That’s why outcome analogs should be the method and not theoretical physics.
Very good data since 1960. 60 years. Let’s say there are 3 Miller As per season, 180 sample size.
Break those 180 down to
1. The low passes within 30 miles of Atlanta.
2. The low is more than 30 miles north of Atlanta
3. The low is more than 30 miles south of Atlanta
then overlay that with
1.The high is 30.30 or higher 12 hours before onset
2.The high is lower than 30.30
You will get a decent sample set of 10-20 matching outcomes in each category. No wailing about sample size, the outcomes are what they are
When you take #1 on the storm and #1 on the high and 20 instances arise and 15 times DC got 3” or more then there is your model replacement
When you take #2 and #2 and 22 instances match and 16 times DC got rain then there is your new and improved, outcome based, non model forecasting.
The will is not there mostly due to entrenched commitment to status quo and embedded money issues.
i promised you all I had a solution and not just bitching. Let the naysay begin.