You could do that, but then that fails to take into consideration smaller/larger events. Plus, since we measure to the tenth, the mode may not capture that many storms (won't be very representative).
The best bet is to just stick with one average, and if you feel the data is skewed, report the standard deviation so we know the average distance of cases from the mean. If you want to get fancy, and your data is in SPSS, you can also report the skewness value to see if some storms are biasing the data set.