As mentioned earlier, I changed my USHCN methodology to match my GHCN methodology – averaging by month for all stations, and then averaging all months for the year. This fixed the discrepancy between my 2014 spike and Anthony’s.
Nick Stokes keeps talking about anomalies, which are necessary when trying to do an absolute comparison of partial years. I wasn’t attempting to do that – I was measuring temperature differences, so his point is irrelevant to what I am doing.
Here are the USHCN final and raw temperatures using the new method. It is almost identical to the old method. They are turning a 95 year cooling trend into a warming trend.
But it is worse than seems. The USHCN “raw monthly data” isn’t really raw. The graph below shows the discrepancy between GHCN monthly “raw” and actual GHCN HCN daily station data – which is truly raw. As you can see, USHCN “raw” data is already adjusted to create a warming trend, and then they do their own adjustments on top of that.
The next graph shows the total adjustments from GHCN HCN daily to USHCN final.
But it is worse than even that seems. NCDC adds additional warming on to the USHCN adjustments. Unfortunately their web site is dead, so I can’t generate that graph right now.