Table 1 Key criteria guiding the global city selection for regional climate model datasets

Area size

A rough estimate was defined of the amount of CORDEX-CORE (0.22°) grid points (so-called gp022) occupied by the urban center, by considering its area (km²) in the GHS-UCDB divided by 600 km² (an intermediate value between the REMO and RegCM grid cell sizes). The selected cities have a minimum gp022 of 0.9, to represent cities with at least one gridbox (except for Europe 0.11 analysis, smaller cities included). This “a priori” estimate of the city size can be compared with the actual urban extent represented by each model.

Giorgi et al.⁵⁴

Teichmann et al.⁵⁶

Remedio et al.⁵⁵

Florczyk et al.²⁶

Population size

The Global Human Settlement Urban Center Database (GHS-UCDB) of cities 2015 was used for population size, containing information for more than 10,000 Urban Centers, regarding city location and extent, and providing a set of geographical, socio-economic, and environmental attributes.

Florczyk et al.²⁶

While this study uses the GHS-UCDB 2019 version, we acknowledge that a more recent release of GHS-UCDB is available⁸¹ that may be integrated in future versions of this work.

Geography

Three different geographic characteristics were considered: coastal, inland, and mountains. The city location and elevation from the GHS-UCDB were used to classify the geographical setting of each city.

Florczyk et al.²⁶

Climate zones

The Köppen-Geiger classification was used to define the climate zone of the cities. We relied on the existing, openly available global classification by Chen and Chen (2013), using long-term climate data (1901–2010) at half-degree resolution. Each city was assigned the climate classification of the grid point nearest to the GHS-UCDB urban center.

Chen and Chen

Florczyk et al.²⁶

Global balance

In line with the global cover of the CORDEX-CORE dataset, a set of cities covering all CORDEX domains was aimed for. A balance was sought between CORDEX domains that tend to have larger cities than others. A CORDEX domain was automatically assigned to the cities via their corresponding IPCC AR6 region.

Diez-Sierra et al.⁷¹

Climate impact

Smaller or less populated cities might still be subject to substantial climate risks or vulnerabilities. This aspect was handled in combination with the city size and global balance to fill gaps in domains.

Observations availability

At this stage, the availability of observations in the urban and surrounding areas was not a defining factor for city selection, as obtaining observations for cities and their direct surroundings remains a challenge at the global scale. In our study, the search for observations was inverted. We first selected the cities following the above criteria and then used the Global Historical Climatology Network (GHCN) database as an observational reference for those cities where observations of this network are available (section “Global in-situ observations for UHI evaluation”).

Menne et al.⁸⁰