More On The Fake NOAA Late Winter Trend

I did some more research on my theory that the NOAA later winters hockey stick since 1990 is due to missing data. In the first graph below I plotted the number of valid days per month for the 100 stations with the fastest trend towards a later winter. As I expected, these stations have dropped from thirty days of valid temperatures per month in 1990, to six days per month now.

This confirms my suspicion that the apparent trend towards later winter since 1990 is largely due to data loss. If a station is missing temperature readings in October and November, that will cause the average date of the first freeze to come later.

Stations used in this analysis :