usdata

NOAA · GHCN-Daily Station Observations

Daily station weather

Global Historical Climatology Network daily summaries: temperature, precipitation, snow, and other elements from land surface stations, served by the NCEI Access Data Service with station and date filtering.

CSVSince v0.2
Quick startWalkthroughGuide
The walkthrough's first look at daily station weather

At a glance

Spatial
Land surface stations; more than 100,000 stations in 180 countries and territories
Time step
Daily
Updates
Daily, reconstructed each weekend from more than 25 data source components
Latency
Real-time streams are replaced by archive-ready sources 45 to 60 days after the end of a month
Files
CSV
Selection
Station observations within inclusive calendar dates; selected elements
You provide
Both dates; station IDs or a geographic query
Full description

Global Historical Climatology Network daily summaries: temperature, precipitation, snow, and other elements from land surface stations, served by the NCEI Access Data Service with station and date filtering. Variables are NCEI element codes and the catalog lists the common ones; the delivered set is open and depends on the station.

Quick start

Terminal

python -m pip install "usdata[pandas]"
usdata fetch noaa:ghcn-daily \
  --start 2024-01-01 \
  --end 2024-12-31 \
  --vars TMAX,TMIN,PRCP,SNOW \
  -p stations=USW00013967 \
  -p units=metric

Python

from usdata import build_query, fetch, get

items = fetch(
    get("noaa:ghcn-daily"),
    build_query(
        start="2024-01-01",
        end="2024-12-31",
        variables=["TMAX", "TMIN", "PRCP", "SNOW"],
        stations="USW00013967",
        units="metric",
    ),
)
data = items[0].open()

Manifest

# dataset.yaml, then: usdata pull dataset.yaml
name: okc-daily-weather
sources:
  - dataset: noaa:ghcn-daily
    start: 2024-01-01
    end: 2024-12-31
    variables: [TMAX, TMIN, PRCP, SNOW]
    params:
      stations: USW00013967
      units: metric

The same query the walkthrough below ran. The pandas extra opens the files.

Walkthrough

Notebook Manifest Source

Saved results from a run against the live service; the notebook records when it ran and the checksums of what it read. Run it yourself.

The Global Historical Climatology Network daily dataset (GHCN-Daily) is NOAA's archive of daily observations from land stations: maximum and minimum temperature, precipitation, snowfall, snow depth, and other elements from more than 100,000 stations in 180 countries and territories, some going back to the eighteenth century. NCEI assembles it from many source networks, quality-controls it, and serves it through its Access Data Service; one request returns one CSV with a row per station and day.

This walkthrough pulls every day of 2024 at Will Rogers World Airport in Oklahoma City: high and low temperature, precipitation, and snowfall. It needs usdata[pandas] and matplotlib, downloads about 30 kB, and runs in a few seconds.

from datetime import UTC, datetime
from pathlib import Path

import matplotlib.dates as mdates
import matplotlib.pyplot as plt
import pandas as pd

import usdata
from usdata import cite_lockfile, pull, verify

# One figure style for every usdata notebook, so previews look alike.
plt.rcParams.update(
    {
        "figure.figsize": (8, 4.5),
        "figure.dpi": 120,
        "axes.spines.top": False,
        "axes.spines.right": False,
        "axes.grid": True,
        "grid.alpha": 0.3,
        "font.size": 10,
    }
)
manifest = Path("dataset.yaml")
print("Executed (UTC):", datetime.now(UTC).isoformat(timespec="seconds"))
print(f"usdata {usdata.__version__}; pandas {pd.__version__}")
Executed (UTC): 2026-09-24T06:03:50+00:00
usdata 0.26.0; pandas 3.0.6

Select

The manifest names one station, four element codes, and a window. Dates are inclusive calendar dates and any time of day is ignored. USW00013967 is the airport's GHCN station ID; a bbox or place in place of stations finds stations through NCEI's search service, which is how to reach the volunteer and cooperative stations near a city. TMAX and TMIN are the day's high and low, PRCP its precipitation, and SNOW its snowfall. units: metric asks for degrees Celsius and millimetres.

