Methods and sources

What a packet does

A packet gathers the federal data a risk assessment starts from; the plan team adds local knowledge.

• Show the plan clock for every jurisdiction FEMA lists in the county's plan: status, approval date, expiry, days to or since expiry, and whether an update is underway.
• List the 18 National Risk Index hazards, which apply and which NRI marks "Not applicable", ranked by modeled expected annual loss, with a list of hazards NRI does not cover (local data needed).
• List federal disaster and emergency declarations with county Public Assistance dollars, flag those new since the last plan's approval, and state "no recorded events" for hazards with neither a Storm Events record nor a federal declaration.
• Summarize NOAA Storm Events records: the scale, observed range and worst recorded event per hazard, and historical frequency by intensity (tornado rating, hail size, wind speed).
• Give FEMA's modeled annual frequency per hazard, with a probability descriptor from a scale defined in the packet. For earthquake, wildfire, coastal flooding and tsunami, FEMA's value is the chance at a location in the county, not how often the hazard affects the county, so it is labeled that way, gets no "1 in N years", and the descriptor comes from Storm Events; with no Storm Events record (always for earthquake) the same scale is applied to the chance at a location, and the county as a whole is at least as likely.
• Count repetitive-loss and severe repetitive-loss flood properties for the county and each jurisdiction's NFIP community (counts only).
• Offer a county-level starter table for assets and impacts, and a checklist of what to gather locally.
• Carry a citation line on every table for the plan's bibliography, and a "What changed since your last plan" box.

FEMA's Local Mitigation Planning Handbook names the same federal sources used here: NOAA Storm Events, OpenFEMA declarations and the National Risk Index.

On format, FEMA's Local Mitigation Planning Policy Guide (2025) says: There is no prescribed method for how to present this information (p. 18) and FEMA does not require any specific format (p. 13).

FEMA's review tool and its Element B items

FEMA reviews every local hazard mitigation plan with the Local Mitigation Plan Review Tool, Appendix A of the 2025 Local Mitigation Planning Policy Guide. The tool lists the requirements a plan must meet, grouped in elements. Element B is the risk assessment. It has nine items, and each packet section says which one its tables support ("Supports FEMA checklist item B1-a: hazard identification").

• B1-a, hazard identification: every natural hazard that can affect the planning area, with a reason for any commonly recognized hazard left out.
• B1-b, hazard location: where each hazard can occur.
• B1-c, hazard extent: how severe each hazard can be (a scale and the range of intensities).
• B1-d, previous occurrences: past events for each hazard, at least the state and federal disaster declarations since the last plan, and a statement for any hazard with none.
• B1-e, probability of future events: how likely each hazard is, with any descriptor defined.
• B1-f, hazards that vary across jurisdictions: in a plan with several jurisdictions, the risks that differ between them.
• B2-a, vulnerability: a summary of each jurisdiction's vulnerability, including its people and assets.
• B2-b, potential impacts: the likely effects of each hazard on each jurisdiction.
• B2-c, repetitive loss properties: the number and types of NFIP-insured properties flooded again and again.

Other items the packets mention: A4-a (the plan's sources) and E1-a (changes in development since the last plan). Georgia packets also show the letter of the GEMA local plan template next to each item.

How to read the labels

Automated baseline

Built automatically; passed the pipeline's quality checks. 3,456 of the 3,488 packets in this release carry this label; the other 32 were also checked line by line against the live sources.

Checked against live sources on (date)

On that date every number in the packet was compared line by line with the live FEMA and NOAA data by an automated review, a separate review rechecked the result, and every issue found was fixed or explained. Its page says: "This packet was checked line by line against live FEMA and NOAA sources on (date)." Release 2026-10 files published before the 2026-10-01 correction carried the earlier wording of this label, "Hand-checked on (date)"; it means the same check.

Supports FEMA checklist item B1-a (and the other items)

The FEMA review item the section's tables are evidence for. The plan and FEMA's review decide whether the item is met.

Local data needed

Information federal data do not hold, which the plan adds from local sources. The packet names it; for B2-a and B2-b it also gives a county-level starter table and a checklist.

