NSF NCAR couples its Atlanta heat dataset to a weather model that accounts for buildings

"Offline" is the operative detail in how the first HUMID was built: an urban model its authors describe as simplistic, run against temperature and humidity measurements for 1980 through 2018, with no account of how the atmosphere interacts with the urban environment ^[5]. The coupled version solves the two together. The WRF-Urban extension carries building height into the simulation, which is what lets HUMID-Atlanta represent how wind moves around buildings and influences cooling ^[6]. The same air temperature does not feel the same on two blocks, because layout changes wind and shade ^[4].

The fidelity costs record length. The original spanned 39 years; HUMID-Atlanta covers 14, and the extension through 2025 brings it to 16 ^[16][8]. Nine years, 2010 through 2018, sit in both datasets ^[17]. For heat-wave epidemiology the count of extreme events sets the statistical power, and a 14-year window holds fewer heat waves than a 39-year one. The published account describes the data as high-resolution without giving the model grid spacing or a numerical comparison of old and new heat values ^[18].

Whether building-scale meteorology changes health conclusions is under test. Earlier work on kidney disease and heat exposure did not separate people living in city centers from those in suburbs or rural areas ^[13]. Andrew Newman, an NSF NCAR senior scientist and a co-author, said: "Not only is acute kidney injury a serious health concern, but the economic burden of treating those with it is substantial" ^[14]. He said the dataset and its successors "have the ability to fill major gaps in heat health research" ^[20]. That analysis is still under way ^[12].

Tzu-Shun Lin, the lead author and an NSF NCAR scientist, said: "This is a useful tool in actually quantifying how human beings experience heat in an urban environment and how factors like wind or humidity influence the impacts" ^[9][11]. The team is "already talking with other major cities and looking into the possibility of creating a version for the entire conterminous U.S.", Lin said ^[10]. Lin also sees the method being developed to forecast future meteorological changes and to feed early warning systems for extreme heat action days ^[15].