How to Read HURDAT2 Hurricane Best Tracks in Python

On this page

Problem statement

HURDAT2 is the National Hurricane Center's official record of every Atlantic tropical and subtropical cyclone since 1851: 1,988 storms and 55,524 positions in the release of September 2026. It is a plain text file, but not a CSV. Each storm starts with a header line, followed by a block of fix lines. Coordinates carry hemisphere letters instead of signs, missing values are -999, and a one-letter code marks landfalls and other special fixes that fall between the regular six-hourly ones. This page parses it into a pandas table, explains each field, and writes the tracks as lines.

Quick answer

Read the file line by line. A line starting with letters (AL011851) is a header: remember its ID and name. Every other line is a fix: split on commas, attach the current ID, and parse date and time together. Turn 28.0N/94.8W into signed degrees and -999 into NaN. The setup's read_hurdat2() does it in about twenty lines.

storms = hd.groupby("sid").agg(name=("name", "first"), start=("time", "min"), fixes=("time", "size"),
                               peak_kt=("wind", "max"), min_hpa=("pressure", "min"))
print(storms.tail(4).to_string())

             name               start  fixes  peak_kt  min_hpa
sid                                                           
AL302005  EPSILON 2005-11-29 06:00:00     43       75    981.0
AL302020    THETA 2020-11-08 12:00:00     33       60    987.0
AL312005     ZETA 2005-12-30 00:00:00     36       55    994.0
AL312020     IOTA 2020-11-12 12:00:00     26      135    917.0

One row per fix becomes one row per storm with a groupby. Here are the last four by ID: the late storms of the record-breaking 2005 and 2020 seasons, which ran through the alphabet (Epsilon, Zeta, Theta, Iota). Zeta of 2005 started on 30 December and belongs to the 2005 season, although most of its fixes are in January 2006.

Map of the North Atlantic, Caribbean and Gulf of Mexico covered in thin curving lines, every best track since 1851, coloured from grey for the earliest decades to blue and red for recent decades, sweeping west from Africa and recurving north-east towards Europe. — Every track in the HURDAT2 Atlantic file, 1851–2025: 1,957 storms with two or more six-hourly fixes.

Headers, fixes, identifiers and missing values

1. Two kinds of line

lines = open("data/hurdat2-1851-2025-092326.txt").read().splitlines()
print("\n".join(lines[:3]))
heads = [l for l in lines if l[:2].isalpha()]
declared = sum(int(h.split(",")[2]) for h in heads)
print(f"header lines: {len(heads):,}; fixes declared in headers: {declared:,}; data lines: {len(lines) - len(heads):,}")

AL011851,            UNNAMED,     14,
18510625, 0000,  , HU, 28.0N,  94.8W,  80, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999
18510625, 0600,  , HU, 28.0N,  95.4W,  80, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999, -999
header lines: 1,988; fixes declared in headers: 55,524; data lines: 55,524

A header holds the storm ID (basin, number in the season, year), the name (UNNAMED before names were assigned in 1950) and the number of fix lines that follow. The 1,988 headers declare 55,524 fixes, exactly the number of data lines, which is a useful check that a download is complete. Each fix line has 21 fields: date, time, identifier, status, latitude, longitude, maximum wind in knots, central pressure in hPa, twelve wind radii and the radius of maximum wind.

2. The record identifier

print(hd.record.replace("", "(blank)").value_counts().to_dict())
print(f"fixes at a synoptic hour (00, 06, 12, 18 UTC): {hd.synoptic.mean():.1%}; "
      f"non-synoptic fixes that carry an identifier: {(hd.record[~hd.synoptic] != '').mean():.1%}")

{'(blank)': 54228, 'L': 1213, 'I': 35, 'R': 11, 'P': 10, 'T': 9, 'S': 8, 'C': 5, 'W': 4, 'G': 1}
fixes at a synoptic hour (00, 06, 12, 18 UTC): 98.0%; non-synoptic fixes that carry an identifier: 100.0%

The third field is usually blank. L marks a landfall (1,213 fixes); the others mark an intensity peak (I), a minimum in pressure (P), a change of status (S), genesis (G) and similar events. 98.0% of fixes fall at 00, 06, 12 or 18 UTC, the synoptic hours. Every one of the remaining fixes carries an identifier: the extra fixes are added only to record such moments. Sums and rates over time must leave them out, or a landfall counts twice.