Modeled

A National Risk Index value from FEMA's model, not a count of events. "FEMA annualized frequency (modeled)" is not events divided by years.

Zone-level

A Storm Events record filed for a National Weather Service forecast zone, not a county. The zone may cover more than one county, so the event may have happened elsewhere in the zone. Zone damage is never added up across counties.

County-level (Vermont towns)

Town packets have their own plan clock and repeat-flood counts; hazard, probability and declaration tables come from the county and say so.

New since last plan and in the 12 months before plan approval

A declaration after the current plan's approval date is new. One in the 12 months before approval is marked separately, because plans are usually drafted before that window.

Nominal dollars

Dollars as reported, labeled with their year; not adjusted for inflation.

How the plan clock is set

• Each jurisdiction is keyed by its type and Census code, so a school district and a city with the same number stay apart.
• The current plan is the Approved or Amending plan with the latest expiration date. A past expiration date wins over an "Approved" label (the packet notes "FEMA status not yet updated"). Without an approved plan, the latest expired plan is shown; without either, "no approved plan on record".
• Approvable Pending Adoption is FEMA's status for a jurisdiction that has not yet adopted its plan. That plan is not yet in effect for the jurisdiction and is never its current plan, even when FEMA has approved the plan itself (the packet then says "approved, pending adoption here").
• A newer plan FEMA's tracker shows as in progress, in review, awaiting revisions or pending adoption is shown as "update underway". The status is FEMA's tracker entry as of 2026-09-30.
• When a jurisdiction is in more than one approved plan, the packet uses the one that expires last and names the others. FEMA's plan status map shows one plan per jurisdiction and may show a different one.
• A plan past its expiration date is expired. Until a new plan is approved, FEMA mitigation project grants (HMGP, FMA) need a FEMA exception (44 CFR 201.6(a)(3)).
• The home map groups county areas by their own plan row: expired, not yet adopted or never covered (a plan FEMA approved or found approvable but not adopted here counts in this group), expiring within 12 months, 12 to 24 months, beyond 24 months, or absent from FEMA plan data (common in New England, where towns hold the plans).

Data-error screens

Values outside physical limits are screened: the packet leaves the value out, says so, and never substitutes a corrected figure; the record still counts as an occurrence unless it is an exact duplicate.

• Thunderstorm wind above 130 knots is not treated as a speed (likely a data entry error, such as a hail size typed into the wind field). It is left out of the worst event, the observed range and the intensity bands, and the county's wind line says a value was left out. One Oklahoma record (Canadian County, 2011-05-24) is a real gust measured as the El Reno tornado passed; it is left out as a tornado wind, with that reason.
• Tornado path length and width. Lengths above 220 miles, 200 miles or more from 1996 on, or (from 1996 on) 50 miles or more when that is over three times the straight-line distance between the recorded start and end points plus 20 miles, print as "length not shown: implausible value in source". Widths above 4,600 yards likewise. A length of 0 is not printed. Lengths never rank tornadoes.
• Duplicate tornado records (same county coding, rating, date, time, places and coordinates, with length and width identical or swapped) count once.
• Zone records cannot be a county's worst. For tornado, hail and thunderstorm wind, a record that reached a county only through a forecast zone covering more than one county is counted, but cannot be the county's worst event or set its observed range, because it may have happened in another county. When no other record exists, the packet says so.
• Hail size. Sizes under 0.20 inch are treated as not measured; the placeholder 9.99 is dropped; sizes of 8 inches or more stay as recorded with a note that 8 inches is about the U.S. record.
• Damage values with no unit. A damage value NOAA recorded as a number with no unit letter (for example "06") is printed as recorded and left out of damage sums and counts. Property damage before 1996 is printed as NOAA's value from its damage ranges. One record (East Carroll Parish, Louisiana, 2009-05-10, $700 million) is printed with the note that its narrative describes local damage.
• Blank fields stored as zero. Before 1996 the Storm Events file stores 0 for tornado length and width on wind and hail rows; they are read as blank. Damage entered as "0T" or "0?" is read as not reported, not zero.
• County codes that are zones. A record coded to a county whose place name is really a forecast zone name is placed as a zone record. A county-coded record whose name names a different county, or whose code is not a county, is left out of county counts (mostly Puerto Rico records from 2000 to 2008).
• Impossible plan dates (such as the years 1025 and 3025) are not shown. A jurisdiction's approval date more than 5 years from the plan's approval is shown with a note that it may belong to an earlier plan.

