First page Back Continue Last page Overview Graphics
3D-Var Parameter Tuning
The first five parameters had by far the largest positive impact on the cool season test case (Jan 23, 00z, 2000). The improvement was largest for the fit of the guess to the radiance data. This resulted from significant increases to horizontal and vertical correlation length for temperature and an increase in amplitude of the background temperature error. This went along with decreases in background error for stream function and decrease in vertical correlation of potential function (divergent wind).
Although these changes seem to focus mainly on radiance data, the preliminary results from the eta22 parallel show consistent improvement in fit of the guess to winds, relative to the operational eta32. We propose one more round of tuning, this time with just the first 8 on the above list, and again only with the cool season test case. This will be finished by Friday, Sept. 8, and we will then be ready for the retrospective run of the January snowstorm and final parallel run before implementation.