As part of his 2016 Congressional testimony, John Christy presented this chart:
I sought to reproduce the plot. I used the USHCN. The results I found match the pattern quite well, though the absolute numbers differ, I suspect, because of station selection. Rather than assemble the various transient stations, I used all the stations. And I excluded any year from a station in which more than 30 days of data were missing. Here is the result:
The station data used looks like this:
Where stations are one line each from top to bottom and years are left to right, from 1895 through 2014. Green indicates a year of good data and grey indicates no data ( or no station ). Obviously, station selection could conceivably change the temperature frequency. However, running the same analysis, but excluding all transient stations and all stations that were not already online by 1930 yielded quite similar results. The spatial distribution of stations in the US has been reasonably good and the spatial coverage has been reasonably dense, so for the US, there is some confidence in this distribution. For the globe, spatial coverage in 1930 was very poor and a global analysis of this sort may not be meaningful.
The number of stations included by year:
What’s interesting also is examining not just extreme heat ( max greater than 100F ) by also moderate heat ( max greater than 80F or 90F ):
Also, minimum temperatures:
Here is an animation of all the stations reaching 100F:
The year 1936 is a peak year of the Dust Bowl, but also drought in the MidWest. The effect on 100F days is visible:
There were some questions as to whether the USHCN data was suitable compared to the GHCN data. The resulting plot using the CONUS stations in the GHCNv3.22 data set produces very similar results:
The different stations are evident in the year by year animation of 100F stations:
The CONUS GHCN stations do appear more numerous and continuous:
Cold days from CONUS GHCN stations:
Here is the comparison of average 100F days versus global average temperature from NCDC:
The Time of Observation (TOBS) bias may be present in the USHCN and GHCN TMAX data sets. If the time of observation coincides with the time of maximum temperature, two days may indicate the 100F threshold. And there was a systemic shift from evening TOBS to morning TOBS in the data. For consecutive 100F days, this is not significant, but the bias persist, but TOBS can have an effect. There are two available measures to obviate or minimize this effect.
One way to obviate TOBS influence is to count the stations which exceed a threshold at least once. This approach is possible because the number of stations for CONUS is relatively constant. This measure indicates area of occurrence more than frequency. The patterns exhibited over time are similar to frequency. Below is the plot of the number of 100F stations plotted over the Christy graphic.
The number of 100F stations also indicates very weak anti-correlation with global average temperature:
And strong correlation with drought:
Another way to remove TOBS bias is to select for only stations which have night or morning TOBS. Below is the same chart comparison of 100F days using only US stations with a TOBS specified and between 10PM and 10AM.
The results are much the same as for the analysis using all stations and for the counts of single occurrence, though before 1930, there are fewer 100F days. That’s likely because there were few TOBS stations meeting the criteria then:
Similarly, the weak anti-correlation with global temperature and the strong correlation with drought reflect the patterns above:
- For CONUS, there is not evidence of global warming causing extreme heat
- For CONUS, there is evidence that droughts cause extreme heat
The GHCN CONUS TMAX data are contrary to claims of extreme heating from AGW. But when one considers the US MidWest average July latent heat flux( as seen below from NCAR reanalysis), upward latent heat flux ranges from 50 to 150 W/m^2. As a drought ensues and soil moisture tends toward zero, so too does latent heat flux tend toward zero. Anthropogenic greenhouse forcing, on the other hand, is on the order of 4W/m^2 for a CO2 doubling. The capacity of natural drought to cause extreme heat will always be much greater than the anthropogenic signal of global warming.
Finally, about attribution, extreme events attributed to AGW carry the emotional weight of the entirety of the event, not the margins. Radiative Forcing is a real phenomenon, and might contribute to margins of a heatwave, but the scale of RF, compared to natural variability, as demonstrated above, is not significant.