CRUTEM4 is cooler than GHCN-V4

The October 2019 global temperature anomaly that I calculate based on the latest CRUTEM4 station data is 0.15C cooler than that using GHCN-V4 (0.79C as  compared to 0.96C). The temperature series are calculated in exactly the same way and both use HadSST3 for ocean temperatures. The only differences lie within the station data. CRU use 7280 stations and give similar results to GHCN-V3. V4 has many more stations (17378) but the data handling is different, even the station IDs have changed and mostly originate from GHCN-Daily.

If we look at the annual averages we can see a clear discrepancy.

Comparing annual temperatures for CRUTEM4 and GHCN-V4. 2019 is the averages of the 10 months Jan-Oct.

The HadCRUT4.6 official temperature is calculated on a 5deg (lat,lon) grid is 0.75C, whereas  I am instead Spherical Triangulation which gives results similar to Cowtan & Way. My 3D-H4 results given below.

About Clive Best

PhD High Energy Physics Worked at CERN, Rutherford Lab, JET, JRC, OSVision
This entry was posted in AGW, Climate Change, climate science, CRU, Hadley, NOAA and tagged . Bookmark the permalink.

7 Responses to CRUTEM4 is cooler than GHCN-V4

  1. Nick Stokes says:

    Usually our results are quite similar. But for October, using TempLS mesh, I got 0.89C, cf your 0.96. GHCN V4 and ERSST.

    • Clive Best says:

      I don’t understand it either.

      Maybe the difference lies in the SST data. I might try using ERSST to check.

      • bindidon says:

        Clive Best

        “Maybe the difference lies in the SST data.”

        Certainly they do! If you would use JMA’s COBE-SST2 instead, the land/SST average would differ even more because they have a very conservative SST processing.

        Btw, thanks again here for having recently published your pretty evaluation of GHCN daily, so I could compare my little layman’s work with your more professional and more mathy approach:

        Given my rather simple work, I shouldn’t be unsatisfied with the result.

        Amazing is that the same software is used to process PSMSL tide gauge data (here too with a simple approach omitting vertical land movement, oscillations etc).

        J.-P. D.

        • Olof R says:

          Are you sure about COBE SST2? Cowtan has produced a C&W version using COBE SST2 instead of HadSST3, to adress a recent cooling bias in the latter.

          And, yes the most recent values of “COBECRUT” are a tad warmer (~0.04 C) than standard C&W.
          But using HadSST4 instead makes the dataset even warmer..

        • Clive Best says:

          Well done! Processing GHCN-Daily is not an easy job, yet I think it is about the closest we can get to the raw data.

          The basic problem is how to integrate a complex and ever changing set of station data. Robert Rhode from Berkeley Earth has done the only other analysis I am aware of.

        • bindidon says:


          Thanks for the valuable hint. Duh… I think I live with a somewhat ‘deprecated’ view on JMA’s SST work.

        • bindidon says:


          Thanks in turn for the convenient reply.

          What disturbs me in my evaluation is not the discrepancy in the historical part of the generated time series, but the much lower trend it shows nowadays (0.19 vs. your 0.27 °C / decade in the sat era).

          The historical difference one could explain with paucity of data I wouldn’t handle properly; but for 1979-2019, there are really enough stations.

          The bias thus must be due to something different.

          I use a simple strategy eliminating all stations
          – with lifetime less than 10 years, or
          – unable to produce anomalies wrt a specified reference period, and give them a 2nd chance if their grid cell has such data.


Leave a Reply