3. The status of the system

print(hd.status.value_counts().to_dict())
tropical = hd.status.isin(["TD", "TS", "HU"])
print(f"fixes when the system was tropical: {tropical.mean():.1%}; subtropical (SD, SS): {hd.status.isin(['SD', 'SS']).mean():.1%}")

{'TS': 20572, 'HU': 15779, 'TD': 9388, 'EX': 6444, 'LO': 1775, 'SS': 758, 'SD': 355, 'DB': 315, 'WV': 138}
fixes when the system was tropical: 82.4%; subtropical (SD, SS): 2.0%

The status says what kind of system the storm was at that moment: tropical depression (TD), tropical storm (TS) or hurricane (HU), their subtropical equivalents (SD, SS), or not a tropical cyclone at all: extratropical (EX), a remnant low (LO), a disturbance (DB) or a tropical wave (WV). 82.4% of fixes are tropical. The 6,444 extratropical fixes include some of the strongest winds in the file, from storms that turned into North Atlantic gales. Category and energy statistics should filter on status as well as wind.

Stacked percentage bars of HURDAT2 fix status by 25-year era from 1851–75 to 2001–25: hurricane and tropical storm shares shrink while tropical depression, extratropical and remnant-low shares grow. — Status of the fixes by era. Early storms were recorded mostly at hurricane and tropical-storm stages; modern best tracks also follow depressions, remnants and extratropical stages.

4. Missing values are -999

raw_p = pd.to_numeric(pd.Series([l.split(",")[7] for l in lines if not l[:2].isalpha()]))
print(f"pressure as read: mean {raw_p.mean():.0f}; after -999 to NaN: mean {hd.pressure.mean():.0f} hPa from {hd.pressure.notna().sum():,} fixes")
by = hd.groupby((hd.season - 1851) // 25 * 25 + 1851).agg(pressure=("pressure", lambda s: s.notna().mean()), radii=("r34_ne", lambda s: s.notna().mean()))
print(by.round(3).to_string())

pressure as read: mean -95; after -999 to NaN: mean 992 hPa from 25,193 fixes
        pressure  radii
season                 
1851       0.015  0.000
1876       0.019  0.000
1901       0.023  0.000
1926       0.055  0.000
1951       0.321  0.000
1976       0.820  0.000
2001       1.000  0.883

Read as numbers without replacing -999, central pressure averages −95 hPa. Replaced, it averages 992 hPa over the 25,193 fixes that have one. The table shows how coverage grew: pressure is known for about 2% of fixes before 1926, 32.1% in 1951–1975 and all fixes since 2001, and wind radii exist only from 2004.

A season table, a track layer and the Pacific file

The data used on this page

The National Hurricane Center's HURDAT2 best-track files for the Atlantic (1851–2025) and the north-east and north-central Pacific (1949–2025).

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 National Hurricane Center HURDAT2 best tracks: Atlantic 1851-2025 (released 23 September 2026) and
# north-east and north-central Pacific 1949-2025 (29 September 2026). NHC keeps every release under its own name.
FILES = ["hurdat2-1851-2025-092326.txt", "hurdat2-nepac-1949-2025-092926.txt"]
for name in FILES:
    if not Path(f"data/{name}").exists():
        Path(f"data/{name}").write_bytes(requests.get(f"https://www.nhc.noaa.gov/data/hurdat/{name}", timeout=300).content)

COLS = ["date", "time", "record", "status", "lat", "lon", "wind", "pressure"] + \
       [f"r{kt}_{q}" for kt in (34, 50, 64) for q in ("ne", "se", "sw", "nw")] + ["rmw"]


def read_hurdat2(path):
    """One header line per storm (id, name, number of fixes), then one line per fix; -999 means missing."""
    rows = []
    for line in open(path):
        f = [x.strip() for x in line.split(",")]
        if f[0][:2].isalpha():
            sid, name = f[0], f[1]
            continue
        rows.append([sid, name] + f[:21])
    h = pd.DataFrame(rows, columns=["sid", "name"] + COLS)
    h["time"] = pd.to_datetime(h.date + h.time, format="%Y%m%d%H%M")
    h["lat"] = h.lat.str[:-1].astype(float) * np.where(h.lat.str[-1] == "S", -1, 1)
    h["lon"] = h.lon.str[:-1].astype(float) * np.where(h.lon.str[-1] == "W", -1, 1)
    for c in COLS[6:]:
        h[c] = pd.to_numeric(h[c]).replace(-999, np.nan)
    h["season"] = h.sid.str[-4:].astype(int)
    h["synoptic"] = h.time.dt.strftime("%H%M").isin(["0000", "0600", "1200", "1800"])
    return h.drop(columns="date")


hd = read_hurdat2("data/hurdat2-1851-2025-092326.txt")
print(f"HURDAT2 Atlantic: {hd.sid.nunique():,} storms, {len(hd):,} fixes, {hd.season.min()}-{hd.season.max()}")

