Atmospheric scientist Andrew Dessler of Texas A&M University created a stir online at Twitter last year when he decided that a graph published by the EPA, and cited by Bjorn Lomborg a year ago “just didn’t look right.”
Science doesn’t follow, “just didn’t look right” as a rule of discovery or investigation, and as a scientist, Dessler should know better. Following a personal viewpoint of “just doesn’t look right” is a fast track to confirmation bias. Here is what Dessler said:
But, like many climate alarmists, Dessler just has a hard time believing that any time in the past could be worse than the present when it comes to weather. There has been so much rhetoric about the age we live in as being “hottest ever” and “unprecedented” that some people actually begin to reject contrary data that has been around for years.
In this case, Dessler would have us believe that the “dust bowl” period in United States history is nothing compared to the ravages of heat waves and drought experienced in the United States today, and he set out to prove his beliefs with a series of Tweets which you can read here.
Recently, during a discussion with ClimateDepot’s Marc Marano, Dessler brought up the thread from last year again.
But what are the real facts about the 1930’s heat wave? Is it merely a statistical aberration or a “cherry pick” as Dessler suggests? Or is it real? And if it isn’t real, but “cherry picking” of data, as Dessler contends, why would the EPA publish it?
Research yields the following undisputable facts about the 1930’s heat waves, contradicting Dessler’s viewpoint. First, the World Meteorological Organization defines a heat wave as five or more consecutive days during which the daily maximum temperature surpasses the average maximum temperature by 5 °C (9 °F) or more.
1. Yes, extended heat and drought did actually occur through a majority of the United States during that period. The year 1936 was particularly bad. Wikipedia notes this:
July 1936, part of the “Dust Bowl,” produced one of the hottest summers on record across the country, especially across the Plains, Upper Midwest, and Great Lakes regions. Nationally, about 5,000 people died from the heat.
https://en.wikipedia.org/wiki/1936_North_American_heat_wave
The U.S. National Weather Service says:
… the summer of 1936 featured the most widespread and destructive heat wave to occur in the Americas in centuries.
https://www.weather.gov/ilx/july1936heat
2. The heat waves of 1934 and 1936 were caused by natural ocean patterns, according to this peer-reviewed paper from 2015 which said,
Two ocean hot spots have been found to be the potential drivers of the hottest summers on record for the Central US in 1934 and 1936. The research may also help modern forecasters predict particularly hot summers over the central United States many months out.
Markus G. Donat, Andrew D. King, Jonathan T. Overpeck, Lisa V. Alexander, Imke Durre, David J. Karoly. Extraordinary heat during the 1930s US Dust Bowl and associated large-scale conditions. Climate Dynamics, 2015; DOI: 10.1007/s00382-015-2590-5
3. The Fourth U.S. National Climate Assessment (NCA) from 2017 did a detailed analysis of the United States heat waves, including cold and warm spells going back to 1900, and found the exact same thing the EPA did. Below is figure 6.4 from chapter 6. Note the very bottom panel.
The NCA added this, bold mine.
Since the mid-1960s, there has been only a very slight increase in the warmest daily temperature of the year (amidst large interannual variability). Heat waves (6-day periods with a maximum temperature above the 90th percentile for 1961–1990) increased in frequency until the mid-1930s, became considerably less common through the mid-1960s, and increased in frequency again thereafter (Figure 6.4). As with warm daily temperatures, heat wave magnitude reached a maximum in the 1930s. The frequency of intense heat waves (4-day, 1-in-5 year events) has generally increased since the 1960s in most regions except the Midwest and the Great Plains. , Since the early 1980s (Figure 6.4), there is suggestive evidence of a slight increase in the intensity of heat waves nationwide as well as an increase in the concurrence of droughts and heat waves.
https://science2017.globalchange.gov/chapter/6#fig-6-4
This 2017 NCA report clearly supports the EPA chart, and contradicts Dessler’s viewpoints of heat waves being worse in the present, and that the 1930’s heat wave spike was a result of some “cherry picking” of data.
4. An independent data analysis of heat wave events also contradicts Dessler.
The figure below is a result from 671 individual U.S. Historical Climatology Network (USHCN) stations with greater than 94 percent data availability. This is the number of days in “heat waves” lasting at least 5 days in length, in which each day is above the 95th percentile of the distribution of the daily values per date.
There were 671 stations which qualified for the 94 percent available data of the total 1218 USHCN stations.
