How often does the model show a forecast of 1 inch of precipitation in the 24-hour period ending 12 UTC over your home? 2 inches? 3 inches?
How does this compare to the frequency such events are actually observed?
Could it be that a model forecast of 3 inches in 48 hours is so unusual that it indicates a rare event, perhaps with far higher actual rainfall amounts? Or could it be that such a model forecast occurs spuriously on an occassional basis so it's nothing to get excited about?
Does it matter which types of observations we use for "ground truth"?
The purpose of these plots is to shed light on these questions with some data.
For example, consider this forecast for the Houston area for tropical storm Allison.
Note that most of the heavy precipitation is falling from the convective parameterization, so this is not a case of spurious grid-scale "precip bombs." Also notice the enormous scale of the 4-inch area, almost a 3-degree by 3-degree area (that's 300 km x 300 km)! Now consider the Eta model precipitation climatology on the 90-km AWIPS #104 grid during the preceding 14 months:
This shows that such a precipitation forecast was unprecedented in this
location – the model almost never produced four-inch amounts over the U.S.
Gulf Coast states during the 14 months preceding this forecast! It's the model's
way of tipping off that something extraordinary might occur, something possibly
well beyond what it is capable of predicting. It does not mean that such amounts
will always occur somewhere or will happen in the location the model indicates
– you have to bring all your forecast tools and knowledge to bear on the
forecast problem. In this case, heavy rain was to be expected where Allison
tracked, and this was suggesting extreme amounts were possible, beyond anything
that happened in the past year anywhere over the south-central states. Moreover,
the total water volume predicted is enormous – imagine if the tropical
storm stalls and all that water is dumped in one place instead of being distributed
over the model's track! Take a look at what was
observed during the 24-hour period ending at the end of this 48-hour prediction.