HURDAT2 Atlantic: 1,988 storms, 55,524 fixes, 1851-2025

Example 1: the 2025 season in one table

s25 = storms[storms.index.str.endswith("2025")]
print(s25.assign(start=s25.start.dt.strftime("%d %b")).to_string())

               name   start  fixes  peak_kt  min_hpa
sid                                                 
AL012025     ANDREA  22 Jun     15       35   1014.0
AL022025      BARRY  28 Jun      8       40   1006.0
AL032025    CHANTAL  04 Jul     16       50   1002.0
AL042025     DEXTER  02 Aug     43       70    988.0
AL052025       ERIN  11 Aug     68      140    913.0
AL062025    FERNAND  23 Aug     21       50    999.0
AL072025  GABRIELLE  17 Sep     45      120    944.0
AL082025   HUMBERTO  24 Sep     29      140    918.0
AL092025     IMELDA  26 Sep     38       80    966.0
AL102025      JERRY  07 Oct     18       55   1000.0
AL112025      KAREN  08 Oct     15       50    996.0
AL122025    LORENZO  13 Oct     10       50   1000.0
AL132025    MELISSA  21 Oct     49      165    892.0

The 2025 season had 13 named storms. Three reached 140 kt or more: Erin, Humberto and Melissa. Melissa's 165 kt and 892 hPa make it the strongest of the season and one of the strongest Atlantic hurricanes on record.

Horizontal bar chart of the peak wind of each 2025 Atlantic storm, labelled with wind and pressure, from Andrea at 35 kt to Melissa at 165 kt and 892 hPa, with Erin and Humberto at 140 kt. — The 2025 season: 13 named storms, three of them category 5.

Example 2: tracks as lines

from shapely.geometry import LineString

trk = hd[hd.synoptic].sort_values(["sid", "time"]).groupby("sid").filter(lambda g: len(g) > 1)
lines_gdf = gpd.GeoDataFrame(storms.loc[trk.sid.unique()], crs=4326,
                             geometry=[LineString(zip(g.lon, g.lat)) for _, g in trk.groupby("sid", sort=False)])
lines_gdf.reset_index().assign(start=lambda d: d.start.astype(str)).to_file("hurdat2_atlantic.gpkg", layer="tracks", driver="GPKG")
print(f"{len(lines_gdf):,} tracks written; storms with a single fix: {hd.sid.nunique() - len(lines_gdf)}")

1,957 tracks written; storms with a single fix: 31

Drawing one line per storm through its six-hourly fixes gives 1,957 tracks. The 31 storms with a single synoptic fix cannot form a line and are left out. Keep them as points if you are counting storms. The GeoPackage carries each storm's name, start, peak wind and minimum pressure.

Example 3: the Pacific file reads the same way

pac = read_hurdat2("data/hurdat2-nepac-1949-2025-092926.txt")
print(f"north-east and central Pacific: {pac.sid.nunique():,} storms, {len(pac):,} fixes, basins {sorted(pac.sid.str[:2].unique())}; "
      f"east longitudes: {(pac.lon > 0).sum()}")

north-east and central Pacific: 1,262 storms, 32,060 fixes, basins ['CP', 'EP']; east longitudes: 633

The Pacific file has the same layout: 1,262 storms from the eastern (EP) and central (CP) Pacific since 1949. It has 633 fixes with east longitudes, from storms that crossed the date line, which is why the parser applies the sign from the letter rather than assuming west.

What a best track is

A best track is a post-season reanalysis. After each season, NHC forecasters re-examine every storm with all the data that arrived late or was not available in real time, such as aircraft reconnaissance, scatterometer winds, ship and buoy reports and satellite estimates, and write down their best estimate of position and intensity every six hours. Older seasons are revised too: the Atlantic Hurricane Reanalysis Project has been working forward through the record from 1851, revising it a few seasons at a time. That is why NHC keeps every release under a dated name, and why a published analysis should cite the release it used, here the one of 23 September 2026.

Parsing details

Whitespace. Fields are padded with spaces; strip each one after splitting.
Names repeat. UNNAMED covers hundreds of storms, and names are reused every six years. Always group by the ID.
The season is in the ID, not the date: Zeta (AL312005) runs into 2006.
Wind is in knots and is a one-minute sustained wind, rounded to 5 kt; pressure is in hPa (millibars).

Internal links

How to map a hurricane season's tracks — the tracks drawn by intensity
Hurricane tracks in the wrong hemisphere — the N and W suffixes
ACE comes out too high — what the record identifier does to sums
How to map a hurricane track from IBTrACS — the multi-agency archive

FAQ

How do I read HURDAT2 in pandas?

Not with read_csv alone: the header lines have a different number of fields. Read the file line by line, carry the storm ID from each header to the fix lines below it, and build a DataFrame from the fixes.

What does -999 mean in HURDAT2?

Missing. Pressure is missing for most fixes before the 1950s, and wind radii exist only from 2004.