A New World City Warmth Island Dataset: World Grids of the City Warmth Island Impact on Air Temperature, 1800-2023


As a follow-on to our paper submitted on a brand new approach for calculating the multi-station moderate city warmth island (UHI) impact on air temperature, I’ve prolonged that preliminary U.S.-based find out about of summertime UHI results to world land spaces in all seasons and produced an international gridded dataset, lately overlaying the length 1800 to 2023 (each and every 10 years from 1800 to 1950, then annually after 1950).

It’s founded upon over 13 million station-pair measurements of inter-station variations in GHCN station temperatures and inhabitants density over the length 1880-2023. I’ve computed the common UHI warming as a serve as of inhabitants density in seven latitude bands and 4 seasons in each and every latitude band. “Temperature” this is founded upon the GHCN dataset per month Tavg near-surface air temperature information (the common of day-to-day Tmax and Tmin). I used the “adjusted” (homogenized, now not “uncooked”) GHCN information for the reason that UHI impact (interestingly) is normally more potent within the adjusted information.

Since UHI results on air temperature are most commonly at night time, the consequences I am getting the use of Tavg will overestimate the UHI impact on day-to-day prime temperatures and underestimate the impact on day-to-day low temperatures.

This then lets in me to use the GHCN-vs-population density relationships to world historic grids of inhabitants density (which prolong again many centuries) for each and every month and yearly since as early as I make a choice. The per month solution is supposed to seize the seasonal results on UHI (usually more potent in summer time than iciness). For the reason that inhabitants density dataset time solution is each and every ten years (if I get started in, say, 1800) after which it’s annually beginning in 1950, I’ve produced the UHI dataset with the similar annually time solution.

For instance of what one can do with the knowledge, here’s a world plot of the variation in July UHI warming between 1800 and 2023, the place I’ve averaged the 1/12 deg spatial solution information to at least one/2 deg solution for ease of plotting in Excel (I don’t have a GIS gadget):

If I take the 100 places with the biggest quantity of UHI warming between 1800 and 2023 and moderate their UHI temperatures in combination, I am getting the next:

Observe that through 1800 there was once 0.15 deg. C of moderate warming throughout those 100 towns since a few of them are very previous and already had huge inhabitants densities through 1800. Additionally, those 100 “places” are after averaging 1/12 deg. to at least one/2 stage solution, so each and every location is a mean of 36 authentic solution gridpoints. My level is that those are *huge* heavily-urbanized places, and the temperature indicators could be more potent if I had used the 100 largest UHI places at authentic solution.

Once more, to summarize, those UHI estimates don’t seem to be founded upon temperature knowledge particular to the 12 months in query, however upon inhabitants density knowledge for that 12 months. The temperature knowledge, which is spatial (variations between within sight stations), comes from world GHCN station information between 1880 and 2023. I then observe the GHCN-derived spatial relationships between inhabitants density and air temperature all over 1880-2023 to these inhabitants density estimates in any 12 months. The per month time solution is to seize the common seasonal variation within the UHI impact within the GHCN information (usually more potent in summer time than iciness); the inhabitants information does now not have per month time solution.

In maximum latitude bands and seasons, the connection is strongly nonlinear, so the UHI impact does now not scale linearly with inhabitants density. The UHI impact will increase relatively impulsively with inhabitants above barren region stipulations, then a lot more slowly in city stipulations.

It should be remembered that those gridpoint estimates are founded upon the common statistical relationships derived throughout 1000’s of stations in latitude bands; it’s unknown how correct they’re for particular towns and cities. I don’t know but how finely I will be able to regionalize those regression-based estimates of the UHI impact, it calls for a big quantity (many 1000’s) of station pairs to get excellent statistical indicators. I will be able to do the U.S. one by one because it has such a lot of stations, however I didn’t do this right here. For now, we can see how the seven latitude bands paintings.

I’m making the dataset publicly to be had since there’s an excessive amount of information for me to research alone. One may just, as an example, read about the expansion over the years of the UHI impact in particular metro areas, equivalent to Houston, and examine that to NOAA’s precise temperature measurements in Houston, to get an estimate of the way a lot of the reported warming development is because of the UHI impact. However you would need to obtain my information information (that are relatively huge, about 117 MB for a unmarried month and 12 months, a complete of 125 GB of information for all years and months). The positioning of the information is:


It is possible for you to to spot them through identify.

The layout is ASCII grid and is precisely the similar because the HYDE model 3.3 inhabitants density information (to be had right here) I used (ArcGIS layout). Each and every document has six header information, then a grid of actual numbers with measurement 4320 x 2160 (longitude x latitude, at 1/12 deg. solution).

Time for Willis to get to paintings.



Please enter your comment!
Please enter your name here

Share post:


More like this