Note that the Western United States (ID+OR+WA, CA+NV and AZ+CO+NM+UT) had their highest number of Heat Wave Days in the last decade, but the rest of the country was near/below the national average. Clearly there have been some regional changes in the last 110 years, suggesting weather pattern changes, but for the nation as a whole (gray bars), the decade of 1932-1941 still stands out as the warmest.
Interestingly, this is a very different result compared to Dessler’s claims, especially when it comes to some of the graphs he posted, such as this one in his Twitter thread:
Dessler’s graph indicates the dataset he is using (Berkeley Earth) has had some temperature adjustments applied. The difference is not geographic as the result from the USHCN stations is weighted by area also. We can get Dessler’s result, more or less, if we add 1°F to each daily temperature from 1980 to 2005 and then 2°F more from there to the end of the data. This gives a rough estimate of the magnitude of the adjustments that have been applied to the dataset he used.
The bottom line: unless you use adjusted data, you can’t get to Dessler’s results.
If Dessler wants to claim that the hundreds of observers that made high temperature measurements over decades should have that hard-earned data adjusted, he can claim that, but he ends up with a result he claims to abhor, one that is “cherry picked” by using adjusted data rather than the actual high temperatures that were recorded through history. Adjusted high temperature data is not the same as actual measured high temperatures, it isn’t even real.
Imagine if your evening news weather report, when reporting high temperatures for your area that day, used adjusted high temperature data to present the results to viewers. The backlash would be swift and vicious.
You can compare all of the facts in points 1-4 above to Dessler’s “just didn’t look right” analysis and make your own judgments about when heat waves were worse in the United States. You can believe the real actual data measured at the time, or the data that has but put through a statistical adjustment mill decades later.
“Heat waves (6-day periods…)
““heat waves” lasting at least 5 days in length”
“It don’t look right” has a place in science, as in it’s different to my back of envelope calculations. In this case, the equivalent is that the Dust Bowl era suggests more extreme heat occurred in the 30s even if just because it was very dry over a vast area. That it’s an inconvenient fact should not trigger the same gut feeling.
Many discoveries come from “that’s odd” or “that doesn’t look right” moments, but the point is that after that moment and before the discovery there’s a typically intense period of research. This guy seems to think he can miss out the middle step.
It’s one thing to apply a “that doesn’t look right” thought process to something you have personally produced, or which you can reproduce (as in an experiment).
Applying it to historical temperature records is pseudo-scientific malfeasance. The recorded instrument readings are “DATA.” Anything that alters that is no longer “DATA” at all.
For Australia, I have used both raw and ACORN-SAT adjusted daily maximum temperatures to analyse heatwaves in the 6 State capitals that host many of the longest historic T records as well as about 70% of our people now. Some measurements go back to the 1860s.
There are many definitions of heatwave. I simply calculate for each year the average Tmax temperature of all consecutive days for lengths of 1, 3, 5 and 10 days. The hottest each year go into a draw for Top 40 to make it simpler to see what is happening. Against my better judgement, I have inserted linear least squares regression trend lines to save wear on eyeballs doing much the same, as is common prectice.
What can be said? This is not a continent-wide survey. To do that requires data I do not have. It uses a simple definition of heatwave that does not involve, for example, Tmin or humidity.I insert a few dummy temperatures for missing data, then check that they probably make no difference to anything except ease of computong.
Apoart from those confessiions, the data seem to indicate a high sensitivity to cherry picking, rather like Kip Hansen finds for sea level changes. You can make assertions about these heatwaves that more or less contradict each other depending on how you cherry pick. Unlike USA, we did not seem to have dustbowl conditions at these locations in the 1930s.
The finding that interests me most is the temperature of the heatwaves. Taking raw data and 5-day waves, the usual temperatures of the Top 40 heatwaves are in degrees C:
Brisbane 34.8
Sydney 33.3
Melbourne 37.5
Hobart 30.2
The long-term average of all daily Tmax is:
Brisbane 25.4
Sydney 22.1
Melbourne 19.9
Hobart 17.0
The latitudes of these places is
Brisbane 27 degrees south
Sydney 32
Melbourne 37
Hobart 42
These 4 cities ate 350 miles apart down the east coast, getting further from the Equator and showing this in their ordinary daiuly average Tmax. But the average for the Top 40 heatwaves does not follow this simple decline as you move away from the Equator. Melbourne heatwave temperatures average 3 degrees C more than Brisbane. Part of the rhe reason is that the hottest, longest heatwaves develop many miles away from these cities, as in desert around Alice Springs (roughly). As they move to Melbourne, the temperature cools, with faster air mass movement giving hotter, shorter heatwaves in Melbourne aand slower air movement giving longer, less hot heatwaves to melbourne.
