by David Hoadley

Storm Track, Sept/Oct 1994

Editor's note: The Convective Outlook is a thunderstorm hazard forecast issued by the National Weather Service Storm Prediction Center, and is intended as guidance for local forecast offices.

It may be helpful to briefly consider the reliability of "DAY-2 convective outlooks" for tomorrow's severe storms. Based on these, I drove excessive distances in 1994, averaging 600-700 miles a day and 4-5 hours sleep a night for extended periods, trying to land in the next day's ideal location. Much of it was wasted effort. Especially frustrating were the tornadoes that occurred where I had just been 400 miles earlier, the day before!

These forecasts are issued daily through NWS offices or via computer accessible data sources in your home or motel. They are initially available by 7AM CDT and occasionally updated around noon or later for the following day.

Each of us has a different chasing style and "comfort level" with various data products. This early morning forecast is especially important to me, since I am usually on the road by mid-morning and lack easy access to later updates. I also prefer making early motel reservations to avoid the evening's tourist rush and assure a good room with data-accessible phone lines. conse- quently, that early morning forecast assumes added importance. Sometimes I call for a noon revision, but NWS doesn't always change it. At such times, I am also preoccupied with today's building storms -- and unwilling to take time and stop for updates (Where is the roadside phone? Does it work? Can I get through? Has the forecast changed?). Also, I don't like to bother NWS staff when severe storms may be building in their area, just when they don't need chaser interruptions. So, for various reasons, I rely substantially on the early morning DAY-2 forecast.

How accurate was it during 1994? Based on available data, not especially so -- and in some cases not at all. I analyzed 13 days for which good records were retained (4/24-25, 6/5-11, 6/25-26, 6/30 & 7/2), listing early morning (122) DAY-2 and actual DAY-1 risk areas ("Slight Risks" were assumed to correspond to DAY-2 severe areas). In determining direction and distance of shifting forecasts, I measured from the approximate mid-point on the western boundary of each area (storms usually begin near there and sweep across the forecast area). Results:

Ten of thirteen DAY-1 risk areas shifted west from the earlier DAY-2 forecast, and three shifted east. Those shifting west were an average of 123 miles closer to the first tornado on 3 days but would have given up an average of 63 miles further for 7 others. In other words, if you had remained overnight at the western (mid-point) boundary of the DAY-2 forecast, you would have been closer to 7 of 10 the next day.

The three that shifted east were an average of only 55 miles closer to the first tornadoes on those days. since these somewhat offset the 3 closer tornadoes to the western shifting risk area, remaining at the boundary of DAY-2 would still have been a good compromise. Surprisingly this would also have been true regarding actual distance from the initial tornado or watch. DAY-2 mid-point western boundaries were an average of 296 miles away versus 313 miles for the later and presumably more accurate DAY-1 forecast!

There is insufficient data from earlier years for comparison, but one day each from 1990 and 1993 again shows westerly shifting risk areas from their DAY-2s, with DAY-1 only 10 miles further from initial tornadoes.

One interesting footnote to 1994 illustrates the greater problems that forecasters had with that season. In 1993, the average distance from DAY-1 mid-points to tornadoes on five days (for which records are available) was 238 miles -or 75 miles closer! The longer distances in 1994 indicated greater uncertainty at the start of each day and -thus- larger risk areas.

Other experience of this chaser over the years suggests that NWS computer models have a recurring bias for too rapid eastward advancement of springtime weather systems. A notable example was the morning of May 5, 1983, when AM Weather (NWS data) forecast a strong central plains low to move to Chicago and trail a cold front across southern Texas. Instead it remained over the central plains and spawned dozens of tornadoes over the following three days.

On the very morning of a major outbreak with panhandle "wedges", the models busted badly on where and how fast that system would move. There have been other, less dramatic instances over the years, but they can be found.

This abbreviated study and recollection is obviously not exhaustive or definitive, and one season does not a pattern make. However, it does suggest that chasers need to take greater care in using DAY-2 forecasts. Each spring will be unique, with different dynamics and challenges to understand.

A good approach to any new season may be to hold back the first few days on rushing to your DAY-2 area. Regardless of how potent sounding, don't try to drive the full distance back from a fading western storm -all the way east to some location well inside the new area. Instead, if anywhere in the central plains, go to -or near- its western boundary, wait, and see what the morning brings. You are then positioned to move a moderate distance west or east. How well are the models dealing with that spring's patterns? Do risk areas jump back and forth or are they a logical progression from the day before? A few days should tell this tale. That is how I will plan my 1995 chase strategy. If '95 is anything like '94, the western boundary of DAY-2 is three times likelier to shift west with DAY-l. Anyway, remaining with the DAY-2 boundary overnight may not lose more than 60-70 miles in either direc- tion from most of the nearest tornadoes, regardless of which direction DAY-1 goes. However, committing to a deep DAY-2 location the night before can lose more if DAY-1 shifts far back to the west. In short, a skeptical chaser may do better in the long run over a gullible one, who pushes the limits of exhaustion and risks losing the insightful edge of experience.