More Anomalous Anomalies
The three primary producers of global near-surface temperature anomalies, NASA GISS(Goddard Institute for Space Studies), NOAA NCEI(National Centers for Environmental Information) and HadCRUT all begin the process with access to the same data. They then select a subset of the data, “adjust” the data to what they believe the data should have been; and, calculate the current “adjusted” anomalies from the previous ”adjusted” anomalies. NASA GISS alone “infills” temperature estimates where no data exist. Each producer then independently prepares their global temperature anomaly product.
The calculated anomaly in any given month relates directly to a global average temperature for that month. The difference between the calculated anomalies in any pair of months is thus the same as the difference between the calculated global average temperatures for those months. However, the differences reported by the three primary producers of global average temperature anomaly products from month to month, or year to year, are rarely the same; and, the changes are not always even in the same direction, warming or cooling.
The global average temperature anomaly differences reported by the three primary producers of global near-surface temperature anomalies for the months of November and December, 2016 are an interesting case in point. NASA GISS reported a decrease of 0.12oC for the period. NOAA NCEI reported an increase of 0.04oC. HadCRUT reported an increase of 0.07oC. Each provider estimates a confidence range of +/- 0.10oC for their reported anomalies. (NASA GISS: -0.22oC / -0.12oC / -0.02oC; NOAA NCEI: -0.06oC / +0.04oC / +0.14oC; HadCRUT: -0.03oC / +0.07oC / +0.17oC) Therefore, the confidence ranges overlap, suggesting that the differences among the anomaly estimates are not statistically significant. However, it is clear that the global average near-surface temperature did not both increase and decrease from November to December.
The second decimal place in each of these reported anomalies is not the result of temperature measurement accuracy, but rather of numerical averaging of less accurate “adjusted” temperature estimates resulting from data “adjustment”. The Law of Large Numbers states that it can be appropriate to express calculated averages to greater precision than the precision of the numbers being averaged, if the errors in the individual numbers are random. However, the nature of the factors which cause individual temperature measurements to be inaccurate and thus require “adjustment” suggests that the resulting errors are not random. Certainly, the “adjustments” made to the data are not random. Therefore, it is highly unlikely that reporting global average near-surface temperatures to greater precision than the underlying “adjusted” temperature data is appropriate.