After examining the record of CONUS stations in the GHCN TMAX record, it is quite reasonable to ask ‘What about the global record?’ I looked the global GHCN data ( excluding the US stations, which dominate the record, especially early on ).
When examining these data,
- Duration of record ( to compare trends in the context of other changes )
- Continuity of record ( to avoid bias from transient stations )
- Spatial coverage ( to capture regional changes )
- Representative distribution of stations
- Reliable data
Further qualities which are not necessarily known from the record are accuracy and being free of biases.
Unfortunately, examining the available station data, it is apparent that international data are very poor in comparison to the US data with respect to duration and coverage:
Further, the representative spatial distribution, coverage, and reliability are very poor for the globe compared to the US record:
The qualities of the data record which are available for US stations since 1895 are simply not present to allow unbiased analysis for the globe.
That may be seen in a slightly different animation of every station which appears for a given year and the number of missing days for that year. This demonstrates the distribution and availability of data, but understates the fragmentary nature of the records: stations come and go, perhaps at the same location, through out the period.