It has been a while since I have posted progress on our DOE-funded research into the Urban Heat Island (UHI) effect in the GHCN [Global Historical Climatology Network] station temperatures used to monitor land-based global warming.
It should be remembered that everything I post on this subject is (as is usually the case) a work in progress. [emphasis, links added]
The bottom line is that there are substantial UHI-based trend (warming) effects in the GHCN station temperatures; the only question is, how much have they affected temperature trends?
Most previously published research on the subject has suggested the effects are small (Hausfather et al., 2013; Wickham et al., 2013; Hansen et al., 2010; Parker, 2010; Jones et al., 2008; Parker, 2006; Peterson & Owen, 2005; Peterson, 2003; Peterson et al., 1999; Gallo et al., 1999; Karl et al., 1988).
As a result, you will find most who defend the “climate crisis” narrative will refer to one or more of those studies as showing the “science is settled”, and that GHCN-based land warming estimates are largely free of UHI warming effects.
I have argued that those studies involved methodologies that were not very good. Identifying the UHI effect is difficult.
I’ve come up with a novel way of quantifying the average UHI effect, even at stations that would be considered “rural” with presumably no UHI effect.
We have a paper in review at Nature Scientific Reports describing the methodology (my blog description of the methodology is here), but I have no idea what chance it has of being published.
I will get right to the results as they stand today. What I show below are for the all-station average of GHCN stations; they are NOT area averages, which are what is needed for climate monitoring.
They just show how much the average GHCN station is influenced by spurious UHI warming.
The stations cover the latitude bands from 20N to 80N but are dominated by U.S. stations (about 80% of the total) due to the huge number of stations we have in this country.
The plots are for four classes of initial GHCN station population density (the first year those stations started operating) during the warm season (May/June/July), and give the cumulative year-on-year temperature increase averaged across all stations in each of the four initial station population classes.
The adjusted (homogenized) GHCN station temperature changes are in green, and my calculated UHI effect is in red.
For the “wilderness to very rural” class (upper-left panel), the UHI effect on temperature trends turns out to be quite small, contrary to what I have recently argued.
Since many of these low-population stations are at high northern latitudes, this would suggest that the UHI effects on the large warming trends reported there are small.
But as we progress to higher population stations, we find that the UHI warming effect becomes larger.
In the highest population density class (“suburban to urban”, lower-right panel) my calculation of UHI warming is virtually the entire GHCN-reported warming signal since 1880, but only a small part of the reported warming since 1980.
If these results stand, what will they mean for reported land warming trends?
I’m guessing that the UHI effect on area-average trends since 1980 (the period of the most rapid temperature rise) will turn out to be relatively small.
But before 1980, it looks like the UHI effect on GHCN temperatures could be substantial.
This would change the nature of the global warming narrative, with little land-based warming for the first 100 years starting in 1880.
What could change these results? First, I do not account for increases in the UHI effect due to per-capita increases in infrastructure and energy use (buildings, vehicles, parking lots, electricity use, and resulting waste heat).
I assume the UHI effect is only a function of population density (partly because we have global gridpoint data on population extending back into the 1800s).
Thus, my UHI warming estimates might be a little low for stations where the population stopped growing but spurious sources of heat continued to increase, such as in Vienna, Austria (R. Bãhm, Climatic Change, 1998).
In any event, I feel like I am finally converging on useful results.
One aspect of this is that the record-high temperatures now being reported in major population centers in the southwest U.S. and southern Europe need to be revisited based upon the very large urban heat island temperature increases seen in the lower-right panel of the above plot at suburban-to-urban stations.
Read more at Dr. Roy’s Blog
There’s a particularly nasty effect of UHI that comes into play when the people running the surface stations realize that urbanization has taken what was once a suburban or rural station into the heat island. They then set up a new station further out from the urban core, run both in parallel for a while and determine the average delta between them attributable to UHI, then subtract that from the historical record of the old station and treat this adjusted-downward history as if it were the history of the new station. A few decades later, the new station’s neighborhood is urbanized, and they repeat the relocation, again adjusting history downward. This creates a warming trend out of thin air.