Finley's Rules for tornado forecasting: how accurate were they?

[Elaine Spencer]

In 1888, Sgt. John P. Finley of the US Army Signal Corps published in the American Meteorological Journal his 15 "rules" for tornado forecasting. These rules were based on several years of observations and surveys of tornado outbreaks, including the "Enigma" outbreak of 1884:

Finley's Rules

The rules/criteria he listed were:

1. Presence of a well-defined low pressure area.
2. Slow progression of the low increasing flow northward of heat and moisture into the southeast quadrant.
3. A north-south or northeast-southwest orientation of a trough-like low.
4. The descent of a well-marked anticyclone in the rear of the low.
5. High temperature gradients.
6. Increasing wind velocities of the southeast, southwest, and northwest quadrants of the low.
7. Northward curve of the isotherms in the southeast quadrant and eastern portion of the southwest quadrant of the low.
8. Southward curve of the isotherms in the northwest quadrant and the northern portion of the southwest quadrant.
9. High temperature gradient between the noses of opposing curves of temperature.
10. Increasing high humidity in the southeast quadrant of the low.
11. Maximum areas of tornado frequency for each state.
12. Occurrence of tornadoes in certain parts of the country, in certain months of the year.
13. Tornadoes frequently occur in groups with parallel paths, within a few miles of each other.
14. Tornadoes always occur in the southeast quadrant of a low several miles southeast of its center.
15. Easterly curve in the southwest and northwest quadrants of a line separating the northerly and southerly surface winds of the low.

I'm assuming that criteria 1-10 would be used to determine the general area where tornadoes could occur -- corresponding to where a tornado watch would be issued today -- while criteria 11-15 would narrow down or pinpoint the area most likely to be hit. Given that Finley was working with 19th century communications and technology, I'm wondering just how accurate these rules turned out to be. Are they fairly close to what forecasters would look for today, or did he get some things totally wrong?

 

[Jeff Duda]

In an absolute sense, his forecasts turned out to be right some 96% of the time. However, given the climatology of tornadoes, when you look deeper into where that 96% actually came from, you see that his forecasts generally did not beat random chance (based on climatology). Read this journal article for a full examination of this: The Finley Affair.

Speaking more colloquially, some of these rules are sensible in the context of synoptic scale weather forecasting. We know that tornadoes tend to occur near synoptic scale cyclone centers and near baroclinic zones (i.e., regions with temperature gradients on constant pressure surfaces). However, these rules are very broad and have poor discriminatory ability. Nowadays we know that tornadoes actually occur when very localized and intense vertical vorticity is generated near the surface, almost always because a particularly strong updraft that is almost exclusively caused by intense thunderstorms tilted existing horizontal vorticity into the vertical and stretched it out. None of these processes can be diagnosed via examination of synoptic scale features. What synoptic scale considerations give us are areas where the thunderstorms that are likely to produce these tornadoes are most favored on the large scale.