print(manifest.read_text())
name: okc-daily-weather
sources:
  - dataset: noaa:ghcn-daily
    start: 2024-01-01
    end: 2024-12-31
    variables: [TMAX, TMIN, PRCP, SNOW]
    params:
      stations: USW00013967
      units: metric

What arrives

The first pull downloads one CSV and writes dataset.lock.json beside the manifest, pinning the file's checksum. NCEI rebuilds GHCN-Daily every weekend and replaces real-time values with archive sources 45 to 60 days after a month ends, so a later pull of the same days can return different bytes; the lockfile records which ones this analysis used.

result = pull(manifest)
(item,) = result.fetched
print("File:", item.path.name)
print("Bytes:", item.provenance.size)
print("Source:", item.provenance.source_url)
print("Retrieved (UTC):", item.provenance.retrieved_at.isoformat(timespec="seconds"))
print("Checksum:", item.provenance.checksum)
File: daily-summaries_2024-01-01_2024-12-31_eab4db696c51.csv
Bytes: 30726
Source: https://www.ncei.noaa.gov/access/services/data/v1?dataset=daily-summaries&stations=USW00013967&startDate=2024-01-01&endDate=2024-12-31&format=csv&units=metric&includeStationLocation=1&dataTypes=TMAX%2CTMIN%2CPRCP%2CSNOW
Retrieved (UTC): 2026-09-24T06:03:51+00:00
Checksum: sha256:0ba8cb9e10a9457eee1d942a3bf0e5445c41ed73137673292dfadf68ac19028a

Open

The generic CSV reader returns a DataFrame with one row per station and day. DATE is a calendar date with no time or time zone: the station's observation day. Elements are columns named by their codes, in alphabetical order. NCEI's CSV has no units row, so the units come from the request: TMAX and TMIN in degrees Celsius, PRCP and SNOW in millimetres. Each row repeats the station's latitude, longitude, and elevation in metres.

frame = item.open_csv(parse_dates=["DATE"])
elements = ["TMAX", "TMIN", "PRCP", "SNOW"]
print(f"{len(frame)} rows, {frame['DATE'].min():%Y-%m-%d} to {frame['DATE'].max():%Y-%m-%d}")
print("Missing values per element:", frame[elements].isna().sum().to_dict())
frame[["STATION", "DATE", *elements]].head()
366 rows, 2024-01-01 to 2024-12-31
Missing values per element: {'TMAX': 0, 'TMIN': 0, 'PRCP': 0, 'SNOW': 0}
STATION DATE TMAX TMIN PRCP SNOW
0 USW00013967 2024-01-01 5.0 -2.7 0.0 0.0
1 USW00013967 2024-01-02 8.9 -4.3 0.0 0.0
2 USW00013967 2024-01-03 10.6 -2.1 0.0 0.0
3 USW00013967 2024-01-04 11.7 -1.0 2.5 0.0
4 USW00013967 2024-01-05 5.0 1.7 5.3 0.0

A first look

A year of daily highs and lows as a band, with daily precipitation below it.

coldest = frame.loc[frame["TMAX"].idxmin()]
wettest = frame.loc[frame["PRCP"].idxmax()]

