The previous post looked at the potential effect of missing months on station annual mean tempertures and anomaly values where one month of data was missing and showed in the last figure just what a high proportion of stations have at least one month missing per year in recent years. Many stations, however, have much more than one month missing. Figure 1 shows the missing months in GHCN data from 1800. There may have been very few stations reporting in the 19th Century, but up to 1875, those that are present in the record had more than 90% reporting rate. The record is 85-95% complete up to ~1960 then drops rapidly: currently only ~70% of station data is complete in each year. The rest? Oh dear.
Some 15-20% of stations now have one month missing per year; 5-8% have two months missing; 2.5% have four months missing; at least 1% have 5 months missing. Remember that deriving an annual temperature requires at least six months of data (at least three seasonal means each composed of at least two months of data), and remember also that the number of reporting stations has dropped to around 1500 stations in recent years, so that is ~450 stations with at least one missing month in each year.