Probably more a practical decision than anything else- if a storm is expected to reside in an area for longer than 3 hours then evcuate. The key numbers probably are travel speed (10-20 mph) and area of exposure (150-300 miles) so a worst case scenario is 30 hours but never expect head on so assume 10%- 3 hours.

DNV says 3 to 6 hours. but it discusses 3 hours as a general case. if you expect duration more than that you should use your value instead of 3 hrs.
 

Guess I am wondering: why does DNV say 3 hours? What is the basis for this particular storm duration?

I think maybe based on observations, One sea state (or storm) normally takes between 3 to 6 hours,
 

Looking at my Barltrop and Adams page 256. It states:-

"Over an interval of 3 hours the statistics do not change very much and the seastate may be described by significant wave height (Hs) and mean zero crossing period (Tz)."

 

And DNV-RP-C205 states

"Three hours has been introduced as a standard time between registrations of sea states when measuring waves, but the period of stationary can range from 30 minutes to 10 hours."

DNV-RP-H103 mentions 3 hours as well.

Remember the 1.86 factor which is said to be the ratio between max wave height and sig  wave. I've used that for years without complaint.

So I guess the stats people think that after 3 hours you just see (statistically) a max wave height that doesn't differ much from using say a 4hour, 5hour... 10 hour etc. storm data.

 

It's in order to capture all possible extremities of the sea state. Generally 3hrs is statistically sufficient to achieve this, using more than this is waste of project time frame.
However for some analysis (e.g. Fatigue analysis), if you can prove that a time duration less than 3hrs will give results that are statistically ok compared to the 3hrs, it will be acceptable.