⚙️ How-To Severe storms & tornadoes Health & disease mapping Notebook By · Updated

How to Count Weather Deaths by Event Type, Place and Age

On this page

Problem statement

Which kinds of weather killed people in the United States in 2025, where were they, and how old were they? The Storm Events database answers that in two places: death counts on each event row, and a separate fatalities file with one row per person, giving age, sex and location such as "in a vehicle" or "mobile home". The two do not agree exactly. Event types split heat and cold across two names each, and one flood can dominate a whole year. This page counts 2025's weather deaths by type, place, age and month, and shows the choices that change the ranking.

Quick answer

Read the year's StormEvents_fatalities file and join it to the details file on EVENT_ID for the event type, state and date. Count FATALITY_TYPE == "D" (direct) separately from "I" (indirect), and merge Heat with Excessive Heat and the two cold types before ranking.

fat = pd.read_csv("data/storm-events-fatalities-2025-c20260819.csv.gz")
f = fat.merge(ev[["EVENT_ID", "EVENT_TYPE", "STATE", "CZ_NAME", "begin"]], on="EVENT_ID", how="left")
print(f"{len(f):,} deaths: {(f.FATALITY_TYPE == 'D').sum()} direct, {(f.FATALITY_TYPE == 'I').sum()} indirect")
print(f[f.FATALITY_TYPE == "D"].EVENT_TYPE.value_counts().head(6).to_string())
895 deaths: 632 direct, 263 indirect
EVENT_TYPE
Flash Flood       206
Tornado            64
Heat               62
Wildfire           61
Rip Current        38
Excessive Heat     37

The fatalities file lists 895 deaths for 2025: 632 direct, caused by the weather itself, and 263 indirect, such as a crash on an icy road or a heart attack shovelling snow. Flash floods caused the most direct deaths (206). Tornadoes (64), heat (62) and wildfire (61) were close behind, followed by rip currents and excessive heat.

Horizontal stacked bar chart of 2025 weather deaths by cause, direct in red and indirect in orange: heat and flood longest, then tornado, wildfire, thunderstorm wind and rip current, with winter weather, cold, lightning, winter storm, dust storm and high surf shorter. Horizontal stacked bar chart of 2025 weather deaths by cause, direct in red and indirect in orange: heat and flood longest, then tornado, wildfire, thunderstorm wind and rip current, with winter weather, cold, lightning, winter storm, dust storm and high surf shorter.
Weather deaths in 2025 by cause, with heat and cold types merged. Floods caused 222 direct deaths and heat 99.

Counting deaths and grouping causes

1. Two places to count deaths

print(f"details file: {ev.DEATHS_DIRECT.sum()} direct + {ev.DEATHS_INDIRECT.sum()} indirect; "
      f"fatalities file: {(fat.FATALITY_TYPE == 'D').sum()} direct + {(fat.FATALITY_TYPE == 'I').sum()} indirect")
per_event = fat[fat.FATALITY_TYPE == "D"].groupby("EVENT_ID").size()
diff = ev.set_index("EVENT_ID").DEATHS_DIRECT.sub(per_event, fill_value=0)
print(f"events where the two disagree: {(diff != 0).sum()}")
details file: 628 direct + 261 indirect; fatalities file: 632 direct + 263 indirect
events where the two disagree: 4

The event rows sum to 628 direct and 261 indirect deaths, and the fatalities file has 632 and 263. Four events disagree. The files were revised at different moments, and a death added later did not always reach the event row. Use the fatalities file when you need people, and say which file you used.

2. Merge heat and cold types before ranking

group = {"Heat": "Heat", "Excessive Heat": "Heat", "Cold/Wind Chill": "Cold", "Extreme Cold/Wind Chill": "Cold",
         "Flash Flood": "Flood", "Flood": "Flood", "Coastal Flood": "Flood"}
f["group"] = f.EVENT_TYPE.map(group).fillna(f.EVENT_TYPE)
print(f[f.FATALITY_TYPE == "D"].group.value_counts().head(6).to_string())
group
Flood                222
Heat                  99
Tornado               64
Wildfire              61
Rip Current           38
Thunderstorm Wind     36

Merging the types changes the ranking. Floods (flash, river and coastal) account for 222 direct deaths, and heat rises to second with 99 once Heat and Excessive Heat are combined. Weather Service offices choose between the two names by local warning criteria, so splitting them understates heat. Tornadoes (64) and wildfire (61) follow.

