Tag Archives: Model Errors

It’s not just snowstorm accumulations that are difficult to predict….

Last Updated

Today’s models are a perfect example of where model prediction of QPF (total accumulated precipitation) is so difficult.

Had this been snowfall, we would have been predicting 1-2 feet of snow for today.  (QPF values were 1-2 inches of water).  Instead, last night’s late models revised that down to 0. 50 inches of water.  So a forecast of 2 foot  snowstorm would have had to be revised down to 5 inches!

(A spectacular change- the evening NAM was predicting 1.29 inches for PHL during the afternoon alone.  The 1 AM model run reduced that to about 0.30!)

Significant errors in forecast QPF  is very a frequent occurrence.   When it rains, no one really notices or cares when the QPF is not forecast properly.  But if it’s going snow, the models let you down in a way that is much more obvious.

It’s often not about “changes in the track of the storm” as the TV people like to talk about.  It’s more about the three dimensional distribution of energy and moisture in a storm that can’t be accurately measured and then modeled.

Today’s forecast is a case in point.

10 AM Update – Yet another turnaround….this morning’s NAM shows a QPF of 2.32 inches for PHL with a band of the heaviest rain right though the Delaware Valley.  A totally different precipitation coverage map compared to last night.   Even last night’s experimental “National Blend of Models” had decreased the precipitation forecast for today in PHL.

Good thing this isn’t falling as snow.

In addition to these large swings in forecast QPF for certain areas, snowfall amounts can be anywhere from QPFx10 or QPF x20!  So errors in QPF are multiplied by a factor of ten or more when we have snow.

While we’re at it, I get a real kick out of the TV “Future Tracker” where they depict the locations of the heavy rain in advance.  I want you all to know those forecasts are not to be ever taken literally.