Source quirks shown as recorded

• Tornado records printed as published: two Brazos County, Texas records of 1995-05-18 (both count; their begin locations differ), and a Berkshire County, Massachusetts F4 of 1995-05-29 whose damage is published as $250 although the event's documented loss was far larger. No value is substituted.
• Zone records from 1996 to 2019 whose zone is in no known National Weather Service zone file cannot be placed in a county and are not counted. An older record is placed by its zone number only when its zone name matches the zone file; records from zones the Weather Service later renumbered (Hawaii, Alaska, Colorado, Oregon, Washington, Wyoming and others) are placed by name from a reviewed list, or left out. Zone files are matched by date from 2020 on.
• Guam, the Northern Mariana Islands and American Samoa records are placed by island name, not by zone number. Records for Palau, the Federated States of Micronesia, the Marshall Islands and Wake Island, which Storm Events files under Guam, are left out. The 33 records Storm Events files for the "Northern Marianas" area (the Commonwealth outside Saipan and Rota; typhoons including Super Typhoon Yutu of 2018-10-24, a volcanic ash fall and two rip currents) are placed in the Northern Islands Municipality only, because a record for the whole area cannot be told apart from one about the northern islands. The Northern Islands Municipality is not a National Risk Index county area, so these records appear in no packet, and Tinian has no Storm Events record.
• Older zone names that list several counties ("Sedgwick and Phillips Counties", Colorado; "Stillwater/Carbon", Montana; "Catalina and Santa Barbara Islands", California; and 48 other older names in the West, Virginia and Puerto Rico) are placed in every county the name lists, not only in the county of the zone that now carries the number.
• Augusta and Albemarle counties, Virginia are missing from the Weather Service zones named after them; Hazard Plain adds them back (and Indiana County, Pennsylvania, which two zone files list under a mistyped code). Every zone file lists the Colorado zone "Saguache County East of Continental Divide" with Rio Grande County under the correct spelling and Saguache County under a misspelling ("Saquache"); Hazard Plain places it in Saguache County only.
• Kalawao County, Hawaii has no Storm Events record at all; its "no recorded events" lines say why.
• Per-location descriptors. Where NRI gives coastal flooding, wildfire, tsunami or earthquake no expected annual loss in a county, no frequency descriptor is assigned for it, even though NRI publishes a small chance at a location.
• Lightning in the island territories: NRI publishes no lightning frequency for American Samoa, Guam, the Northern Mariana Islands and the U.S. Virgin Islands, so no descriptor is given.
• Connecticut: plans and declarations use the former counties; NRI uses the nine planning regions. Declarations are listed on every overlapping region, labeled "former county". A few Connecticut flood properties match no town and are reported at state level.
• FEMA plan labels that lag: some plans still labeled Approved have passed their expiration date; the date is used.

Method details

The packets keep short notes; the full text of these rules is here.

Probability: FEMA annualized frequency

FEMA annualized frequency (modeled) comes from the National Risk Index v1.20 (December 2025). It is not recorded events divided by years, and what it measures depends on the hazard. For most hazards it estimates how often the hazard affects this county each year: it counts events or event-days (drought, heat wave, cold wave, winter weather, ice storm and inland flooding are in days per year; lightning is the number of days per year with a recorded cloud-to-ground strike, from the NOAA Storm Prediction Center daily lightning climatology); it may count events in the surrounding area (hail, strong wind, tornado, hurricane and ice storm use a 49 km grid and buffers) and is averaged over the county's area; and some hazards use a minimum value for places with no recorded event. Earthquake, wildfire and coastal flooding are different: FEMA computes them for each census block and averages them over the county, so the value is the chance (for coastal flooding, the modeled number of floods per year) at a location in the county, which is lower than the chance that the hazard strikes somewhere in the county. Earthquake is the chance of at least minor-damage shaking (U.S. Geological Survey model); wildfire is the chance that a location burns in a large fire (U.S. Forest Service burn probability); coastal flooding comes from FEMA flood zones and NOAA high-tide flooding, so it can be above 1 where high-tide flooding is frequent. Tsunami is the value at the county's most exposed location, from recorded events and modeled return periods. For these four hazards the packet gives no "1 in N years" and takes the descriptor from Storm Events unless the value is 1 or more a year; with no Storm Events record (always for earthquake) it applies the same scale to the chance at a location, which the county as a whole at least reaches. Landslide is a U.S. Geological Survey modeled county rate spread by slope susceptibility. Volcanic activity is the chance that a volcano affecting the county erupts in a year. Drought counts only extreme and exceptional drought (D3 and D4). Source: FEMA, National Risk Index Technical Documentation v1.20, sections 5.2.2 (Table 5), 5.2.3, 7.4, 10.4, 16.5, 17.1, 17.3.1, 17.5, 20.3.1, 21.4, 22.4 and chapters 6 to 23.

