Five stages, one of which is preliminary and one of which is permanent, and the documented reasons the archive is not a fair sample of what happened.

In this lesson you will

  • Trace the report lifecycle from spotter to WFO to Local Storm Report to SPC to Storm Data.
  • Distinguish a preliminary Local Storm Report from the certified Storm Data record.
  • State two documented biases in the report databases and why wording and accuracy matter.
  • Describe how this platform cross-checks a report against radar for plausibility.

The five stages

Your report does not arrive somewhere and stop. It moves through a chain, and at two points along it a human being can change what it says. Knowing where those points are is what tells you which parts of your report have to be right the first time.

Five stages left to right connected by arrows. Spotter, then forecast office, then Local Storm Report, then SPC preliminary reports, then Storm Data at the national archive. A vertical dashed line falls between the SPC stage and Storm Data, with everything to its left labeled preliminary and minutes to hours, and everything to its right labeled certified and months later. A note under the forecast office and Storm Data stages marks them as the two points where the content can be changed.
The two amber stages are where a human reads your words and decides what they meant. Everything you can control happens before the first one. WxAlerts
  1. You. A phone call, a form, a net. What you say is the only evidence that exists at this point.
  2. The forecast office. A forecaster reads it, decides whether it is plausible and what it means, and may act on it immediately in a warning.
  3. The Local Storm Report. The office publishes it, often within minutes. This is the version the world sees first.
  4. SPC preliminary reports. Aggregated nationally the same day, which is where the familiar dots-on-a-map come from.
  5. Storm Data. Weeks to months later, after quality control, into the national archive at NCEI. This is the version that still exists in thirty years.

Preliminary is not a hedge, it is a status

NWS Instruction 10-517 says it in two sentences and they are worth reading carefully, because a great deal of public confusion comes from skipping them:

LSRs are preliminary in nature. The final report of verified weather
events will be listed in monthly Storm Data reports in accordance with
NWSI 10-1605.

NWS Instruction 10-517. Quoted in full.

The misconception this kills
"The LSR is the official record." It is not. It is the fast record, published while the storm is still going, and it gets corrected. The certified record is Storm Data, and the two routinely disagree: surveys downgrade tornadoes, re-rate them, merge two reports into one track, or delete an event that turned out to be straight-line wind. None of that means anybody lied.

Local offices say the same thing on their own event pages. The Mobile office's page for the 10 May 2025 tornadoes, which lesson 2 uses as a worked example, carries this while the surveys were still being revised: "The data provided on this page are considered PRELIMINARY as we continue to analyze all available data." That is a forecast office telling you, in advance, that its own numbers will move.

What is actually in a Local Storm Report

An LSR is a fixed set of fields, and every one of them is something you either supplied or the office had to guess. That is the reason to know the list.

FieldWhere it comes from
TIMEYou. The time it happened, not the time you called
EVENTYou, via what you described
CITY LOCATIONYou. Distance and direction from a known place
LAT/LONYou, or the office deriving it from your description
DATEFollows from the time
MAGNITUDEYou. The size or speed, and whether it was measured or estimated
COUNTY LOCATION/STFollows from the location
SOURCEYou. Trained spotter, public, law enforcement, and so on
REMARKSYou. Everything the other fields cannot hold

The magnitude field carries a tag of Measured, Estimated or Unknown, a format in place since 8 February 2006. Unknown is what you get when you do not say, and it is the weakest of the three.

Why the archive is not a fair sample

Storm Data is built from reports, and reports are made by people, so the archive records where people were as much as where the weather was. This is measured, not suspected.

  • Population. More people means more reports of the same event. Allen and Tippett found this plainly in the hail record.
  • Roads. A tornado crossing a highway is seen. The same tornado two miles into timber is not.
  • Distance from the office. Reporting density falls off with range from the WFO, which is a property of the reporting network rather than of the weather.
  • Night. The single largest gap, for the obvious reason.

Potvin and colleagues put a number on how far this goes: modeling the tornado database, they estimated that in some rural areas only about 45% of tornadoes are reported at all. More than half of them simply are not in the record.

Which is the argument for your report
If the archive were complete, one more report would be a rounding error. It is not complete, and the gaps are systematic and known. A report from a place that does not usually generate reports is worth more than one from a place that does, and the same goes for one made at three in the morning.

What this platform does with it

Every report submitted here is checked against radar and the surrounding data for plausibility. Module 5 said it for hail and wind and module 6 said it for flooding, and the rule is the same one: a report that disagrees with the radar is flagged for a human to look at. It is never automatically thrown away.

The instruction the offices work to reasons the same way. Storm Data quality control reconciles report times against radar signatures, and the directive is explicit about what the archive is for: "the primary purpose of Storm Data is to accurately describe events, regardless of the impact on verification scores." The record is meant to be true, not flattering. Lesson 5 is about what you can do to make yours survive that check.

Check yourself

A Local Storm Report from last night lists an EF2 tornado. Three weeks later Storm Data lists the same event as EF1. What happened?

  1. Normal. The LSR is preliminary and the survey revised the rating for the certified record
  2. The original report was false and has been retracted
  3. Two separate tornadoes were merged, which always lowers the rating
  4. Nothing. Storm Data and the LSR are the same record shown two ways
Hint

One of the five stages is described by its own directive in a single word.

Answer

Normal. The LSR is preliminary and the survey revised the rating for the certified record

NWS Instruction 10-517 states that LSRs are preliminary and that the final report appears in Storm Data. A rating changing between the two is the process working exactly as designed, and it reflects a damage survey rather than any judgment about the person who called it in.

Sources

  1. NWS Instruction 10-517, multi-purpose weather products specification
  2. NWS Instruction 10-1605, Storm Data preparation (current edition 26 July 2021)
  3. NWS Mobile, 10 May 2025 severe weather event and damage surveys
  4. Potvin et al. 2019, Wea. Forecasting 34(1), 15–30: a Bayesian hierarchical modeling framework for correcting reporting bias in the US tornado database
  5. Allen & Tippett 2015, Electron. J. Severe Storms Meteor. 10(3), 1–31: the characteristics of United States hail reports, 1955 to 2014
  6. NWS Service Change Notice 05-37: modified preliminary LSR format, effective 8 February 2006