The Declaration Ledger

I selected FEMA's Disaster Declarations Summaries v2 through the allowlisted Data.gov provider. Data.gov marks the federal dataset public and links the license to the U.S. government-works page. FEMA asks users to follow its OpenFEMA terms and citation guidance. The source lists official FEMA declarations and one row for each designated area and its program flags.

I queried the OpenFEMA API for major disaster declarations dated from January 1, 2000 through December 31, 2025. The four ordered responses contain 30,160 unique area records. Canonical record byte counts and SHA-256 hashes are pinned in the source note; this removes only the API's changing request time and insignificant JSON formatting. A standard-library Python builder verifies every page, the OpenFEMA entity and version, the filter, the schema, all record IDs, and the expected 1,587 disaster numbers before reducing the data. It groups rows by disaster number, combines program flags within each named area, counts declarations once, and retains area names for the selected record. Repeated builds produced the same 1,144,628-byte browser snapshot with SHA-256 `0c6a0692dc76cf9a1edf20b1c0f14e319fe7cb2fa777de7a4a886170a4c4d9f1`; the browser makes no runtime request.

Years use the calendar year of the declaration date, not FEMA's fiscal year. Incident rankings use FEMA's primary incident type even when secondary types are listed. The project excludes emergency and fire-management declarations. FEMA describes the dataset as raw, unedited NEMIS data subject to a small percentage of human error. It also notes that tribal declarations are included within state data; the interface identifies the 62 declarations that FEMA marks as direct Tribal Nation requests. Counts do not measure disaster frequency, severity, damage, population affected, spending, need, response quality, or recovery outcomes. Program flags show authorization by designated area, not applications, approvals, obligations, or dollars.