I think his data supports the latter.

I saw this.

Two points:

1. As CO2 has increased... the actual number of hot days in the US has declined. https://t.co/nIlPSjtLQX

2. The claim here is clearly the urban heat island effect and NOT CO2.

Sorry climate bedwetters. @DLeonhardt

https://t.co/IxsbUXS3uv

— Steve Milloy (@JunkScience) July 23, 2019

and went to Boston Has Become New York: The reality of global warming by David Leonhardt to check and see whether his claim was being misrepresented. It isn't. From Leonhardt.

The temperature in Washington has topped 90 degrees for 12 straight days. While I was sitting inside during one of those days trying to avoid the heat, I spent some time making a chart. You can see it above.

It shows the average number of days per year when the temperature cracked 90 degrees in various cities, during the first eight decades of the 20th century (before global warming became more severe), and then in the past 10 years.

I chose 15 major cities from the National Weather Service’s database, without knowing exactly what I’d find. In four of the cities, mostly in the Midwest, the numbers are virtually unchanged. But in the other 11, there has been a substantial increase. Houston, for example, used to have 89 days above 90 degrees in a typical year; it now has 115. Atlanta has gone from 36 to 56, and Denver from 27 to 48.

By this measure, the Boston of today feels like the New York of the 20th century. Washington is on its way to resembling the Memphis of old. And Miami is more like Dallas used to be. (To see a larger version of the chart, click here.)

And the planet is only going to keep getting hotter until our political leaders — that is, Republican leaders, who are the obstacle to action on climate change — do something about it.

Without checking, I am happy to stipulate that the chart reflects the data.

There are at least four things wrong with his approach however. He is offering this data to illustrate the validity of the hypothesis that increasing CO2 concentrations are driving higher global temperatures.

His first error is pretty basic. His chosen sample does not represent the universe of the claim. What is happening in the US does not tell us anything about what is happening globally.

The global surface area is 510m square kilometers. The surface area of the US is 9.5 million square kilometers. In other words, the US is 1.8% of the global surface area. Leonhardt is assuming that anything happening in the US must be representative of what is happening globally. It is not.

That is bad framing of the problem, flying in the face of everything we know. Some areas experience prolonged periods of warming while other areas simultaneously experience long periods of cooling. You have to have comprehensive measurements of all locales to be able to determine whether there is net cooling or warming. As an example, this summer has been about normal for the UK, sweltering in continental Europe and below average in the extended landmass of Russia. You cannot take the measure from a single area and know anything about the whole.

The second error is that there is no rationale for his time-frame choice. When you look at time-series data, you can make it show anything you want to by choosing the start and stop dates. You have to have a reason for the chosen start date and the chosen stop date and Leonhardt does not provide a reason. That is at best sloppy thinking or lazy practice.

The issue of stop and start dates is especially important when you are looking at short time-frames for long time-frame events. Climate change is measured in multi-decades not years. A major volcano, a particular point in the solar-cycle, El-nino, can all affect climate temperatures for years at a time. You only see the long run change in multi-decadal averages. Comparing last week's stock market average against last month's stock market average is a similar error, given that the stock market tends to move over a 66 month business cycle.

The third error is Leonhardt's choice of measurement. Is number of days above 90 a good proxy for increasing AGW temperatures? No. Clearly no.

To see why, think about flu season. We hypothesize that there is a new flu strain and that this will be a worse flew season than last year. Which is the better measure, the total number of flu cases or the number of days when new flu diagnoses exceed a standard deviation above the long run norm, say 1,000 new cases a day? Clearly it is total number of flu cases. The flu season may last longer (which would drive up number of cases but never exceed the strike point) or shorter. It might be significantly below average for most the season but then have a couple of weeks sharply above. Alternatively, it might run close to the average all season and with only a normal number of above 1,000 new cases a day but the excess above 1,000 new cases a day might be so high as to push the total number of cases for the season well above norms even though the number of 1,000 new cases a day was normal.

