All findings · Storm Blackout and Restoration
Where does Milton rank on the sum of county outage peaks?
Milton's sum of county outage peaks ranks fifth: 3.7 million.
Milton (AL142024) sums to 3,736,269 peak customers across 51 counties (5th largest sum). Each county contributes its own peak.
These numbers are Storm Blackout and Restoration version 1.2, the published package (672 storms in version 1.2).
Table
From the table
| Figure | Value |
|---|---|
| Sum of county peaks | 3,736,269 |
| Counties with a peak | 51 |
| Rank | 5 |
Reading
What the data shows
For AL142024, the sum of numeric peak_customers_out is 3,736,269 across 51 counties. That ranks 5th, behind Irma, Helene, DR-4586, and Ian. Each county contributes its own peak. Those peaks are not one simultaneous count of customers out.
Context
Why it matters
Milton is a Florida landfall in this package. Compare its summed peaks with the other storm events in this table.
Method
How it was measured
The sum is numeric peak_customers_out for AL142024. The rank is that sum's place among event sums in the release utility table, largest first. Method on the dataset page.
Limits
What this cannot tell you
Night light is not a utility outage count. County peaks can double-count customers across overlapping service territories when summed. Disclaimer word for word: Not for real-time emergency response or for any decision about one address or property.
Check
Check it yourself
- File
storms/utility_by_county.csv- Column
peak_customers_out- Row
event_id AL142024, peak_customers_out nonempty, rank 5 by summed peak- API and assistant
- Same key, same data
This API call reads the same cells.
curl "https://api.heimdallresearch.com/v1/storms/AL142024" \
-H "Authorization: Bearer $HEIMDALL_API_KEY"This assistant prompt asks for the same cells.
MCP https://api.heimdallresearch.com/mcp question: What is the sum of peak_customers_out for AL142024, and where does that sum rank?Citation
Dataset citation
Heimdall Research (2026). Storm Blackout and Restoration (Version 1.2) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23010005 Concept DOI https://doi.org/10.5281/zenodo.23010004.
- Version
- 1.2
- DOI
- 10.5281/zenodo.23010005
Notes on the wording and the comparison
The sentence on this page is the owner-picked wording. The marked numbers are filled from the check on this page. A later release stays in its own question file until that version is published. The published package is the version named in the citation on this page. Sorting the named file by the named column reaches the same cells. The comparison uses that published file and leaves later releases out. The data dictionary shipped with the package defines the column. A second person can repeat the aggregation from the file alone. Empty cells stay empty. The dataset page states the sensor and the public record behind the file. Rows left out of the aggregation are described in the method section. The download named in the citation is the file to open. The figure changes only when that published file changes. The same cutoff and the same columns are used for every row that enters the figure. A reader who wants the row list can filter the file with the check line on this page. The published package is the version named in the citation on this page. Sorting the named file by the named column reaches the same cells. The comparison uses that published file and leaves later releases out. The data dictionary shipped with the package defines the column. A second person can repeat the aggregation from the file alone. Empty cells stay empty. The dataset page states the sensor and the public record behind the file. Rows left out of the aggregation are described in the method section. The download named in the citation is the file to open. The figure changes only when that published file changes.
The sentence on this page is the owner-picked wording. The marked numbers are filled from the check on this page. A later release stays in its own question file until that version is published.
The figure is the aggregation named in the check. A different row filter would be a different figure.
A later published version can be set beside this one once that version is the cited release.
Related
Related findings
-
After Maria, night-light brightness took 121 days to return to the usual range in the typical county.
For Maria (AL152017), the median of days_to_normal is 121 days across 46 counties with a value. This column is satellite night-light brightness.
Breakdown
In the county summary table for AL152017, the median of days_to_normal is 121 across 46 counties with a value. days_to_normal counts days of satellite night-light brightness. It is separate from utility hours_to_90pct_restored.
-
After Irma, the typical county reached 90% of customers restored in 53 hours.
For Irma, the median county hours to 90 percent restored is 52.9 hours (280 counties with a numeric value).
Breakdown
For AL112017, the median of numeric hours_to_90pct_restored is 52.9 hours across 280 counties. Blank and non-numeric cells are left out. A row with reporting_drop true stays in the median when the hour cell is numeric.
-
44 of 672 storm events are tropical cyclones or typhoons.
These 672 storm events include 36 with kind tropical and 8 with kind typhoon (44 together). Kind tropical includes hurricanes. Other counts: severe storms 347, floods 186, tornadoes 24, ice storms 21, winter storms 21, snowstorms 20, straight-line wind 6, coastal storms 3.
Breakdown
Counting kind on storms.csv yields the mix above. The three largest summed county peaks are Irma, Helene, and DR-4586. Kind tropical includes hurricanes.
Estimates from public records and satellite night light. Not for real-time emergency response or for any decision about one address or property.
Provided as is, without warranty of any kind.
Source storms/utility_by_county.csv total peak_customers_out 3,736,269.
Source storms/utility_by_county.csv counties 51.
Source storms/utility_by_county.csv rank peak_customers_out 5.