Automated weather reports in court: what survives a challenge

Two Florida courts excluded figures nobody could verify

A weather report survives a challenge when the person citing it can say where every figure came from, how it was made, and which independent records agree with it. A report that cannot answer those three questions is not weaker evidence; in the two rulings below it was no evidence at all, because the witness who relied on it was not allowed to give the figure.

The circuit court for Miami-Dade County, in Hernandez v. Citizens Property Insurance Corporation, struck a policyholder's engineer from testifying to the wind speed at the property on the claimed date of loss. The engineer had taken the speed from a commercial report.

The engineer is precluded from giving any expert testimony regarding the wind speeds at the subject property on the purported date of loss as the [vendor's] report he relies upon and the methodology by which they calculate the data therein cannot be verified, as is required by the court under Daubert, thus deeming the testimony unreliable.

Hernandez v. Citizens Property Insurance Corp., No. 2016-015721-CA-01 (Fla. 11th Cir. Ct. 2021), quoted with the product name omitted

Florida's Fourth District Court of Appeal later reached the same place from the other direction. In Universal Property and Casualty Insurance Company v. Navlen the insureds had won at trial on an expert's wind speed taken from a commercial data product. The appellate court reversed and ordered a new trial.

The only attempt to establish that data's credibility was the expert's assertion that he "uses it all the time."

More is required than an expert stating a test is reliable because they use it.

Universal Property & Casualty Insurance Co. v. Navlen, No. 4D2022-1590 (Fla. 4th DCA, 20 September 2023)

Neither court found the figure wrong; neither was asked to. Each found that the figure arrived without a method anyone in the room could examine, and that an expert's habit of relying on it is not a method.

Rule 702 asks four things of the figure and of the person citing it

In federal court the rule is Rule 702 of the Federal Rules of Evidence. Florida's section 90.702 was rewritten to follow the federal rule's earlier text and asks the same three things of the testimony: sufficient facts or data, reliable principles and methods, and a reliable application of them. Most states have adopted a version of the federal rule; the remainder, New York and Pennsylvania among them, still ask whether the method is generally accepted in its field.

A witness who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise if the proponent demonstrates to the court that it is more likely than not that: (a) the expert's scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue; (b) the testimony is based on sufficient facts or data; (c) the testimony is the product of reliable principles and methods; and (d) the expert's opinion reflects a reliable application of the principles and methods to the facts of the case.

Federal Rule of Evidence 702, as amended 1 December 2023 (the Florida text predates this amendment)

Daubert is the Supreme Court decision that made the trial judge the gatekeeper for this rule. It gave judges a short list of questions to ask of a method, and none of them is about the answer.

• Whether the method can be, and has been, tested.
• Whether it has been submitted to peer review and publication.
• Its known or potential rate of error.
• Whether standards exist that control how it is applied, and whether they are maintained.
• Whether it is generally accepted in the relevant scientific community.

The proprietary figure in Navlen failed that list entirely: no testimony that the data had been tested, reviewed, published or measured for error, and no standard that governed it. Put the same questions to a federal record and they answer themselves. A spotter report is filed under the Weather Service's published directive for preparing storm data; a station reading comes from a maintained, calibrated instrument; a radar-derived size comes from a published, peer-reviewed algorithm whose bias its own authors stated. A report that cites the record inherits those methods. A report that replaces the record with its own computation has to defend the computation itself, to the judge, as a question of admissibility, before the figure reaches a jury.

Independent corroboration is a count of sources, not of copies

The Navlen court noted that the expert offered no independent corroboration for the data he relied on. The phrase is used loosely in claims work and precisely in court: a figure is corroborated when a second record, made by a different method or a different office, agrees with it. The count that matters is the count of independent sources, not the count of documents that repeat one.

Independent

Made by a different office, instrument or method. A spotter's measurement and a radar signature over the same place on the same afternoon are two; a feed that republishes the spotter's line is still one.

Corroborating

Agreeing on the thing being proved: that hail fell, on that day, near that place. Two sources that agree on the day and differ on the size still corroborate occurrence and date.

Not corroboration

A second figure computed from the same radar volume by a second vendor. Two computations from one input are one source with two names.

This is why the free pages on this site count how many independent sources carried each storm day. A day carried by one source is one office's record that it happened. A day carried by three can be checked against itself, and when one of the three is challenged the other two are still standing.

