[contravariant]

I'm not sure the added precision is helpful if it's just 2.5 days ahead. An extra week of accurate forecasts is a bit pointless if it takes 8 days.

[sp332]

Simulating 8 days would take 8/2.5 = 3.2 days.

[bernardv]

When you’re trying to assess imminent danger from an approaching tropical cyclone, days and hours count. A lot.