Storm Events coverage and zone matching

Storm Events records are more complete after 1996; before 1996 only Tornado (from 1950), Hail and Thunderstorm Wind (from 1955) exist. Zone-level records are placed with NWS zone-county files, which exist from 2020 on; older, renumbered and unmatched zone records are handled as described under Source quirks. Zone records whose zone is in no NWS zone file are not counted; this applies to records from 2020 on as well. Records before 2020 use the zone map of the earliest available NWS file; NWS has since redrawn some zones. A record from a zone in no known file whose zone name is the name of exactly one county of its state (for example "Pocahontas", West Virginia) is placed in that county as a zone-level record; names that a county shares with an independent city (Baltimore, Richmond, Fairfax, Franklin, Roanoke) are not placed. Within a county, zone records that share a NOAA episode, event type and begin date count as one record (for example a tropical storm logged for both the Coastal Levy and Inland Levy zones of Levy County, Florida); the record kept is a county-coded record of the same episode, type and day if there is one, else the strongest zone record; a county-coded record is merged only when it repeats another tornado record of the county exactly (same path, swapped or copied length and width). The most damaging record in B1-c reads every record, merged or not, and zone damage is never added across counties.

Places named by Storm Events records

A county-coded record names a place when NOAA's begin or end location is the place's name. NOAA gives many locations as a distance and direction from a reference place ("76.69 miles ENE of Keysville"); the record still counts for the place it names, and the packet prints NOAA's distance and direction beside every place a record names, and on each jurisdiction's row how many of its records are at a distance and how far.

Public Assistance matched to jurisdictions

FEMA files Public Assistance by applicant name. A row counts for a jurisdiction only when its name and kind identify exactly one jurisdiction; FEMA's county field stays the basis of every county total. Beyond the basic forms, these name forms count: a county's own operating departments filed under the county's name (highway, roads, secondary roads, road and bridge, public works, engineer, emergency management, emergency services, parks, facilities; not the sheriff, clerk, auditor, treasurer, coroner, health departments, boards, commissions, districts, authorities or utilities), with a note on the county's row; FEMA's slash form "applicant / sub-unit" when the applicant before the slash has an explicit kind ("Wadena, City of / Police") and the sub-unit is not one of those separate offices or bodies or another government; Kansas USD and rural water district numbers, and a "Schools" form when one school district of that name is in the county (also when only one of the two names ends in a district number); a Massachusetts city that Census names "X Town city" for "Town of X" or "City of X"; a township filed as "X Township, Y County" under its own county Y; "X, City of" for a city whose own name ends in City (Lake Havasu City) when the state has no place named X; three reviewed names of one consolidated government (San Francisco, Anchorage, Opa-locka); four Nebraska school districts' legal names that their NCES state district IDs prove ("Douglas County School District 0001" is Omaha Public Schools); a town or township that Census records as consolidated with the place of the same name and land area (functional status C: "Derby, Town of" is Derby city); and Connecticut and St. Louis rows filed with no county. A row FEMA files under one county that names, by name and kind, exactly one jurisdiction of another county is added to that jurisdiction's row with a note naming the county FEMA used. A bare name that two or more jurisdictions of a county share ("Woodstock" for a town and a village) counts for none of them; each of their rows says how much FEMA files under that name. Department names that start with a city's or town's name but carry no applicant form FEMA uses for the government itself (for example "Omaha Public Works") are not counted; the jurisdiction's row gives their total. In the District of Columbia and Guam, Public Assistance FEMA files as Statewide or with no county is the unit's own; in American Samoa, the Northern Mariana Islands and the U.S. Virgin Islands it is not split by island, and each declaration row says how much FEMA files territory-wide.