Leonhardt has chosen a crude proxy which is dramatic (people notice really hot days) but which is not itself predicative of the overall heat trend lines.

The fourth error, and probably most profound, is that Leonhardt is conducting motivated research. He wants to find support for a given hypothesis rather than consider which hypothesis actually fits the data best.

In this instance, there are easily advanced alternative hypotheses for the data he presents which are a much better fit.

Click to enlarge.

Leonhardt attaches significance to the fact that in eleven of fifteen cities, the number of days above 90 increased in the past decade compared to the average for the eighty years prior. However, that of course prompts the question, if there is a global rise in temperatures why are the four cities in the midwest exempt. Which ought to refocus the mind on the issue of the sample not representing the whole.

But there is something even more obvious when you look at the cities. Where is there the largest rise in number of days above 90? Miami, Houston, Memphis, Denver, Atlanta, Dallas, etc. Which cities have grown the most in the past decade? Miami, Houston, Memphis, Denver, Atlanta, Dallas. What goes with urban growth? Increased impermeability, decreased vegetative cover, increased dense infrastructure (highways, buildings, etc.) - i.e. all the ingredients of urban heat island effect. Of course they are warmer.

My city of Atlanta has gone from 1.5 million in 1985 to 6.5 million in 2019 with all the attendant deforestation, housing and highway infrastructure, etc. Of course it is warmer in the city.

Interestingly, four or five years ago, someone did a review of the past century's temperature data for all measurement locations in the US. If you took the average of all locations, there was a very low, but steady, increase in temperature. However, if you separated urban from rural measurement locations and looked only at the rural data - flat line. All the heat rise was associated with urban heat island effects.

And as a counter test of Leonhardt's data, which cities have declined in population or only held their own? Chicago, New York, Minneapolis, San Francisco. The cities with little or no increase in degree days over 90.

Leonhardt isn't measuring AGW. He is measuring the urban heat island effect.

I have been arguing for the past couple of decades that we don't know, and frankly, might not be able to know, whether there is AGW. We have insufficient measurement coverage, there are inconsistencies with our means of measurement (land surface, satellite, tree ring, etc.), we don't have enough data, there are large areas of the globe which are effectively unmeasured, etc. All the concern about AGW is based on forecasting models of a complex system (climate) which are notoriously sensitive to quality of data, completeness of data, assumptions, and initiating assumptions.

Even if we had confidence in the models (which have consistently overestimated since they first came to prominence in the nineties), given the noise in the climate system, we won't actually know if the models are reasonably correct until a century or two from now. Probably long past the point where we can do anything about it.

My perspective is that we ought to be seeking to minimize all externalities such as CO2 as simply a matter of good stewardship but that there is yet insufficient data to shape long term policy with confidence. And simply doing something enthusiastically without being confident that it is effective is just as likely to make things worse. See the theory and practice of economic development for the past fifty years as an illustration of the negative impact of well-intentioned and enthusiastically implemented bad policy.

There is, however, a very clear aspect of climate change which is measurably real. Micro-climates, almost always based on local externalities (pollution) and land use. We know if we dam up rivers, creating lakes, we are going to change the ecology, the humidity, and the heat of the region. We know if we change farming practices, we will affect temperatures and riverine conditions. We know if we build a lot of impermeable surfaces and concrete buildings and highways, we will increase temperatures.

The forecast accuracy of all these activities are reasonable well observed, the mechanisms are reasonably well known as are the means to mitigate.

We could significantly improve the environment and living conditions by focusing on land use and pollution rather than on an as yet unproven model-generated forecast about the hypothesized role of CO2.

Why don't we do that? It forces hard trade-off decisions, it generates democratic engagement in decision-making and it reduces the money and power of the technocratic center.

Raising taxes and regulations for everyone else based on the choices of a backroom few is the solution for the AGW enthusiasts.

Making people make the best choices they can with the information they have given the things they want is the solution for addressing micro-climate change.

It is a choice between a determinist statism and a free, engaged citizenry making local choices.

Leonhardt is effectively arguing for the former. I think his data supports the latter.