A BeyondClaim report is built the way a certified meteorologist would build one: the primary radar and observation products, matched to the property's parcel boundary, with every figure traced to the product, station and time it came from. This page states exactly what is read, what creates an event, how confidence is scored and how hail is calibrated.
Every number here was verified against the detector code and production data on September 5, 2026 (methodology statement v1.1.0) and is reviewed whenever a detector changes.
What is read, what makes an event, what the report flags.
Detection and reporting are different thresholds on purpose. The archive keeps every signal the products emit; the report filters to what matters for a claim and says which filter it used.
Hail
What we read
NEXRAD Level III hail detection (via NOAA SWDI) joined with MRMS MESH sampled at the storm cell and at the property; NWS Local Storm Reports and NCEI Storm Events for ground truth. Between radar publication and ground review, an hourly MESH sweep covers the gap.
What creates an event
Any radar hail signature within 25 km of the location, any NWS hail report, or MESH of 0.50" or more at the property. There is no size floor on detection; the report filters, the archive does not.
What the report flags
1.00" or larger at the roof, on the larger of the calibrated radar size and a measured ground report. Tiers: marginal 0.50", severe 1.00", damaging 1.50", destructive 2.50", catastrophic 4.00".
Confidence label
High at 85, Medium at 60
Wind
What we read
NWS Local Storm Reports as the trigger, with the measured peak gust from the nearest ASOS station and NCEI Storm Events. The headline gust is the best-measured source available, in that order.
What creates an event
A reported gust of 58 mph or more. Narrative-only reports are kept without a speed.
What the report flags
58 mph and above. Tiers at 58, 74, 96 and 111 mph.
Confidence label
High at 80, Medium at 60
Tornado
What we read
NEXRAD Level III tornadic vortex signatures (via NOAA SWDI), NWS tornado reports, and the surveyed EF rating, path and duration from NCEI Storm Events.
What creates an event
Any vortex signature or NWS tornado report. A radar-only signature is labeled EFU and shown separately from a surveyed tornado; it never inherits a rating.
What the report flags
Any surveyed EF rating. Radar-only signatures are excluded once the severity floor is set to severe.
Confidence label
High at 85, Medium at 65. A radar-only signature can never score High.
Flood
What we read
Daily precipitation at the nearest ASOS station, USGS stream gauge stage against flood stage, and NCEI Storm Events flood records.
What creates an event
A daily total of 2.00" or more, a gauge at or above flood stage, any NCEI flood record, or a peak hour of 1.00" or more (tagged flash).
What the report flags
Data-limited: the report shows what was recorded and says when nothing was.
Confidence label
High at 80, Medium at 60
Winter
What we read
Hourly temperature, freeze-thaw cycles, hours of ice accretion and snow depth at the nearest ASOS station, plus NCEI Storm Events winter records.
What creates an event
A minimum temperature below 32°F or any NCEI winter record. Severity is set from temperature; ice and snow amounts are recorded from NCEI narratives but do not tier the event.
What the report flags
0.25" of ice or more when an accumulation was recorded; otherwise data-limited.
Confidence label
High at 80, Medium at 60
Hurricane
What we read
NHC best track (HURDAT2) for past seasons and live NHC advisories in season, with storm-surge and hurricane records from NCEI Storm Events and ASOS wind and rain.
What creates an event
A track within 300 km of the location, with category stepped down one tier per 100 km beyond 50 km, or an NCEI record. The floor is tropical depression.
What the report flags
Category 2 and above.
Confidence label
High at 80, Medium at 60. Live advisory rows score Medium at most until the season's best track is published.
Confidence
A point ledger you can read, kept apart from corroboration.
Each event's confidence score is an additive ledger: points for the source that found it, the size or speed, the strength of the radar signature, agreement between products, and distance from the location. For hail, wind and tornado the ledger itself is stored with the event, so the score can be audited line by line.
The label cutoffs are set per peril, because the perils do not score alike: High from 80 or 85 and Medium from 60 or 65, as listed in each card above. The score is not a probability and it does not measure whether a human confirmed the event. That is a separate column on every row: NWS-corroborated, from a Local Storm Report, an NCEI record, a gauge or a track, or radar-only.
