the models for the time-span in question (1980-2010). Some questions we hope to answer are: 1.) Are the statistical models similar (i.e. do they use the same PDF). Why or why not?, 2.) Are they within each others uncertainty bands?, 3.) Is stationarity (or non-stationarity) preserved?, 4.) Is there spatial correlation and if so, what are the defining features (i.e. latitude, altitude, topography, etc.) that contribute to the correlation, and 5.) what data / variables compare will and which do not? Why or why not?

Any others?

Our hope is that by initially concentrating directly on the data, that we'll begin to identify other comparisons that could require developing hydrological models of some sort. Directions where this could head could be, 1.) reconstruct river hydrographs from the climate model output and compare those to data, 2.) create systems level performance models in areas of high and low spatial correlation (e.g. environmental impacts, human impacts, etc.), 3.) forcing existing watershed/hydrology models using the model output, 4.) constructing a detailed watershed model from scratch, and 5.) some combination of the above. The models mentioned would all be populated with 4 sets of inputs being the two refined climate change model output sets, the standard model output, and observation data.

This second part is purposely vague as we are reserving the right to change our minds as we learn more from the first exercise!

Our current effort is now focusing on understanding the GCM model outputs (what do they really mean) and collecting any and all observational data. With regards to the latter, if anyone has any ideas or banks of data just sitting around waiting to be used, let us know.