I recently posed five questions to the HadCRUT5 team down at the Met Office about how they use land station data to calculate their monthly global temperatures with. I got a reply this morning and here is what their expert said. Those without a nerdy disposition should look away now.
Is this land station data really used in the calculations for HadCRUT5?
Yes, these files are the station data source for HadCRUT.18.104.22.168.
Why has the number of UK stations fallen in recent years?
There are two main reasons for the recent fall in station numbers. The first is that station observations may not immediately be communicated outside of the collecting country through near real-time reports or inclusion in national data publications / datasets on which the CRUTEM database is based.
The second reason relates to the processing method used. CRUTEM5 and HadCRUT5 process station data as temperature anomalies, requiring data to be available within a 30-year climatological normal period. These data sets use a climatological normal period of 1961-1990. The use of 1961-1990 is beneficial as the majority of stations in the CRUTEM database have data available in this period, allowing anomalies to be computed and gridded. However, recently established meteorological stations may not have data in this period. This prohibits their use in the gridded data set and leads to a reduction in the total number of gridded stations for recent years.
Doesn’t the loss of stations result in a loss of accuracy?
CRUTEM5 and HadCRUT5 include uncertainty assessments for global and regional time series. These uncertainty assessments do not indicate any notable reduction in accuracy in recent years. This is in part because the many modern stations are situated in regions of the world that are already sampled. This is reflected by the number of station observations gridded in CRUTEM5 declining less rapidly than the global coverage of CRUTEM grids (see the below figures, extracted from Figure 6 of the CRUTEM5 paper: Osborn et al., 2020).
There is no notable reduction in grid coverage or increase in uncertainty in global or regional time series for the HadCRUT5 analysis.
Do you calculate on the fly the 1981-2010 LTA because I notice that most of the stations use 1961-1990 normals?
HadCRUT5 data are provided as anomalies with respect to average conditions in the 1961-1990 period. Anomalies with respect to other long-term averages are computed from the HadCRUT5 1961-1990 anomaly series as a post-processing step.
Will the CRU update this file periodically?
Yes, CRU will publish an update of this file with each update of HadCRUT.22.214.171.124.