Circular reasoning in temperature adjustments
You’d think that the average weather of a location is climate? You’d be wrong. The real deal goes like this: there is something called climate which influences weather the effect of which we see in everyday readings such as temperature.
Speaking from a climatologic sense, both statements may be true. From a standpoint of measurement however, the second one makes no sense. It is false. The ‘climate’ is unknowable to us except via measurement of the weather points that constitute it.
This distinction is lost in several climate discussions and even amongst brilliant scientists. An outcome of the distinction, and a basic principle one might add, is that measurements need to be performed independent of effects the generated data are used to infer. Measurement precedes inference.
Standard climate thinking however has proceeded in opposite directions when it comes to creation of temperature series. Under this method, a local station will be ‘adjusted’ if it is deemed, by some metric, to not reflect underlying climate. (Whereas you might think underlying climate is inferred by a measurement of a given station)
I am stating nothing new here. David Stockwell presented the same basic earth-shattering logic (for climate science practitioners, that is) in this elegant write-up: Circularity of homogenization methods
He writes (emphasis mine):
If S is the target temperature series, and R is the regional climatology, then most algorithms that detect abrupt shifts in the mean level of temperature readings, also known as inhomogeneities, come down to testing for changes in the difference between R and S, i.e. D=S-R. The homogenization of S, or H(S), is the adjustment of S by the magnitude of the change in the difference series D.
When this homogenization process is written out as an equation, it is clear that homogenization of S is simply the replacement of S with the regional climatologyR.
H(S) = S-D = S-(S-R) = R
While homogenization algorithms do not apply D to S exactly, they do apply the shifts in baseline to S, and so coerce the trend in S to the trend in the regional climatology.
Stockwell further states (emphasis mine):
I would think the determination of adjustments would need to be completely independent of the larger trends, which would rule out most commonly used homogenization methods
There is proof confirming this diagnosis. Censorship-meisters Realclimate published a post by Zeke Hausfather on his paper co-authored with an US federal government NOOA employee.
They state the rationale underpinning the paper’s methodology:
Any major changes over time in individual stations that are not reflected in nearby stations are likely due to local (rather than regional) effects such as station moves, instrument changes, time of observation changes, or even such things as a tree growing over the thermometer stand. By removing any artifacts of individual station records not shared with other stations in their region, we can get a more accurate estimate of regional climate changes.
anomalies only work well IF the station records are not subject to localized changes due to non-climatic factors. In practice, at least over time spans of decades, this is rarely the case. So additional work (e.g. homogenization) must be done to remove any local perturbations that are not reflected in the regional climatology. Again, because longer-term climate changes occur regionally (not locally), and perturbation of a local record not reflected in other nearby stations is likely a non-climatic factor and should be removed if your goal is to calculate an unbiased estimate of regional climate changes over time.
This is circular.
An unbiased estimate of ‘regional climate change’ should emerge from well-curated, unadjusted temperature records of stations. Hausfather thinks the signal of climate change should be teased out by the guiding hand of adjustments made to stations.
What adjustments you do to a station, should have nothing to do with climate. Any reasoning that violates the above fails. Such adjusted series may per chance be representative of regional climate. But we would lack the means of knowing them to be so.
It is sobering to realize that most homogenization methods in use could be afflicted by this logic.