Hazards of a declaration

FEMA records a declaration's hazards in its incident type, in its designated incident type codes (from 2023 on) and in its title. In Severe Storm and Severe Winter Storm titles, "winds" maps to Strong Wind and "heat" to Heat Wave. When the codes are present, a hazard the title names that the codes do not is added as "also designated (named in FEMA's title)", so declarations with the same title map the same way before and after 2023.

Tornado path length

From 1996 on, tornado records are per-county segments and length is the path length within the county (NWS Instruction 10-1605). Records before 1996 may give the whole path length, including the path outside the county, and before 1996 a tornado is listed once, under the county where it began, with the whole path's rating; counties farther along the path may not list it. Length is not used to rank tornadoes.

Strongest recorded tropical storm

For hurricanes and tropical storms the packet names the strongest record: records with damage, a wind speed or direct casualties come first, then the storm type, category and wind. Storm Events files every zone of a storm's episode under the storm's event type and category, so a hurricane record (NOAA event type "Hurricane (Typhoon)"), especially an inland or remnant one, does not by itself mean hurricane-force winds in the county.

Hurricane records NOAA filed under other types

NOAA sometimes files a tropical cyclone's damage under another event type: in Florida in 2004, Hurricanes Charley, Frances and Jeanne are High Wind and Strong Wind records, and Hurricane Harvey's 2017 Houston-area flooding is Flash Flood records. NOAA's type is kept. A High Wind, Strong Wind, Thunderstorm Wind, Heavy Rain, Flood, Flash Flood, Lakeshore Flood, Coastal Flood or Storm Surge/Tide record is also counted under Hurricane when the record's event or episode narrative names a storm in NOAA's HURDAT2 best-track list for that year (Atlantic, Northeast and North Central Pacific) and the record begins between 2 days before the storm's first best-track position and 5 days after its last. The episode narrative describes the whole weather episode and is shared by every record of it, so most matches are in the episode narrative. The narrative names the storm in one of these forms (any capitalization): "Hurricane", "Tropical Storm", "TS" or "T.S.", "Tropical Depression", "Tropical Cyclone", "Post-Tropical Cyclone", "Subtropical" or "Extratropical Storm", "Hurricane/Tropical Storm" or "remnants of" before the name; or the name alone, only when that same narrative also says "tropical", "hurricane", "remnant" or "landfall", and not when the name is part of a place name (Florence County, Port Arthur) or the record's own county or location name. For years after the newest HURDAT2 season, only the forms with words before the name count, and the name must be a HURDAT2 storm name of six years earlier, since NHC's name lists repeat every six years. These records are matched by name only, with no track dates yet, and the packet's note for the row says so. Such a record is labeled with NOAA's type and the storm as the narrative names it ("NOAA type High Wind; narrative names Hurricane Charley"), is listed in the Hurricane extent row "Wind, rain, flood or surge records whose NOAA event or episode narrative names a tropical cyclone", and counts once in every total over all hazards. A few records name the storm only as context (for example a thunderstorm "between a cold front and Hurricane Arthur off the Delaware coast"); they are counted by the same rule, so read the narrative before attributing a record's damage to the storm (the packet's Hurricane row note and Storm Events history say this too). Misspelled names, West Pacific storms (not in HURDAT2) and storms recalled from earlier in a season do not count.

