Home/Blog
John Christy and I just submitted a paper for publication in the AMS Journal of Applied Meteorology and Climatology called “Average Summer Heat Island Effects Relative to Landsat Impervious Surface Area for 50 Major U.S. Cities, 1985-2025“. You can tell what it’s about from the title.
As you might recall, a couple of years ago we published a paper relating GHCN station air temperatures to population density, and I made maps of the global distributions of UHI warming based upon those calculations.
But population density is not the best choice as a proxy for urbanization when it comes to the urban heat island (UHI) effect. The reason I used it is that there are global gridpoint datasets of population density going back to at least 1800, and how much UHI warming there has been in cities should be compared to what temperatures would be if the city was not there. The very first published study of the UHI effect was in 1818 by Luke Howard for London.
So, a couple years ago, I started a new analysis using Landsat satellite-based impervious surface area (ISA), which should be the single most direct physical influence on UHI warming. ISA measures the spatial coverage (in percent) of roads, parking lots, roofs, and many more types of human-built structures and modified soil surfaces (e.g. railways). The spatial resolution is 30×30 meters, and the dataset begins in 1985, now updated through 2025.
Methodology
I analyzed all of the June, July, and August GHCN station data within 30 to 100 km of the center of the 50 most populous Metropolitan Statistical Areas (MSAs) used by the U.S. Census Bureau. The following image shows those station locations.
This was a tedious, time-consuming process, partly because I also wanted to see if the UHI effect was magnified in these cities during excessively hot weather (warmer than the 97th percentile threshold temperature in Tmax). Determining that threshold from many stations in each of the 50 MSAs (which averaged 27 stations) was not straightforward.
I also simplified the methodology compared to our previous population-density related work: the example plots that follow are based upon how much each GHCN station’s daily Tmax or Tmin temperature departs from the average of all stations within the MSA (rural, suburban, and urban). This is basically what many of us experience on a daily basis as we commute from more rural locations to more urban locations, or drive across the city. Those individual station departures from the MSA-wide multi-station averages were then related to the impervious surface area centered on that GHCN station at 7 different averaging spatial resolutions (from 30×30 m to 10×10 km). The ISA averaging resolution producing the greatest magnitude correlation (positive or negative) was then chosen for further analysis.
The First City: NYC and its Strong Nighttime Warming, But Slight Daytime Cooling
It was quite a coincidence that two of the most extreme, and yet somewhat opposite, examples came from the two most populous MSAs: New York City and Los Angeles. The following plot shows how the New York City MSA GHCN station temperatures related to their impervious surface area.
As in virtually all of the 50 cities I analyzed, there was quite a bit of scatter in the resulting plots. This is because there are other influences I could not quantify, such as local spurious heat sources (brick walls, heat pumps, etc.), local microclimate effects, the use of only one “best” spatial averaging resolution out of seven, and station temperature sensor biases. For stations with substantial differences in elevation, I made lapse rate adjustments based upon regressions for that MSA.
For daily minimum temperature (Tmin) on non-excessive heat days (Fig. 1a) the average warming sensitivity to ISA across all stations was +3.1 deg. C per 100% ISA coverage. LaGuardia Airport looks like it has the most warming for its ISA coverage (over 3 deg. C at ~45% ISA coverage over a 2×2 km area). Central Park is also quite warm, as is Brooklyn and Bronx.
During excessively hot days (exceeding the 97th percentile threshold in Tmax for that station’s entire period of record) the UHI warming was amplified somewhat, increasing from +3.1 to +3.9 deg. C per 100% ISA coverage.
But for Tmax on non-excessive heat days (Fig. 1, panel c) there is a slight urban cool island effect (yes, it’s a real thing), at -0.5 deg. C per 100% ISA coverage. The daytime cool island effect is due to the same reason why the nights are so warm in urban areas: the much higher heat storage capacity of impervious surfaces such as roads and parking lots causes a time delay in the air temperature response compared to natural soil surfaces or vegetation. Just as it takes time overnight for roads and parking lots to release their extra heat energy, it takes time for those surfaces to heat up during the day from solar energy input after cooling off overnight. This creates a slight cooling effect during the day relative to rural areas (even though both obviously heat up during the day). But that effect seems to disappear on excessively hot days (Fig. 1d) where there is an average warming of +2.1 deg. C per 100% ISA coverage, albeit with a lot of variation between stations (low explained variance in the regessions, R-squared).
