With every tornado outbreak comes a bevy of tornado statistics. Often times these stats will arrive within a day of the outbreak ending, and they are almost always based on preliminary information. They also are sometimes missing that important “preliminary” caveat. In order to understand the difference between tornado reports and actual tornadoes, you need to be familiar with the process of how a tornado gets confirmed by the National Weather Service (NWS).

  1. A tornado, funnel cloud, and/or damage is sighted and reported to the NWS, or NWS doppler radar detects a debris signature. This can either get logged as a “tornado” or “thunderstorm wind damage” event type, depending on the confidence level as to whether or not a tornado actually occurred, and sent out as a preliminary Local Storm Report product. The main purpose is to transmit information about damage and severe weather occurrences, rather than an exact tabulation of the number of tornadoes.
  2. Damaged areas are surveyed and reports are further scrutinized after the event is over. This will lead to an official determination as to whether or not a tornado occurred. If a specific area is impacted by numerous and/or long-tracked tornadoes, this process can take some time.
  3. Several months after the event, the tornadoes that were determined earlier are officially certified in NWS Storm Data. The tornadoes and associated statistics that are put into Storm Data are the basis for the official US tornado and severe weather record.

The preliminary reports, or what is outlined in step one above, are what is immediately available and frequently make it into summaries and articles in the following days. The reports are conveniently collected on a wonderful website database managed by the Storm Prediction Center. The maps and preliminary counts on that website give a good sense as to the overall magnitude of the event and the location of the greatest reports.

Taking the little red number off the SPC Storm Reports page and saying “X number of tornadoes occurred yesterday” is usually inaccurate.

However, it is important to remember that the preliminary tornado numbers can vary quite a bit from the actual number of tornadoes that occurred, depending on how Local Storm Reports were entered during the event. There could be under-counting of tornadoes if many were entered in as “thunderstorm wind damage”. These will often be accompanied by a comment such as “possible tornado”, but would still show up in report databases – such as the one linked above – as a blue wind damage dot.

There can also be over-counting, and this tends to be the more common problem. One tornado can be reported by multiple individuals, or a longer-tracked tornado can produce damage in multiple communities – resulting in several reports. Therefore, taking the little red number off the SPC Storm Reports page and saying “X number of tornadoes occurred yesterday” is usually inaccurate. It would also be inaccurate to say “X number of tornadoes were reported yesterday“, because that still implies that X was the number of tornadoes. The most accurate way to convey this information would be to say “X reports of a tornado were received, but the actual number of tornadoes may be different.

The ratio of tornado reports to actual tornadoes in the United States each year over the decade from 2005 to 2014. On average, there were around 87% as many confirmed tornadoes as compared to preliminary tornado reports.

If you consider the past decade, from 2005 to 2014, there were 13,822 preliminary tornado reports logged via Local Storm Report, but only 12,019 confirmed tornadoes that made it to “step three” on the process I outlined. This means that on average there were approximately 87% as many confirmed tornadoes when compared to preliminary reports, or a general over-counting. Others have noted this trend, as well as a general inflation in the number of preliminary tornado reports as time goes on. Additionally, the over-counting has been fairly consistent from year-to-year over the past decade, as you can see on the graph above.

87 percent may not seem like a significant deviation, but more substantial differences arise with regularity in specific events. The problem of over-counting can particularly be exacerbated in significant tornado outbreaks, as that is usually when longer-tracked tornadoes occur – the type of tornadoes that can have numerous reports along their track. Those also happen to be the cases when there tends to be greater interest in totaling up the number of tornadoes and putting things in a historical perspective. In other words, it seems more likely that inaccurate tornado figures will get cited in the biggest events.

292 preliminary tornado reports were logged during the April 27th "Super Outbreak".

292 preliminary tornado reports were logged during the April 27th “Super Outbreak”.

A notable example of over-counting is the 27 April 2011 “Super Outbreak”. Many violent, long-tracked tornadoes occurred that day, each affecting multiple communities along their damage path. In total, 292 preliminary tornado reports were received from 8 AM EDT on April 27th to 8 AM EDT the following day. However, there were 173 confirmed tornadoes in that same time period – still a significant number but only 59% of the preliminary figure. The long-track nature of the tornadoes contributed to the number of reports. 23 of the tornadoes had at least a 25 mile long path, with 7 of those tracking at least 50 miles.

The bottom line is that preliminary severe weather reports are a great resource, but like all great resources be sure to use it wisely! The initial numbers can provide some useful information with the proper context. However, usually if you wait about a week you will have a fuller picture of how many tornadoes actually occurred in a specific event, and that will allow you to place it in a historical context much more accurately.