I went back and scanned the Rohde paper on BEST methodologies. this paragraph stuck out.
The median length of a temperature time series processed by the
Berkeley Average was only 5.9 years. Further, the inner 50% range for
station record lengths was 2.3 to 10.8 years, and only 4.5% of records
were longer than 30 years. This compares to GHCN data before the
scalpel was applied where 72% of the time series are longer than 30
years and the median length is nearly 50 years. As already stated, the
current climate change analysis framework is designed to be very
tolerant of short and discontinuous records which will allow it to
work with a wide variety of data.
I have said this before. how can you find climatic signals (ie. 30 years) from records that are cut up into little chunks and then sewn back together in such a way as to 'meet expectations'.
I think we have moved so far away from actual data that Global Temperature Datasets should come with a warning like food products. instead of "contains 10% real juice", we could have "contains 4.5% real data, from concentrate".
I think you don't have a fucking clue.