fig, (warm, wet) = plt.subplots(
    2, 1, sharex=True, layout="constrained", gridspec_kw={"height_ratios": [3, 2]}
)
warm.fill_between(
    frame["DATE"], frame["TMIN"], frame["TMAX"], color="#b45631", alpha=0.35, linewidth=0
)
warm.plot(frame["DATE"], frame["TMAX"], color="#b45631", linewidth=0.8, label="Daily high")
warm.plot(frame["DATE"], frame["TMIN"], color="#2f6f8f", linewidth=0.8, label="Daily low")
warm.axhline(0, color="black", linewidth=0.8)
warm.annotate(
    f"High of {coldest['TMAX']:.1f} °C, {coldest['DATE']:%d %b}",
    (coldest["DATE"], coldest["TMAX"]),
    xytext=(20, -4),
    textcoords="offset points",
    fontsize=8,
)
warm.set_ylabel("Temperature (°C)")
warm.set_title("Will Rogers World Airport, Oklahoma City: daily weather, 2024")
warm.legend(loc="upper right", fontsize=8, frameon=False)
wet.bar(frame["DATE"], frame["PRCP"], width=1, color="#2563a6")
wet.annotate(
    f"{wettest['PRCP']:.1f} mm, {wettest['DATE']:%d %b}",
    (wettest["DATE"], wettest["PRCP"]),
    xytext=(6, -10),
    textcoords="offset points",
    fontsize=8,
)
wet.set_ylabel("Precipitation (mm)")
wet.xaxis.set_major_formatter(mdates.DateFormatter("%b"))
wet.set_xlabel("Date (2024)")
plt.show()
Saved plot from Daily station weather
total = frame["PRCP"].sum()
wet_days = int((frame["PRCP"] > 0).sum())
top_days = frame.nlargest(3, "PRCP")
print(f"Precipitation in 2024: {total:.0f} mm on {wet_days} days")
top_total = top_days["PRCP"].sum()
print(f"Three wettest days: {top_total:.0f} mm, {top_total / total:.0%} of the year")
print(f"Wettest day {wettest['DATE']:%Y-%m-%d}: {wettest['PRCP']:.1f} mm")
print(f"Days with a high of 35 °C or more: {int((frame['TMAX'] >= 35).sum())}")
print(f"Days with a low below 0 °C: {int((frame['TMIN'] < 0).sum())}")
snow_days = frame.loc[frame["SNOW"] > 0, "DATE"].dt.strftime("%d %b")
print(f"Snowfall: {frame['SNOW'].sum():.0f} mm, on {', '.join(snow_days)}")
Precipitation in 2024: 942 mm on 73 days
Three wettest days: 318 mm, 34% of the year
Wettest day 2024-08-11: 167.4 mm
Days with a high of 35 °C or more: 41
Days with a low below 0 °C: 51
Snowfall: 87 mm, on 14 Jan, 11 Feb, 12 Feb, 27 Mar

Oklahoma City's 2024 precipitation came in bursts: most days were dry, and three days supplied about a third of the year's total, led by 167.4 mm on 11 August. The coldest stretch was the mid-January Arctic outbreak, when the high on 14 January stayed at -13.2 °C. A daily record is what shows these single-day events; the monthly summaries built from it fold them into totals and means.

Pin and cite

verify checks the cached file against the lockfile's checksum. Because NCEI revises GHCN-Daily, keep the cache with the manifest and lockfile; a checksum cannot recreate bytes upstream has replaced. usdata cite dataset.yaml prints the same citation.

assert verify(manifest) == []
for citation in cite_lockfile(manifest):
    print(citation.as_text())
noaa:ghcn-daily
  Menne, M.J., I. Durre, R.S. Vose, B.E. Gleason, and T.G. Houston, 2012: An overview of the Global Historical Climatology Network-Daily Database. Journal of Atmospheric and Oceanic Technology, 29, 897-910, doi:10.1175/JTECH-D-11-00103.1
  homepage: https://www.ncei.noaa.gov/products/land-based-station/global-historical-climatology-network-daily
  license: US Government Work (public domain)
  terms: https://www.ncei.noaa.gov/metadata/geoportal/rest/metadata/item/gov.noaa.ncdc:C00861/html
  retrieved: 2026-09-24; 1 checksummed asset (30,726 bytes) pinned by usdata 0.26.0
  sources: 1

What was awkward

  • The CSV carries no units. SNOW under units: metric is millimetres of snowfall, not centimetres, and the manifest's units is the only record of either, so it has to travel with the data.
  • DATE has no time zone or observing time. An airport's day runs midnight to midnight local standard time, but many cooperative stations report a 24-hour period ending in the morning, so the same DATE at two stations can cover different hours.
  • Values change after the fact: real-time values are replaced 45 to 60 days after the month ends, so a recent window pulled twice can disagree with its own lockfile.
  • Station discovery runs through a separate search service that can fail while data requests still work. Explicit stations avoid it, at the cost of finding the ID elsewhere first.

Used in these studies

Reference

Cite as Menne, M.J., I. Durre, R.S. Vose, B.E. Gleason, and T.G. Houston, 2012: An overview of the Global Historical Climatology Network-Daily Database. Journal of Atmospheric and Oceanic Technology, 29, 897-910, doi:10.1175/JTECH-D-11-00103.1