How far back do surface weather observations go?

It depends on the station. The longest airport records in this archive begin in the 1930s and 1940s; most were in place by the 1970s, and the automated network as it stands today was installed through the 1990s. Each station shows its own first and last day, and a date outside that range says so rather than returning an empty page.

What weather data is available for a past date?

Hour by hour: temperature, dew point, relative humidity, apparent temperature, wind speed and direction, gusts, pressure, visibility, cloud ceiling, precipitation and the reported weather. For the day as a whole: the official high, low and mean temperature, precipitation, snowfall and depth, peak gust, and which phenomena were observed, including thunder, hail and fog.

Can I compare a date to the same date in other years?

Yes. Every station shows that calendar date in every year it recorded one, as a chart and a table, with the record high and low for the date, the 1991 to 2020 normal, and where the year you are viewing ranks among all of them.

Where does this weather data come from?

Hourly observations come from the Iowa Environmental Mesonet's ASOS archive; daily records and climate normals come from NOAA's Global Historical Climatology Network. Both are public data, and both are fetched by your browser directly from those archives. Readings that failed NOAA's own quality checks are discarded rather than displayed.

Is historical weather data free to look up here?

Yes, and no account is required. The underlying archives are public data published by NOAA.

Surface weather, by place and date

What the weather actually was, at nearly three thousand automated stations across the United States, on any day they were recording. Hour by hour, plus the same calendar date in every year on record. The longest records begin in the 1930s. Anywhere else, type the town: you get the nearest station’s readings and a model’s answer for your own coordinate, each labeled as what it is.

Loading the station list…

This section needs JavaScript. The observations are fetched from NOAA and the Iowa Environmental Mesonet by your browser rather than assembled on a server, which is what makes the whole archive available without an account or a charge.

What is here

Hourly: Temperature, dew point, relative humidity, apparent temperature, wind speed and direction, gusts, pressure, visibility, cloud ceiling, hourly precipitation and the reported weather, as a chart and as a table, in the station’s own local time.
Daily: The official high, low and mean temperature, precipitation, snowfall and snow depth, peak gust, mean wind, and the phenomena observed that day: thunder, hail, fog, freezing rain, blowing snow, damaging wind.
Every year: That calendar date in every year on record, with the record high and low for the date, the 1991–2020 normal, the all-years mean, and where your date ranks. Measured at a station, or reanalyzed back to 1940 at a coordinate that never had one.
Anywhere: Type a town or a ZIP code rather than an airport code and the page answers for that coordinate: the nearest recording station’s readings and a gridded model’s output for the exact spot, drawn on one chart so you can see where they disagree, and never presented as the same kind of thing.
And the radar: For dates from 1991, a link into the Level II archive at the nearest WSR-88D: the storm itself, next to the numbers it produced.

Common questions

How far back does this go?

At a station, it depends on the station. The longest airport records here begin in the 1930s and 1940s; most were in place by the 1970s. Each station states its own first and last day, and asking for a date outside that range says so rather than showing an empty table. At a place, the modeled column reaches 1 January 1940 everywhere at once, whether or not anybody ever built an instrument nearby.

Are these measurements or estimates?

Both, and the page always says which. A station page is measurements only: every value is a reading an instrument filed at a named place and a stated minute, nothing is modeled or reconstructed, and readings that failed NOAA’s own quality checks are discarded rather than shown — which is why an occasional day is blank where the days either side of it are not.

A page about a town or a ZIP code has to answer differently, because almost no town has an instrument standing in it. So it shows two things side by side and labels them: the nearest station that was recording, with how far away it is and in which direction, and a gridded model’s answer for your exact coordinate. The first is real and in the wrong place. The second is in the right place and was computed. Every other weather site shows you the second one and calls it your weather; this one would rather you knew.

Where do the numbers for a place with no station come from?

For now and the last two weeks, NOAA’s High-Resolution Rapid Refresh on a 3 km grid — the finest analysis published for the Lower 48, and the same model this site draws as its forecast layer. For anything older, ECMWF’s ERA5 reanalysis, hourly, back to 1 January 1940 anywhere on earth. ERA5 is a 31 km grid, which is coarse enough to be worth saying out loud: it knows the valley, not the hillside. Both arrive by way of Open-Meteo, whose data is published under CC-BY 4.0.

Why is the hourly record thinner than the daily one?

They come from different archives kept for different purposes. A station can have a complete daily climate summary from the 1940s and hourly observations only from the 1970s, because the daily figures were transcribed from paper records and the hourly ones were not.

What time zone are the times in?

The station’s own local time, including the correct offset for that date. A day in January is read on standard time and a day in July on daylight time, as they were lived.

Open the live radar · Historical radar by date · What each radar product shows