Severe-hail probability and peak reflectivity, up to +30
AGREEMENT
MESH at the cell and an NWS match +15
DISTANCE
Within 5 km +5 · beyond 20 km −10
LABEL
High 85 · Medium 60 · Low below
Hail calibration
Radar estimates hail aloft. The report sizes it at the roof.
Raw
MRMS MESH is an estimate of hail aloft and tends to overstate what reaches the ground. The raw value is stored with every event and printed on the report beside the calibrated size, never replaced.
At the roof
A piecewise curve fitted to 2,000 NWS hail reports across Texas and cross-checked against 91 meteorologist-sized roofs, then adjusted for the melting layer from the RAP wet-bulb-zero height on the day. Identity below 1.00" raw; a typical, median-unbiased size above it.
Elsewhere
For storm-cell history away from a specific roof, a radar-site and season factor. It is a literature-range setting, not a fitted curve, and the report says so.
Every report prints its calibration version. Current: v4-roofcurve-2026-09-03. A calibration change re-renders the whole archive on the next read, so two reports for the same roof never disagree because of when they were run.
Evidence chain
What is stored with every event.
The raw radar size at the exact cell, the 5-by-5 neighbourhood and the property, with the sample time and the MRMS product version.
The radar signature: probability of hail and of severe hail, peak reflectivity, vertically integrated liquid, azimuth and range from the station.
Hail kinetic energy and the per-scan profile at the property, where the scans were captured.
The ground truth matched to the event: NWS report size and narrative, NCEI event id, size, damage estimate and distance.
The melt assessment: wet-bulb-zero and freezing-level heights on the day, and the resulting melt risk.
The confidence ledger, so the score can be recomputed by hand.
Not stored, and never claimed: polarimetric radar moments and storm motion. Reports do not cite them.
Coverage
Texas, 2021 to today, audited day by day.
Radar mosaic coverage is continuous for the archive: every UTC day since January 1, 2021 checked, no missing days, one partial day on July 19, 2021. Twelve NEXRAD sites sit inside Texas and border-state sites serve the edges by nearest-station routing.
Absence of a detection is not evidence of absence. Same-day radar products publish one to two days late, and NWS storm reports are reviewed for two to three months before they are final. A live event is marked as reported until that review lands, then confirmed in place.
NEXRAD Level III hail and tornado products via NOAA SWDI, MRMS MESH, NWS Local Storm Reports, NCEI Storm Events, ASOS station observations, USGS stream gauges, and NHC best track and advisories, with RAP model soundings for the hail melt assessment. Every figure in a report names the product, the station and the time it came from. Nothing is modeled from a forecast.
What is MESH, and why is the size on my report different from the swath map?
MESH is the Maximum Estimated Size of Hail from radar, an estimate of hail aloft. It tends to overstate what reaches the ground, so the size on a BeyondClaim report is calibrated: at the roof, a curve fitted to Texas ground reports and adjusted for the melting layer; elsewhere, a radar-site and season factor. A swath map shows the raw radar estimate. Both the raw and the calibrated figure are on the report, with the calibration version.
Why can a confirmed tornado show Medium confidence?
Because confidence and corroboration are different things, and the report keeps them apart. The confidence score measures how strong the radar signature and the matching evidence are. Whether the event is backed by an NWS report or a survey is shown separately, on every row. A surveyed EF0 tornado with a short path can score Medium while being fully confirmed; a strong radar-only signature can score higher while being unconfirmed. Read the corroboration column first.
What does none detected mean?
That no product in the archive placed the peril at the property on that date, and the report shows the nearest detection and how far away it was. It is a defensible result and it is never inflated. It is also not proof of absence: same-day radar products publish one to two days late and NWS storm reports are reviewed for two to three months, so a recent date can fill in later.
Is the radar archive complete?
Texas, 2021 to today. The MRMS mosaic has been audited day by day since January 1, 2021: no missing days and one partial day, July 19, 2021. Because MRMS blends many radars, a single station outage degrades a day rather than blanking it. Twelve NEXRAD sites sit in Texas and border-state sites are used by nearest-station routing.
Does the report say whether a roof is damaged?
No. A report verifies that the weather occurred at the lot and how severe it was there. It never claims damage, never estimates a loss, and never infers cause. Those remain the inspection's job, and the report is built to sit beside it.
Statement v1.1.0, verified September 5, 2026. Reviewed against the detector code whenever a threshold, product, confidence rule or calibration changes.