Some surface station temperature data links:

1. USHCN (United States Historical Climate Network) monthly data through 1994 (obsolete, see item 3, but included for completeness.) is the main online documentation file.

2. USHCN daily data through 2002 (daily max/min, you get to calulate means, monthly, etc.) is the main online documentation file.

3. USHCN monthly data through 2002 is the main online documentation file, and provides an overview.

USHCN includes a station history file with summaries of station changes, including location, time of day of observation, etc. GHCN (see below) does not include such "metadata", which is quite unfortunate. The USHCN monthly files include various sets with successive adjustments, some of which are briefly discussed in the aforementioned: The stations in the three USHCN groupings have varied slightly; check the respective "inventory" files.

The above USHCN collections include data from stations within the contiguous 48 US states only. For data from stations around the globe, including USHCN stations, see below.

4. GHCN (Global Historical Climate Network)

Monthly data including their most recent data available (excluding some recent data for some stations for which they include older data, such as the South Pole (Amundsen Scott, which is available at:, several stations from Iceland, (available at: and who knows what else).

Temperature files in pkzip compressed form are at: and in unix, or gzip, compressed form are at: is the main documentation file, with a bit more in a readme file.

Don't fall off your chair when you read: "The reason why GHCN mean temperature data have duplicates while mean maximum and minimum temperature data do not is because there are over 100 different ways in which daily mean temperature has been calculated by meteorologists."

A brief summary on GHCN "quality control" is at:

The GHCN has a country codes file, but not a station history file, whereas the USHCN has station history files, with more detailed "metadata", and more extensive documentation overall. They both have station "inventory" files, which list the id codes for each station, as well as such things as lattitude and longitude. In GHCN terminology, there are stations, locations, sources, and duplicates. It appears to me that duplicate refers to multiple sources for one location, or station, but might also refer to multiple stations in, or near, one location. In this text I use the word station as a synonyn for location, but there is much ambiguity among these terms.

For many stations the GHCN has no recent data; their old data is included simply to preserve it in a convenient place for researchers.

5. GISS has GHCN monthly mean temp data conveniently available (i.e. by station, rather than all/many stations in one big file) at: but watch out for their adjustments. GISS no longer includes USHCN "raw" data, just adjusted data, and GISS does not include any min/max data.

Update 18 Jan 2005: It now appears that GISS excludes temperature data for years prior to 1880 from the files available at that link.

Annual means at GISS are Dec-Nov, while the USHCN uses Jan-Dec. The GHCN lets you roll your own. The USHCN uses Fahrenheit, GISS uses Celsius, and GHCN omits the decimal point, so its files are in tenths of degrees Celsius. All of these temperature data files are ASCII fixed format files. GISS files, and the USHCN daily files, are blank delimited; the others are not. GISS files each refer to one "station" or "location"; the others include many stations in each file.

About station changes: over periods of many decades, there will usually occur various kinds of station changes, for examples: kind of thermometer, location of thermometer, time of day of reading, or observation, and/or station operator. In the history files that accompany the USHCN collections, there are summaries of such changes (metadata). If you examine, for example, the USHCN mean monthly temperature file of the most recent USHCN collection (item 3 above) you can notice that virtually every station change results in a change to adjustments to the temperatures for that station. While one may question the accuracy of those adjustments, at least there are data on which to base most of them. On the other hand, the GHCN collection does not include such metadata, and the GHCN adjusted temperatures cannot be expected to be as accurate as USHCN adjusted temperatures, however accurate the latter may, or may not, be.

Happy data dunking! Caution: it may become habit forming.

Jerry Brennan
18 January, 2005