A report has to state its own method

The practical test that falls out of the rulings is whether the document can be cross-examined without its author in the room. Each item below answers a question that will otherwise be put to the witness.

1. The sources searched, named individually, each with the period and radius searched and the address of the record, so the other side can pull the same line from the archive.
2. The sources that found nothing. A report that found no record is only as strong as the list of places it looked. Without that list a clean report is indistinguishable from one that never searched.
3. Measured kept apart from estimated, on every line and in every figure. A document that prints the two in one column has already merged what the court will want separated.
4. Distance and bearing from the address for every record, or the county where the record was filed without a point, so the reader can see which records are about the property and which are about the day.
5. What the document does not claim. Occurrence, date, size and distance are the record's; cause, extent and cost of damage are not, and a report that states no opinion on them has nothing there to be struck.
6. A fingerprint. A checksum of the records cited, the figures they carry and the findings stated, printed on every page, with a public page that recomputes it on request. A copy that has been edited, or that rests on a record its source has since revised, no longer matches.

Four reports for one roof did not agree with each other

A forensic engineer at EDT Engineers ordered four commercial hail history reports for his own house and compared them with each other and with what he had seen from his own driveway in the years he had lived there. The reports were of the common kind: a radar-derived hail size computed for the roof's grid cell, delivered within minutes.

They did not agree. The four documents listed different numbers of hail days, different dates and different sizes for the same roof, and two were nearly identical to each other, which says more about a shared input than about the property. Hail at the sizes they described had left no mark on the soft metal of the house or on the cars outside. His conclusion was that a historical hail report is a general guideline rather than a definitive record of damage, and that the inspection decides what happened to the roof.

The lesson is not that radar is useless. It is that a figure computed from one input by one vendor is a single unverified estimate, and four of them are not four sources.

The federal record can be certified

The last question a court asks of a document is whether it is what it claims to be. For the federal weather record the answer predates any vendor. The National Centers for Environmental Information certifies data held in its archive, under the Department of Commerce's seal, specifically to meet the authentication requirements for records submitted as evidence, and will send the certified copy to a courthouse or a subpoena room on request. Climatological observations, cooperative observer records and certain radar products can be certified this way.

Under the federal rules a record of a public office setting out matters observed under a legal duty to report is excepted from the hearsay rule, and a certified copy of a public record authenticates itself without a witness. That is the footing a weather report should stand on: it cites records in an archive a court already knows how to admit, and gives the reader the address of each so the certified copy can be ordered. A report is not itself a certified record and should not pretend to be; its job is to be a faithful, checkable index of records that can be. A figure that exists only inside a vendor's computation cannot be certified by anyone, which is where the experts in Hernandez and Navlen were left.

What the report shows

A Titan Weather report for an address lists every record it found, with its source, its date and time as the source filed them, its size or speed in the source's own units, and its distance and bearing from the address. Records within 3 miles are banded as primary and records within 10 miles as contributing. Every measured size is set apart from every radar-derived size, in the text, the tables and the map, and the two are never added together.

The sources and method section names each source searched with the period, the radius and the public address of the record, and a source that found nothing is listed as searched and empty. The statement of compilation says that records were reproduced as their sources published them, that estimated values are labelled as such wherever they appear, and that no opinion is offered on the cause, extent or cost of any damage. The fingerprint is printed on every page and can be recomputed by anyone holding a copy.

Questions this guide answers

Are automated weather reports admissible in court

Only when the person citing one can say where each figure came from, how it was made and which independent records agree with it. Two Florida courts excluded wind figures from commercial reports because nobody in the room could verify the method, not because the figures were shown to be wrong.

What does Rule 702 ask of a weather figure

That the testimony rests on sufficient facts or data, on reliable principles and methods, and on a reliable application of them to the case. A figure that cites the federal record inherits published methods; a figure computed privately has to defend its own.

What counts as independent corroboration of a storm day

A second record made by a different office, instrument or method that agrees on the day and the place. Two documents that republish one spotter report are one source; a second computation from the same radar volume is one source with two names.

Can the federal weather record be certified for evidence

Yes. The National Centers for Environmental Information certifies data from its archive under the Department of Commerce seal for legal proceedings, and a certified public record authenticates itself under the federal rules. A report that cites the record gives the reader the address to order the copy from.