Next, you will fit models for wind event data that have been separated by meteorological cause.
Non-Thunderstorm Data
Start with the non-thunderstorm data first (dataset “CID_Independent_NT-CID-WIND-SPEED”) and create a new Distribution Fitting Analysis called “PDS NT” in the same manner as the all storms dataset previously. Filter the data so that they are above 42 mi hr-1 and evaluate the goodness of fit of the collection of models when estimated using the method of L-moments.
Non-Thunderstorm Type – Partial Duration Series
Generalized Pareto Distribution (L-moments)
Parameter
Value
Location (ξ)
42.30465
Scale (α)
7.68928
Shape (κ)
0.16318
Sample Size
232
After fitting the partial duration series for the non-thunderstorm type, fit the annual maximum series in a new Distribution Fitting Analysis called “AMS NT” and filter to annual peaks the same way as in the "all storms" dataset in the previous task. Ensure the year starts on 01Oct.
Non-Thunderstorm Type – Annual Maximum Series
Generalized Extreme Value Distribution (L-moments)
Parameter
Value
Location (ξ)
54.44436
Scale (α)
6.19255
Shape (κ)
0.15911
Sample Size
31
232/31 = 7.48 events per year. The number of events per year determines how close the annual maximum distribution gets to the GEV distribution (more events per year results in nearer convergence).
Thunderstorm Data
Repeat the process for the thunderstorm data (dataset “CID_Independent_Thunderstorms-CID-WIND-SPEED”), creating a Distribution Fitting Analysis named “PDS Thunderstorm”. Filter the data to a partial duration series using a Minimum Threshold of 42 mi hr-1. Create a Distribution Fitting Analysis named “AMS Thunderstorm” and filter the data to an annual maximum with a year start date of 01Oct.
Thunderstorm Type – Partial Duration Series
Generalized Pareto Distribution (L-moments)
Parameter
Value
Location (ξ)
42.32534
Scale (α)
12.25859
Shape (κ)
0.25561
Sample Size
91
Thunderstorm Type – Annual Maximum Series
Generalized Extreme Value Distribution (L-moments)
Parameter
Value
Location (ξ)
55.611231
Scale (α)
7.201729
Shape (κ)
0.127287
Sample Size
29
91/29 = 3.14 events per year. There are many more non-thunderstorm events per year (about 2.3 times as many.)