You readers ahould all be doing thius simple type of analysis for your own regions. The data are open to the public, the computing is dead easy. Here are 96 graphs to enage your analytical minds. Please tell us what you deduce from all this.
Geoff S
http://www.geoffstuff.com/sixcity2022.xlsx.
Any heatwave index that counts number of readings above a cutoff is just going to amplify half a degree higher readings on average, probably the result of a change of equipment in Australia or heat island effects. Good to see that you avoided that.
If you are going to do that to show that heatwaves are becoming more common rather than a bit hotter, you would detrend the data first or have cutoffs that increase with time. That this is ignored suggests that it’s deliberately propaganda.
RB,
Thanks for your input. I am happy to send you the raw digital data if you want to improve on what I showed. Are you able to calculate for your own region to extend this type of analysis?
The BOM were well aware of UHI when making the adjustments I show. Australia has no significant Time of Observation bias, US does. I stand by my method. There is some circularity in detrending T data sets by adjusting some T that cause the trend; you would need a method to remove the unwanted UHI trend without data to do it.
I do not do propaganda, I merely criticise others who do. Geoff S
The detrending is to avoid coming to the conclusion of more frequent heatwaves because the same weather pattern gives 0.2°C above the 90 percentile instead of 0.2°below, for example, if there has been a half degree increase in average maximum temperatures. You are not going to sense half a degree hotter but being told that the number of extreme days has gone up 10% sounds very much distressing. Still a half-adze job but good enough to nullify propaganda.
State high temperature records are still heavily dominated by the 1930’s.
Tom.1
If that is a comment about Australia, here are some old Sydney and Melbourne graphs of hot days that do not show a 1930s blip. There are possibly locations in central and north Australia which have them, but I cannot recall any so they might not stand out so much. Geoff S
http://www.geoffstuff.com/century_days_sydmelb.jpg
Sorry, no. I meant the US: U.S. state and territory temperature extremes – Wikipedia
That “feeling” Dessler had is called cognitive dissonance. It’s when your belief system smacks head on into reality.
Dessler, like Mann, is nothing more than an activist with a PhD who calls himself a scientist. They use models to manipulate data to achieve a predetermined desired outcome and call it science.
According to NOAA records, 24 of our US states set their current all time high temperature records in the summers of 1930 thru 1936. About half of the remaining states set their current records prior to 1930.
I don’t think the difference in his result is due to adjustments. The other datasets are also adjusted. He just picked a completely different measurement. The number of “hot” days in a year at every grid cell. Who knows how many different measurements he tried before he found one that looked right.
N.N.
Have a look at the graphs I posted. Half are adjusted, half are not. I do not know which are closer to actual, correct. Geoff S
You’re absolutely right this has nothing to do with adjustments. The reason for the differences between Dessler’s and Kunkel’s metrics is largely due to the fact that Dessler’s metric isn’t arbitrary. It actually plots how heatwaves have increased in gridded data. Kunkel’s metric considers only 876 thermometers without any area weighting. If those 876 thermometers represented the entire US, I suppose this would be fine, but but Dessler showed that they don’t.
Dessler publicly stated that if observations differed from model output the observations must be wrong.
Ah climate science… When the raw data doesn’t support the “consensus” make adjustments to the data.
Kunkel’s metric, which is on the EPA’s website and in the NCA4 report, is based on 876 thermometers. What Dessler showed is that those thermometers do not represent the US accurately. Kunkel’s metric is arbitrary, and arbitrary metrics yield arbitrary results. If you use gridded data, heat waves have been worse in recent decades than they were in the 1930s. And as his graphs show, it’s false to claim that “Dessler would have us believe that the ‘dust bowl’ period in United States history is nothing compared to the ravages of heat waves and drought experienced in the United States today.” In point of fact, his graphs show that the 1930s heatwaves were pretty bad. They’re just worse now.
I summarized Dessler’s actual argument and analysis here: https://woodromances.blogspot.com/2022/02/were-1930s-dust-bowl-heatwaves-worse.html