Pressure Gradient Breakthrough

I decided to investigate the Cordova-Anchorage pressure gradient a little more given the relatively weak correlation compared to the intuitively strong association, as well as the lag that exists between strong wind events and the time of the maximum gradient.

I realized that the change in the gradient over time might have a better physical connection as it would account for the isallobaric wind. Without pulling any new data, I just used the gradient at the time of the report and the maximum gradient to determine the change over time. I plan on trying to analyze the pressure gradient change centered around or just prior to the wind report next.

The result was a 0.56 correlation to the wind magnitude. This was a notable increase from the maximum gradient correlation. More importantly, when replaced in the empirical formula, it makes a significant difference.

In addition, I have rerun the pressure gradient program to be able to pull data from one event that was missing from the initial database. This is the weakest case, and thus an especially important case.

First of all, it actually reverses the dominant group ... the pressure correlation is 0.61 for the wet class and 0.54 for the dry class. The new correlation for the bestfit values is 0.84 overall, which is excellent ... 0.85 for the wet class, and 0.83 for the dry class. This is a huge improvement. Case-by-case, this also removes the biggest outlier! There are two cases with error greater than 10kt. The average error is 4.4kt or 6.0%