Tuesday, July 24, 2012

The Flipping, Flopping Severe Prediction Odds

Yesterday's severe weather event in the Northeast wasn't widespread (it wasn't predicted to be) but thunderstorms did produce severe criteria wind and hail across the northern and western parts of the region but once again fizzled as they approached the city.  You can thank the prolonged dry spell in place over Philadelphia, Southeastern Pennsylvania, and Northern Delaware for enhancing the fizzling trend as these storms moved into the region.

While yesterday wasn't a textbook severe weather day and was more typical of summertime storms that can go severe in spots, one point of contention that I've had from time to time and am having again is with the constantly shifting Storm Prediction Center outlooks over the course of a day.  The SPC issues severe weather outlooks a few times daily -- usually an early morning outlook around 2 AM, with updates around 8:30, 12:30, and 4:30 through the remainder of the day.  Those updates can adjust some of the basics with what is expected...but to see a significant jump in percentage odds in some locations is a bit concerning.

I posted yesterday's 2 AM outlook, followed by the 8:30 update and the 4:00 PM update from yesterday afternoon.  You can see a pretty pronounced shift in the slight risk area (the yellow shading, 15% or greater odds) from the early morning update, to the one done at 4:30 today.  The early morning update is the one that local forecasters rely on to communicate news on severe storm potential to their audience during the morning news and is arguably the more critical update of the day...the 4 PM update is effectively a nowcast as much of the day's severe weather has already popped or is in the process of doing so.

The decision at 8:30 to add Philly and North Jersey effectively added millions of people to the risk area that the Storm Prediction Center deemed could see a severe storm.  While strong and severe storms were far more prevalent to our north than they were locally, severe weather was really not an issue in the city.  That said, it's not the first time that the Storm Prediction Center has been adjusted severe weather possibilities over our region during the course of a day.  The day of the derecho was another such example...and even then there was a failure to adjust by the Storm Prediction Center even into the evening hours.  A derecho is a tougher beast to forecast but the adjustments were slow through the day...and for those in South Jersey it took until 9:30 PM for a slight risk to get issued...just a couple of hours before the line moved in.

Severe weather is not an easy thing to forecast as you are dealing with localized factors inhibiting or enhancing thunderstorms...and in many cases in our region the scattered nature of severe weather (the more common type of storms we see) can yield a nasty storm in one town but a miss five miles down the road.  This type of stuff is part of the nature of the beast; however, the pattern of shuffling with outlooks (adjusting in and out of slight risks, or changing from slight to moderate risks on a day that isn't optimal for severe weather) leads to forecast confusion with the public.    Normally, the Storm Prediction Center does a pretty good job on forecasting severe weather threats and sometimes are great at catching threats that aren't readily noticed on computer modeling.  However, the shuffling and "shift" in forecasts has been a trend that seems to have increased somewhat on their end in recent months.

To me, consistency is a more prudent course of action and getting it right initially should be the goal.   You will see changes as conditions warrant...and there will be instances where an area that's on the fence will move from no risk to slight through the course of the day.  These nudges should be minimal and not as frequent as they have been over the past few months.  The Storm Prediction Center's goal should be to get the first forecast right and only adjust when there's a major reason to do so (derecho or something of the sort that's outperforming model guidance).   To err on the side of caution by putting a borderline area into a slight risk even if the risk is iffy might be better for us who communicate information so we can share what the potential is and get the word out appropriately.

A possible solution to this is to lower the slight risk threshold from 15% to 10%...this would seem to help boost areas that are on the fence that shift into a slight risk during the course of the day that would start out not being in one and probably more effectively and consistently forecast severe weather threats.  Regardless, less waffling and more consistency on days where severe weather to be a broad and scattered threat would go a long way to deliver a more consistent message...instead of catching people off guard.