AsukaWasHere/noaa-storm-events

---
license: other
license_name: public-domain
task_categories:
- tabular-classification
- tabular-regression
- time-series-forecasting
tags:
- weather
- climate
- natural-disasters
- insurance
- geospatial
- united-states
pretty_name: NOAA Storm Events Database 1950-Present
size_categories:
- 1M<n<10M
---

# NOAA Storm Events Database 1950–Present

Every storm event recorded by the NOAA National Centers for Environmental
Information (NCEI) since 1950 — cleaned, structured, and linked across
3 relational tables. 1.94 million+ events covering tornadoes, floods,
hurricanes, hail, lightning, and 58 other event types across all US states
and territories. $662 billion in total documented damage.

**This repository contains a 1,000-row sample of the Details table (Public Domain).**
Full dataset (all 3 tables — Details, Fatalities, Locations — in CSV + Parquet)
available at [claritystorm.com/datasets/noaa-storm-events](https://claritystorm.com/datasets/noaa-storm-events).

## Quick Start

```python
from datasets import load_dataset
ds = load_dataset("claritystorm/noaa-storm-events")

import pandas as pd
df = pd.read_csv("sample_1000.csv")
print(df["EVENT_TYPE"].value_counts().head(10))
print(df.groupby("YEAR")["DEATHS_TOTAL"].sum())
```

## Schema (Details table, selected fields)

| Field | Type | Description |
|-------|------|-------------|
| EVENT_ID | int | Unique event identifier (links to fatalities and locations) |
| EPISODE_ID | int | Episode grouping related events |
| YEAR | int | Event year (1950–2024) |
| MONTH_NAME | string | Month name |
| BEGIN_DATE | string | Begin date (YYYY-MM-DD derived) |
| END_DATE | string | End date (YYYY-MM-DD derived) |
| STATE | string | US state name |
| CZ_TYPE | string | Zone type: C (county), Z (forecast zone), M (marine) |
| EVENT_TYPE | string | Storm type (Tornado, Flood, Hail, etc.) |
| DEATHS_DIRECT | int | Direct fatalities |
| DEATHS_INDIRECT | int | Indirect fatalities |
| DEATHS_TOTAL | int | Total fatalities (derived) |
| INJURIES_DIRECT | int | Direct injuries |
| INJURIES_INDIRECT | int | Indirect injuries |
| INJURIES_TOTAL | int | Total injuries (derived) |
| DAMAGE_PROPERTY_USD | float | Property damage in USD (parsed from NOAA strings) |
| DAMAGE_CROPS_USD | float | Crop damage in USD (parsed from NOAA strings) |
| TOTAL_DAMAGE_USD | float | Total damage in USD (derived) |
| MAGNITUDE | float | Storm magnitude (wind speed, hail size, etc.) |
| BEGIN_LAT | float | Begin latitude |
| BEGIN_LON | float | Begin longitude |
| END_LAT | float | End latitude |
| END_LON | float | End longitude |
| EPISODE_NARRATIVE | string | Episode description |
| EVENT_NARRATIVE | string | Event-specific description |

## Tables in Full Dataset

- **details** (~1.94M rows) — one row per storm event with type, location, damage, and casualties
- **fatalities** (~23.8K rows) — one row per fatality with age, sex, and location
- **locations** (~1.75M rows) — one row per affected location with county and coordinates

All three tables join on `EVENT_ID`.

## Source

NOAA National Centers for Environmental Information (NCEI) — Storm Events Database.
NOAA storm events data is a US federal government work in the **public domain** (17 U.S.C. 105).

Processed by [ClarityStorm Data](https://claritystorm.com).