[NARCCAP-discuss] High observed vs. modeled errors for precipitation

[Kelli Walters]

Hello,

I am evaluating which NARCCAP model runs best fit my observed data to  
determine which ones I should use as climate projections in a  
hydrologic model. I am comparing the past data set (1970-2000) from  
the models to observed data for temperature (max and min) and  
precipitation. For temperature, the models seem to accurately reflect  
the observations, with Nash-Sutcliffe efficiencies around 0.5 (r2 of  
0.5-0.6) for daily comparison and in the range of 0.6-0.9 (r2 of about  
0.85) for monthly averages.

However, for precipitation, my errors have been much larger. I am  
getting Nash-Sutcliffe efficiencies of around -0.4 to -1 (r2 of 0.02  
and smaller) for both daily values and monthly averages. What I am  
wondering is: are these error values normal or in the same range as  
other people are getting for precipitation?

Do you have any papers or references that give these error metrics for  
precipitation when comparing the modeled data to observations? I have  
been looking for literature to back up what I am finding, but I can't  
seem to find what values are considered "acceptable" for precipitation  
error. Any help or sources you can provide on this would be much  
appreciated.

Thank you,
Kelli


-- 
KELLI WALTERS, EIT
M.S. Candidate
Civil Engineering | Water Resources
Oregon State University
waltersk at onid.orst.edu