Expected annual loss far above the neighboring counties' median

The packet copies FEMA's National Risk Index expected annual loss as published and ranks hazards by it. Where the figure is far above the neighboring counties' median and NOAA's records, the value and its rank stay as they are and a caveat is printed beside it (B1-a, the summary on page 1 and the site page). All of these must hold: the hazard is one Storm Events records as such (not earthquake, landslide, lightning, tsunami, volcanic activity or wildfire); the hazard is not drought, since Storm Events records drought damage poorly and its drought history is no test of FEMA's figure (heat wave and cold wave stay in, since NOAA records heat and cold deaths); the loss is at least $100,000 a year; it is at least 10 times the median of the neighboring counties that have a loss for the hazard (at least 2 of them; Census county adjacency file, 2024); its loss per dollar exposed is at least 3 times their median, so a county that is only larger is not flagged; and every NOAA Storm Events record of the hazard placed in the county, including same-day records of the same episode that the history merges into one, shows no direct deaths or injuries, with recorded damage below 10% of one year of the modeled loss. The caveat says what it counts: the records counted under the hazard (and how many of them are NOAA types outside FEMA's definition, such as Frost/Freeze under Cold Wave), the count from 1996 on when the records start earlier (the period of the B1-a records column), and any merged same-day records. The NOAA test also reads zone records that no NWS zone file places when the zone name names the county (for example "KERN CTY MTNS"); they are not counted anywhere else. Such a record does not count for a county when its NOAA narrative places it only in other counties, by county name or by a town that lies in one other county: the event narrative's sentences on the death, injury or damage are read first, then the places the rest of the event narrative names, and the episode narrative's sentences on the loss only when the event narrative names no place at all (a 2005 Utah avalanche record whose zone name lists twelve counties, and whose narrative places both deaths in Sanpete County, does not count for the other eleven).

B1-f hazards by jurisdiction table

Under B1-f, a packet whose plan lists cities, towns, townships or the county has a table of the National Risk Index hazards by jurisdiction. It is a draft from federal data for the planning team to confirm and trim. A row covers every 2020 census tract that holds part of the jurisdiction's area in this county (land or water). The tracts come from the Census 2020 block assignment files: each block code carries its tract, and the same files give the block's county subdivision and place (for townships in Illinois, Kansas, Nebraska, North Dakota and Ohio, the county subdivision comes from the block records of the 2020 redistricting data).

A cell has one of three states. X: at least one of the row's tracts has an NRI rating from Very Low to Very High for the hazard. o: none is rated, but at least one is NRI "No Rating", meaning NRI models no expected annual loss there (no exposed buildings, people or crops, or a modeled frequency of zero). That is not the same as "cannot occur": NRI counts drought losses to crops only and models drought frequency from extreme and exceptional (D3 and D4) drought weeks since 2000, so a town with little farmland, or with no D3 to D4 drought on record, reads o for drought. Blank: NRI rates the hazard Not Applicable or Insufficient Data in every tract of the row. That is not proof the hazard cannot occur either: Insufficient Data means NRI lacks the data to rate it, not that it is absent. Hazards that are blank in every tract of the county are left out of the table and named under it.

A city that crosses a county line counts only its area in this county. The county's own row covers the area in no incorporated place and in no town listed in the table (or the whole county when there is no such area). Special and school districts take the row of the town or place their name contains, and the team confirms the service area; districts whose name contains no listed town or place are named under the table. Census tracts are coarse, so in rural counties several towns share a tract and show the same row. Connecticut: NRI keys tracts by planning region, and a block's region comes from its town. No table is built for American Samoa, Guam, the Northern Mariana Islands or the U.S. Virgin Islands, for which the Census publishes no 2020 block assignment file.

Repetitive loss: the OpenFEMA fields

The FMA counts use OpenFEMA's fmaRl and fmaSrl flags, which follow the Flood Mitigation Assistance definitions of 44 CFR 77.2 (the ones the Local Mitigation Planning Policy Guide cites); the field description of fmaRl leaves out the Increased Cost of Compliance clause of 44 CFR 77.2(i). The NFIP counts use nfipRl and nfipSrl, the insurance definitions (repetitive loss: 2 or more claims over $1,000 in 10 years), which can change to false once a property is mitigated. FEMA's NFIP Multiple Loss Properties file carries its own data date, which can be a few days before the file's version date.

