The DOE Report on Urban Heat Islands

I've written a lot on the effect of urbanization and urban heat islands (UHI) on GMST and CONUS trends, and I don't want to duplicate those efforts here, but I think this section of the DOE report deserves its own post. The DOE report (authored by the Climate Working Group) has a section on "Urbanization influence on temperature trends" on pp. 20-21 (section 3.3) that engages in a bit of rhetorical slight of hand that I find dishonest. In particular, their treatment of UHI involves two studies authored or co-authored by members of the Climate Working Group (CWG), but the report misrepresented one and casually dismissed the other. The report misrepresented a study by Roy Spencer and casually dismissed a paper co-authored by Judith Curry, despite the fact that this study provided conclusive evidence against the position taken in the DOE report. I consider this particularly dishonest. Their treatment of UHI begins with:

Historical temperature data over land has been collected mainly where people live. This raises the problem of how to filter out non-climatic warming signals due to Urban Heat Islands (UHI) and other changes to the land surface. If these are not removed the data might over-attribute observed warming to greenhouse gases. The IPCC acknowledges that raw temperature data are contaminated with UHI effects but claims to have data cleaning procedures that remove them. It is an open question whether those procedures are sufficient (p. 20).

I find it odd that CWG chose to use the terminology of "data cleaning procedures" for the removal non-climatic biases. There's no reason not to use the proper terms "bias correction" or "homogenization," and these procedures have been demonstrated to effectively remove urbanization biases (see below). What I find dishonest is their claim that Spencer's paper from 2025[1] found a significant UHI bias in U.S. temperatures. In the words of the DOE report,

Spencer et al. (2025) used newly available historical population archives to undertake such an analysis and found evidence of significant UHI bias in U.S. summertime temperature data (p. 21).

This is absolutely not true of published datasets for CONUS temperatures, which undergo bias correction prior to publication, and Spencer explicitly acknowledges this multiple times, including in the paper itself. Spencer's paper only reviewed raw data prior to bias correction, and the paper explicitly says this: "It is beyond the scope of this paper to determine how much (if any) urban warming remains in the adjusted (homogenized) GHCN data; here we will compute UHI effects from the raw (unadjusted) version of the dataset."[1] Spencer acknowledges this again on his blog post about this paper. He said that when the paper was undergoing peer review, he was told to use raw data instead of homogenized data because homogenization removes the bias that Spencer was looking to find. Here's an excerpt:

What Does This Mean for Urbanization Effects in the Official U.S. Temperature Record?

That’s a good question, and I don’t have a good answer.

One of the reviewers, who seemed to know a lot about the homogenization technique used by NOAA, said the homogenized data could not be used for our study because the UHI-trends are mostly removed from those data. (Homogenization looks at year-to-year [time domain] temperature changes at neighboring stations, not the spatial temperature differences [space domain] like we do). So, we were forced to use the raw (not homogenized) U.S. summertime GHCN daily average ([Tmax+Tmin]/2) data for the study.

I have more to say on this in my review of Spencer's paper, but the DOE report is clearly indicating that Spencer's paper at least helps to answer the "open question" about whether "data cleaning procedures" are "sufficient" to remove urbanization biases, and the DOE report indicates that Spencer's paper shows significant bias remains. But Spencer has admitted on multiple occasions that his paper does not even try to address that question. Spencer's paper also did not use any area-weighting when he quantified the effect of urbanization bias on raw temperature station data. In my review I calculated that given standard estimates for the percentage area of rural, peri-rural and urban areas, Spencer's method showed only a ~10% increase in trends in CONUS raw summer temperatures. Since UHI is strongest in the summer, if he used annual means, that value would be even less. If he used homogenized temperatures, the bias would be effectively eliminated.

These results have been confirmed by numerous studies, but the DOE report appears to give only passing references to these and then casually dismiss them without serious consideration. The following cites Wickham et al 2013[2], which was co-authored by Judith Curry:

Some studies providing evidence against UHI contamination compared warming rates between rural and urban locations (Jones et al. 1990, Peterson et al. 1999, Wickham et al. 2013). It is not known whether such methods would be capable of detecting UHI bias even when present (p. 21).

Since Judith Curry was a co-author of Wickham et al 2013, you would think she would be familiar with the contents of that paper. The paper identified stations globally in the Berkeley Earth dataset that could be classified as "very rural" and compared these to all land stations. They found that "very rural" stations warmed marginally more rapidly than "all" stations.

Very Rural Stations Warm More Rapidly than All stations in Wickham et al 2013

Since the DOE report cites Wickham et al 2013, and since Curry was a co-author on the paper, I think they need to respond with more than just a casual dismissal of these findings. And the findings of this paper have been confirmed by later analysis. Robert Rohde at Berkeley Earth extends this comparison to the end of 2024.

From Robert Rohde on X

As you can clearly see, a comparison of "all stations" vs "very rural" stations shows virtually the same amount of warming, with the largest differences occurring prior to 1890 or so. So if you're concerned that global warming (or CONUS warming) trends have been affected by urbanization, just use Berkeley's "very rural" subset of weather stations. You'll get virtually the same results. What the DOE report considers an "open question" was actually conclusively answered by a study co-authored by Curry. 

This point can also be seen with global maps of anomalies. Below I show how much the last decade has warmed above the 1941-1970 mean. If urban heat islands were adding a strong bias to global warming, we'd see the darkest red in highly populated areas, but that's not the case. The Arctic is warming the most rapidly despite being largely devoid of cities.

It seems that at least one of the scientific studies cited by the DOE report actually conclusively answers the "open question" mentioned in the opening paragraph, and Spencer admitted that his 2025 paper didn't even address the this question. I know of no rebuttal to the above analysis by Wickham et al, especially when updated through 2024. There is no evidence of any urbanization bias adding a significant amount of warming to either global or CONUS temperature trends.

References:

[1] Spencer, Roy W., John R. Christy, and William D. Braswell. "Urban Heat Island Effects in U.S. Summer Surface Temperature Data, 1895–2023". Journal of Applied Meteorology and Climatology (published online ahead of print 2025). https://doi.org/10.1175/JAMC-D-23-0199.1.

[2] Wickham C, Rohde R, Muller RA, Wurtele J, Curry J, et al. (2013) Influence of Urban Heating on the Global Temperature Land Average using Rural Sites Identified from MODIS Classifications. Geoinfor Geostat: An Overview 1:2. doi:10.4172/2327-4581.1000104
https://www.scitechnol.com/2327-4581/2327-4581-1-104.pdf