3. Where people died

d = f[f.FATALITY_TYPE == "D"]
print(d.FATALITY_LOCATION.value_counts().head(6).to_string())
print("flood deaths in vehicles:", ((d.group == "Flood") & (d.FATALITY_LOCATION == "Vehicle/Towed Trailer")).sum(),
      "of", (d.group == "Flood").sum())
FATALITY_LOCATION
Unknown                  111
Vehicle/Towed Trailer     99
Outside/Open Areas        98
Permanent Home            90
In Water                  85
Camping                   46
flood deaths in vehicles: 62 of 222

The location is unknown for 111 direct deaths. Of the rest, vehicles (99) are the most common place, followed by outdoors (98), at home (90) and in water (85). For floods, 62 of the 222 deaths were in vehicles, people who drove into flooded roads or whose cars were swept away. The 46 deaths while camping are almost all from the Texas flood of 4 July.

Stacked horizontal bars showing where people died for six causes: flood deaths are spread across vehicles, outdoors, in water and camping; heat mostly outdoors; tornado deaths at home and in mobile homes; wildfire mostly unknown; rip current in water. Stacked horizontal bars showing where people died for six causes: flood deaths are spread across vehicles, outdoors, in water and camping; heat mostly outdoors; tornado deaths at home and in mobile homes; wildfire mostly unknown; rip current in water.
Where people died, by cause. 62 of the 222 flood deaths were in vehicles.

4. Age and sex

print(f"age known: {d.FATALITY_AGE.notna().mean():.0%}; median {d.FATALITY_AGE.median():.0f}; sex: {d.FATALITY_SEX.value_counts(dropna=False).to_dict()}")
print(d.groupby("group").FATALITY_AGE.median().loc[["Flood", "Heat", "Tornado", "Rip Current", "Cold"]].to_dict())
age known: 82%; median 50; sex: {'M': 337, 'F': 212, nan: 83}
{'Flood': 45.5, 'Heat': 61.0, 'Tornado': 62.0, 'Rip Current': 38.0, 'Cold': 59.0}

Age is known for 82% of direct deaths, and the median is 50. 337 of the dead were male and 212 female. The median age differs by cause: 38 for rip currents and 45.5 for floods, against 61 for heat and 62 for tornadoes. Heat and tornadoes kill older people, who are less able to cope with heat or to shelter quickly. Floods and surf kill younger people who are out in them.

The deadliest events and the yearly cycle

The data used on this page

The National Centers for Environmental Information's Storm Events Database details and fatalities files for 2025.

from pathlib import Path

import geopandas as gpd
import numpy as np
import pandas as pd
import requests

Path("data").mkdir(exist_ok=True)
# NOAA NCEI Storm Events Database: the 2025 details and fatalities files (revision c20260819) and the 2024 details file
# (revision c20260728), saved on this site on 15 October 2026 because NCEI replaces a year's files whenever it revises
# them and the old names disappear. Live: https://www.ncei.noaa.gov/pub/data/swdi/stormevents/csvfiles/
FILES = ["storm-events-details-2025-c20260819.csv.gz", "storm-events-fatalities-2025-c20260819.csv.gz",
         "storm-events-details-2024-c20260728.csv.gz"]
for name in FILES:
    if not Path(f"data/{name}").exists():
        Path(f"data/{name}").write_bytes(requests.get(f"https://data.spatialworkflow.io/{name}", timeout=600).content)
ev = pd.read_csv("data/storm-events-details-2025-c20260819.csv.gz", low_memory=False)


def dollars(s):
    """Damage is stored as text: '10.00K', '1.60B', '0' or blank. A blank means no estimate, not zero."""
    s = s.str.strip()
    return pd.to_numeric(s.str.rstrip("KMB"), errors="coerce") * s.str[-1].map({"K": 1e3, "M": 1e6, "B": 1e9}).fillna(1.0)


def times(df):
    """Event times are local to the county or zone ('CST-6'), with a two-digit year."""
    df["begin"] = pd.to_datetime(df.BEGIN_DATE_TIME, format="%d-%b-%y %H:%M:%S")
    df["end"] = pd.to_datetime(df.END_DATE_TIME, format="%d-%b-%y %H:%M:%S")
    df["utc_offset"] = df.CZ_TIMEZONE.str.extract(r"(-?\d+)$")[0].astype(int)
    df["begin_utc"] = df.begin - pd.to_timedelta(df.utc_offset, unit="h")
    df["prop"], df["crops"] = dollars(df.DAMAGE_PROPERTY), dollars(df.DAMAGE_CROPS)
    return df


ev = times(ev)
print(f"{len(ev):,} Storm Events records for 2025, {ev.EPISODE_ID.nunique():,} episodes, {ev.EVENT_TYPE.nunique()} event types")
72,360 Storm Events records for 2025, 9,846 episodes, 48 event types