Scope of the federal data

• The plan team writes the plan: its narrative, mitigation strategy and goals. Items marked "Local data needed" are filled from the community's own records.
• Assets, impacts and vulnerability for each jurisdiction (B2-a, B2-b) come from local data; the county-level starter table gives county figures only.
• Dollars are nominal (not adjusted for inflation). The packet estimates no losses for individual buildings and models no future conditions.
• State-declared emergencies are not in federal data (there is no national source); each packet names them as local data needed and points to the state.
• National Risk Index values are modeled and fixed at version 1.20 (December 2025). They describe the county as a whole; the tract range shows spread, not location.
• Storm Events holds the events the National Weather Service recorded; before 1996 it holds only tornado, thunderstorm wind and hail. Damage figures are the Weather Service's estimates and are often blank.
• Drought extent uses National Weather Service drought reports, not U.S. Drought Monitor county statistics. Earthquake and other hazards without a Storm Events measure name the local data needed.
• FEMA's plan tracker is updated on FEMA's schedule; some "plan in progress" records carry no date.
• Public Assistance for a jurisdiction is matched by applicant name; an applicant whose name does not identify one jurisdiction counts in the county total only.
• State plans, tribal plans, statewide Public Assistance applicants and tribal declarations are on each state page and in the state files (state.xlsx and state.json in the state folder and the state ZIP), not on county pages. Public Assistance that FEMA files under a county field that is not a county of the state (for example an Alaska school area) is placed by applicant where the name identifies one place; the rest is listed on the state page. Public Assistance rows that FEMA files under a tribal government in place of a state (for example Fort Mojave Indian Tribe) have no state and are not listed on a page or in the files of this release.
• A place that spans counties appears on every county page FEMA lists it under.
• Repeat-flood data are counts for the county and each community. No property detail (addresses, dates or dollars) is published.
• The map leaves out American Samoa, Guam, the Northern Mariana Islands and the U.S. Virgin Islands (not in the Census map file used); their state pages list every unit.

Rules for edge cases in this release

Rules applied where the source data allow two treatments:

• Vermont Public Assistance filed under a bare name. An applicant filed with no "Town of", "City of" or "Village of" (for example "Colchester" or "Bethel") is credited to a jurisdiction when exactly one plan jurisdiction in the county has that name. Its dollars also count in the county's Public Assistance.
• Vermont towns with no row in FEMA plan data show no town Public Assistance figure (a blank, not zero).
• Michigan, May 2020 Midland dam failures. DR-4547 is counted as Inland Flooding because its title names severe storms and flooding (FEMA's incident type is Dam/Levee Break). This applies in Arenac, Gladwin, Iosco, Midland and Saginaw counties.
• Alaska forecast zones. A zone record filed under the former Valdez-Cordova area (split in 2019 into the Chugach and Copper River census areas) goes to the census areas that the first zone file after the split lists for that zone: the Prince William Sound zones to Chugach, the Copper River Basin zone to both. Zone records in Anchorage, Bethel, Fairbanks North Star, Matanuska-Susitna and Yukon-Koyukuk are placed by the same zone-county files. Two zones that no later file lists for either census area (AK224 and AK226) stay in both, with a note on each record, because the 2020 zone outlines that could decide it are no longer published.
• Puerto Rico forecast zones. Flood and wind zone records in Dorado, Guayama, Humacao and Vieques are placed in municipios by the zone-county files.
• Southeastern Connecticut Planning Region. The Public Assistance total of the region's towns includes $115,589.80 of applicants matched to its towns by the Connecticut town rules.
• Bedford County, Virginia. Records filed under the former city of Bedford, which joined the county in 2013, are counted in the county.
• Wakulla County, Florida. The worst hail is 2.75 inches on 1995-06-10, as NOAA records it for the county.
• Perry County, Illinois and Washington County, Virginia. The hail count (Perry) and the wind count (Washington) each leave out one record that NOAA codes to the county.
• Other rules. An area whose only plan is waiting for local adoption shows that plan as an update, not as the current plan. Vermont and Connecticut applicants filed as "Town of X", "X, Town of" or "X (Town of)" are credited to the town named. County-coded Storm Events records in "City" units (Baltimore city, Carson City, Charles City County) count in that unit. A Vermont town with no NFIP community on record shows a blank repeat-flood count, not zero. Older zone names in western states are matched to a county from a reviewed list. One Davidson County, North Carolina tornado pair counts once, as a duplicate.