The relationships shown in Fig. 1 are strictly based upon spatial differences between stations, averaged across all summers, 1985-2025. A natural question is, how has the UHI effect changed over those years? While not in the paper, I’ve just started to look at this. In Fig. 2, I show how the +3.1 deg. C multi-station average Tmin warming per 100% ISA, when applied to the yearly average ISA in every county contained in that metropolitan statistical area, results in an estimate of how the Tmin warming has increased in time over the 1985 to 2025 period.
As can be seen in Fig. 2, all 22 counties in the New York City MSA experienced increasing impervious surface area, which then translates into increases in UHI warming. It should be emphasized that these calculations are based upon a single Tmin-vs-ISA relationship (+3.1 deg. C per 100% ISA) representing an average relationship across all GHCN stations (Fig. 1a). There is no way to know how accurate the individual county UHI warming amounts shown in Fig. 2 are.
The thick purple curve in Fig. 2 is the population-weighted average UHI warming increase across all counties. It is not very large: about +0.1 deg. C across 40 years, which is only +0.025 C per decade. So, while the UHI effect in NYC is quite large (+3 to +4 deg. C at several GHCN stations), since 1985 the increase in UHI has been rather weak, at least using the small increases in impervious surface area coverage as a proxy for increased urbanization. Other potential influences (e.g. increased localized spurious heat produced by more traffic and air conditioning) are not accounted for.
The Second City: Could Los Angeles Daytime Temperatures Decrease with Warming?
The second most populous U.S. MSA, Los Angeles, shows some stark differences compared to New York City (Fig. 3).
The biggest difference with New York City is in the daytime maximum temperature (Tmax). In Fig 3d we see see very strong cooling of the urban areas relative to the rural areas on excessively hot days, with an average relationship of -6.8 deg. C per 100% ISA coverage. This is likely because when hotter than normal summer air masses settle over the U.S. Southwest, the daytime thermal low pressure which forms inland, away from the Pacific Ocean, increases the strength of the sea breeze. As a result, the most urbanized locations (which tend to be nearer the coast) that are normally cooler anyway during the daytime (Fig. 3c) cool even more on days when the all-station average temperature across the Los Angeles metropolitan statistical area is well above the normal summer average temperature.
The other Pacific Coast cities I analyzed show similar results, although not quite as strong as L.A. Given this strong UHI-cooling effect from the sea breeze, there is no way to determine just how much daytime UHI warming (or cooling) is due to the presence of impervious surface area. The signal is confounded by the sea breeze since the more urbanized station locations tend to be closer to the sea breeze-influenced coast.
If you are confused at this point about how hotter weather causes the cooler temperatures shown in Fig. 3, remember that these plots (Figs. 1 and 3) are for how individual GHCN station’s temperatures depart from the all-station average temperatures. Thus, they are temperature differences, and say nothing about what the absolute temperatures are. So, while excessively hot summer days might actually be warmer than usual in downtown Los Angeles, there is a strong offsetting cooling effect from stronger sea breezes on those days compared to inland stations, which tend to be less urbanized.
On this point, I have looked at the official NOAA daytime Tmax temperatures for the summers in Los Angeles (LAX airport), and I do not see any evidence of this cooling effect; NOAA data shows the long-term warming at LAX has been at least as strong as at other southern California stations. I have no explanation for this discrepancy, other than the possibility that the homogenization process NOAA uses might make rather large trend adjustments to the LAX temperatures if the LAX trends depart from the surrounding stations farther from the coast. Homogenization is essentially a “voting” process and the few coastal GHCN stations might have gotten out-voted.
Regarding nighttime minimum temperatures (Tmin), Los Angeles shows a modest UHI warming effect during non-excessive heat days (+1.9 deg. C per 100% ISA, Fig. 3a), increasing to +2.3 deg. C on excessively hot days (Fig. 3d). Of course, the cooling effects of the sea breeze mostly disappear at night.
Thanks To Everyone, and The Future
The 50-city UHI study paper might be the last research article I publish if we are unsuccessful in getting a new contract to keep me employed. I have funding for several more months, then retirement might be a necessity.
As you already know, our “peers” who will review our current proposal don’t like us very much, and government funding of science contracts and grants is unavoidably political. The political winds in DC have just shifted strongly to the left, which will impact, directly or indirectly, congressional appropriations for what kinds of scientific research are deemed acceptable.
I’d like to thank everyone who has been contributing to this website on a monthly basis. I get an email every month for every contribution, along with the contributor’s name, and it always helps to brighten my day. The fact that people take the time to contribute means more to me than the actual $$ amount.
We will see how the funding situation sorts itself out in the coming months.
