Case study · built from the live record

Hurricane Isaias, as it happened

Latest advisory 18, Oct 11 03:00Z: 20 kt, Depression. Record read Oct 21 07:01Z.

1. The storm, advisory by advisory

Not in the record: advisories 2, 16. The hurricane worker reads NHC every 30 minutes and missed them; nothing is filled in.

2. State by state

When each state was first put under a watch and a warning, and how many National Weather Service forecast zones it covered. Watch and warning lead times are counted to the first advisory with the center over US land (Oct 10 03:00Z).

3. By FEMA region

4. Calchis’s forecast against what happened

At each advisory Calchis forecasts how strong the storm will get. Isaias peaked at 105 kt (advisory 12). From advisory 3 (Oct 7 09:00Z, 66 hours before the center reached land), the forecast peak was within 15 kt of that. It then overshot as the storm kept strengthening, to 125 kt at advisory 11: the model extends a strengthening trend, the same weakness its validation record shows on Hurricane Milton. Its probability of Category 4 or worse peaked at 23%; Isaias never reached Category 4. Mean error of the forecast peak before the peak: 12 kt.

How this was produced. 3 of 16 forecasts were recorded at the moment they were made; the rest are replayed, not recorded live. The forecast model was changed after Isaias (v2, 10 October 2026): it now reads NHC’s official forecast and no longer stacks the intensification trend on the rapid-intensification jump. The forecast column is the v2 replay, given NHC’s forecast as NHC issued it at each advisory; the v1 column is what this page first published, kept so the change cannot rewrite it. A replayed row runs the same model with only the advisories issued up to that one, the rule its seven-event validation record follows; the NHC forecast track is not given to it, and no learning correction is applied. Since 10 October 2026 every forecast is recorded when it is made, so later storms are scored on forecasts frozen in advance.

5. The wind field against what airports measured

Calchis draws the storm’s wind field from the track and intensity; the same field drives its loss estimates. Here it is held to the winds 16 airports inside its 166 km reach recorded as Isaias came ashore. It read high. The field gave the median station 2.2 times the sustained wind it recorded, and at 12 stations the field’s sustained wind was above the strongest gust the station measured.

9 stations report gusts at least twice the highest sustained wind, against the 1.3–1.5 typical in a hurricane, which suggests gaps in the sustained record at the peak; the gust comparison does not depend on that. Only 1 station recorded tropical-storm-force sustained wind, the rule this check fixed in advance for its summary figure (1.59×): too few for a verdict. On Hurricane Milton the same field read 1.10× at nine stations. Across every US hurricane landfall from 2017 to 2024 it reads 1.32× at airports that recorded 45 kt or more and 1.65× at airports that recorded 20–44 kt, so a high reading away from the core is the field’s known bias rather than something peculiar to this storm; correcting it is open model work. NHC’s Tropical Cyclone Report will replace these observations; this section will be re-run on it.

How this was produced. Observations: National Weather Service API station observations (api.weather.gov), 2026-10-09T12:00Z to 2026-10-11T00:00Z, fetched 2026-10-10 (2-minute average sustained wind (ASOS), peak gust). The field is run once for the coastal crossing taken from the recorded advisories, the same way the Milton check runs it; the station list, the crossing and the summary rule were fixed before any value was read.

6. Estimated damage against what is reported

Calchis estimate, published 2026-10-10: $2.53 billion in residential wind damage (FL $2.24 billion, AL $0.29 billion), from the storm’s wind field over the 2020 housing units in 1,603 census tracts within its reach, valued at Calchis’s default replacement cost and read through the Hazus wind damage curves (model 3.5, the method of the Hurricane Milton benchmark). It was made before any reported figure existed and is not revised afterwards.

Read it with two cautions, stated now rather than after the numbers arrive. It is residential wind only, while the reported totals below include businesses, autos, storm surge and flood, so it should sit below them. And the wind field read high at the airports (section above), so the estimate may read high as well.

What this record cannot tell you yet

• The exact landfall. “First advisory over land” is when an advisory’s position first falls on US land, not the moment the center crossed the coast. NHC’s Tropical Cyclone Report gives that, months later.
• Damage and declarations. FEMA declarations, NOAA Storm Events damage reports (published about 75 days after the month ends) and NFIP claims are added to Calchis’s record as they are published.
• How the rest of the forecast did. The wind field Calchis drew will be held to airport and buoy observations, the storm surge estimate to NOAA tide gauges, and the zones to the damage reports, as each is published, the way Hurricane Milton’s wind field was checked against its stations.
• Which storm a warning was for. NWS tropical products do not name the storm. These are attributed to Isaias by place and time: tropical watches and warnings in Gulf and Atlantic states issued while it was active. No other Atlantic storm was active in that window, so the attribution holds.
• Zones are forecast zones. A count of NWS zones is not a count of counties or people. 2,011 watch and warning versions are in the record.

Sources: National Hurricane Center public advisories, as read by Calchis every 30 minutes; National Weather Service tropical watches and warnings (api.weather.gov), every version kept; US land mask from Census TIGER boundaries; FEMA regions as FEMA publishes them.

See also Hurricane Milton, replayed.