How to cite

Each jurisdiction page has a "Cite this" line with the exact text. The pattern is:

Hazard Plain (year). Jurisdiction, ST: risk-assessment baseline packet, release 2026-10 (data as of 2026-09-30; unit type and code). Automated baseline.

On a packet checked against live sources the label in that line reads "Checked against live sources on (date)", as on its page.

Packets, tables and pages are CC BY 4.0 (reuse with credit to Hazard Plain and the release); the federal source data are public domain, and OpenFEMA data are also subject to the OpenFEMA terms. Each packet, state file and bulk file carries the reuse line, a "Cite this" line and the corrections contact.

For the plan's bibliography (FEMA element A4-a), cite the federal sources themselves. Every table in a packet carries its source line; the sources of this release are:

• FEMA, OpenFEMA HazardMitigationPlanStatuses v1, version 2026-09-29T134039Z, pulled 2026-09-30, https://www.fema.gov/api/open/v1/HazardMitigationPlanStatuses (API)
• FEMA, OpenFEMA DisasterDeclarationsSummaries v2, version 2026-09-29T162832Z, pulled 2026-09-30, https://www.fema.gov/api/open/v2/DisasterDeclarationsSummaries (API)
• FEMA, OpenFEMA PublicAssistanceFundedProjectsSummaries v1, version 2026-09-29T150754Z, pulled 2026-09-30, https://www.fema.gov/api/open/v1/PublicAssistanceFundedProjectsSummaries (API)
• FEMA, OpenFEMA NfipMultipleLossProperties v1, version 2026-08-04T164045Z, pulled 2026-09-30, https://www.fema.gov/api/open/v1/NfipMultipleLossProperties (API)
• FEMA, FEMA National Risk Index 1.20.0 (December 2025), county and census tract tables, version 1.20.0, pulled 2026-09-30, https://www.fema.gov/about/reports-and-data/openfema/nri/v120/NRI_Table_Counties.zip (download, 25 MB), https://www.fema.gov/about/reports-and-data/openfema/nri/v120/NRI_Table_CensusTracts.zip (download, 635 MB)
• NOAA NCEI, NOAA NCEI Storm Events Database, details files (csvfiles v1.0), version c20260918, pulled 2026-09-30, https://www.ncei.noaa.gov/pub/data/swdi/stormevents/csvfiles/
• NOAA National Weather Service, NWS county-public forecast zones correlation file, version bp03mr20 to bp16ap26 (11 files; newest bp16ap26), pulled 2026-09-30, https://www.weather.gov/gis/ZoneCounty
• U.S. Census Bureau, Census 2024 Gazetteer, 2020 place by county, CT county-to-cousub crosswalk, version 2024-gazetteer, pulled 2026-09-30, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_counties_national.zip, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_cousubs_national.zip, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_place_national.zip, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_unsd_national.zip, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_elsd_national.zip, https://www2.census.gov/geo/docs/maps-data/data/gazetteer/2024_Gazetteer/2024_Gaz_scsd_national.zip, https://www2.census.gov/geo/docs/reference/codes2020/national_place_by_county2020.txt, https://www2.census.gov/geo/docs/reference/ct_change/ct_cou_to_cousub_crosswalk.xlsx
• U.S. Census Bureau, Census 2020 Block Assignment Files (county subdivision and place), version baf2020, pulled 2026-10-02, https://www2.census.gov/geo/docs/maps-data/data/baf2020/, https://www2.census.gov/programs-surveys/decennial/2020/data/01-Redistricting_File--PL_94-171/
• U.S. Census Bureau, Census county adjacency file, 2024 vintage, version 2024, pulled 2026-10-03, https://www2.census.gov/geo/docs/reference/county_adjacency/county_adjacency2024.txt
• NOAA National Hurricane Center, NHC HURDAT2 best-track data (Atlantic 1851 to 2025; Northeast and North Central Pacific 1949 to 2025), version 2025-092326, pulled 2026-10-03, https://www.nhc.noaa.gov/data/hurdat/hurdat2-1851-2025-092326.txt, https://www.nhc.noaa.gov/data/hurdat/hurdat2-nepac-1949-2025-092926.txt