In search of the Hockey Stick – part 2
In part 1 global temperatures have been represented as raw values. Other than the large variability of the global temperature measurements, both seasonally and location-wise, and some curious evolutions in the weather stations used to track them, there is little drama and no sign of a hockey stick.
Let’s look further. Most climate change presentations use the difference between a reference value at some chosen time and a related value at another time. This difference is called an ‘anomaly’ (*). There are a few issues with this.
First, what value ? Most graphs show a few, simple and clear lines. The only way to obtain this apparent simplicity is to smooth and average multiple and dispersed measurements into a single value.
Second, the reference value is set at a chosen time. The available measurements span 300 years or ~3600 months, giving as many possible reference times to arbitrarily pick from.
Third, one would prefer to compare reference and related values and that are actually comparable. In that regard, apples and oranges are more comparable than the thickness of tree rings and thermometer values or smoothed averages of different sets of unrelated weather stations across the globe.
Given these issues, let’s try something else here: a scatter plot of the difference between the temperature of month M in year Y for station S (the related value) and the mean temperatures of month M for station S over all available years (the reference value) versus the year Y.
The differences with the common representations are
- The plotted values are not averages, but individually calculated station measurements.
- There is not a single reference value obtained at a handpicked date, but timeless multiple reference values (one per station) imposed by data set itself.
- Reference and related values are fully comparable, since they come from the same station. Long term trends can only be established by long-lived stations.
The graphs below are produced from the dataset v2_mean.
It is not exceptional to see January temperatures that are 10C colder or warmer than the mean whereas July temperatures tend to stay within +/-2C of the mean. Due to the paucity of stations and their smaller latitudinal diversity in the southern hemisphere, no conclusions of the effect of seasons there can be made.
Calculating a difference between two uncertain values doubles the uncertainty. Temperature measurements have practical precision of at best 0.5C. Sure, sophisticated thermometers are much more accurate in theory, but the reality of a changing environment frustrates turning this accuracy into precision. This means that any difference less than 1C is not very significant in the large scale of things.
So the large temperature fluctuations from one year to another are real. Looking closely at the last decennia, 5-year ENSO cycles are clearly visible.
On the other hand, these plots reveal no clear long term trend and still no hockey stick…
(*) From Wikipedia:
an anomaly is any occurrence or object that is strange, unusual, or unique. It can also mean a discrepancy or deviation from an established rule or trend.
In this case the most neutral definition – a deviation from a trend – is applicable. One wonders, however, why a term with alarming, negative or authoritarian connotations was used.