Example 1: deadliest events

top = d.groupby(["EVENT_ID", "EVENT_TYPE", "STATE", "CZ_NAME"]).size().sort_values(ascending=False).head(6)
print(top.to_string())
kerr = d[(d.STATE == "TEXAS") & (d.begin.dt.strftime("%m-%d").between("07-03", "07-06")) & (d.group == "Flood")]
print(f"Texas floods of 3-6 July 2025: {len(kerr)} direct deaths, {kerr.FATALITY_AGE.lt(18).sum()} under 18")
EVENT_ID  EVENT_TYPE   STATE       CZ_NAME                              
1287369   Flash Flood  TEXAS       KERR                                     61
1287451   Flash Flood  TEXAS       KERR                                     45
1292941   Tornado      KENTUCKY    LAUREL                                   17
1230303   Wildfire     CALIFORNIA  LOS ANGELES COUNTY SAN GABRIEL VALLEY    17
1230302   Wildfire     CALIFORNIA  EST SAN GABRIEL MOUNTAINS                17
1230301   Wildfire     CALIFORNIA  EST SANTA MONICA MOUNTAINS               12
Texas floods of 3-6 July 2025: 123 direct deaths, 40 under 18

Two Kerr County flash-flood events, 61 and 45 deaths, are the year's deadliest, and the Texas floods of 3–6 July killed 123 people in all, 40 of them under 18, many at summer camps on the Guadalupe River. The Laurel County, Kentucky segment of the 16 May tornado that struck London (17 deaths) and the January Los Angeles fires follow. The fires' deaths are split across three fire-weather zones, so a count by zone breaks each fire into pieces.

Example 2: deaths by month

print(d.groupby([d.begin.dt.strftime("%m").rename("month"), "group"]).size().unstack(fill_value=0)[["Flood", "Heat", "Tornado", "Cold"]].to_string())
group  Flood  Heat  Tornado  Cold
month                            
01         0     0        0    10
02        15     0        2     7
03         3     0       21     1
04        16     2        9     0
05         5     5       26     2
06        24    29        6     3
07       135    38        0     0
08         7    19        0     0
09         8     6        0     0
10         4     0        0     0
11         2     0        0     0
12         3     0        0     2

Each cause has its own season. Cold deaths come in January and February, tornado deaths from March to May, and heat deaths from June to August. July 2025 had 135 flood deaths, almost all in Texas. One event can change a year's ranking: without the Texas floods of early July, floods and heat would be level.

Stacked monthly bar chart of 2025 direct weather deaths by cause: a January spike from wildfire, small spring bars mostly tornado, a June bar of heat and flood, and a very tall July bar almost entirely flood. Stacked monthly bar chart of 2025 direct weather deaths by cause: a January spike from wildfire, small spring bars mostly tornado, a June bar of heat and flood, and a very tall July bar almost entirely flood.
Direct deaths by month and cause. July's 135 flood deaths are almost all the Texas floods.

What counts as a weather death

Storm Data counts a death as direct when the weather itself caused it: drowning in flood water, being struck by debris or by lightning, or heat stroke. A death is indirect when the weather set up the circumstances, such as a car crash on ice, carbon monoxide from a generator, or a cardiac arrest during clean-up. Offices learn of deaths from medical examiners, the media and emergency managers, and heat deaths in particular are hard to attribute: public-health counts based on death certificates are several times higher than Storm Data's.

So the file is good for comparing causes within itself (who dies in floods, where tornado victims were) and weaker as a total of weather deaths. Heat is the cause it undercounts most.

Before publishing a death count

  • Say direct or indirect. Mixing them changes totals by about 40%, and other sources mostly report direct deaths.
  • Merge synonym event types, at least Heat/Excessive Heat and Cold/Wind Chill/Extreme Cold/Wind Chill.
  • Late additions. The fatalities file is revised with the details file. A count read early in the following year can be short.
  • Unknown age and sex are blanks, not zeros. Drop them before averaging.

FAQ

What weather killed the most people in the US in 2025?

Floods, with 222 direct deaths in NOAA's Storm Events data (206 from flash floods), most in the Texas flood of 4 July. Heat was second with 99 once both heat types are combined.

What is an indirect weather death?

A death the weather contributed to but did not directly cause, such as a crash on an icy road. 2025 had 263 indirect deaths in the fatalities file.