Define subnational units for projections

Chris Jacobs-Crisioni primarily undertook this task, with our team providing support in specific areas. The approach outlined in Dijkstra and Jacobs-Crisioni (2023) was used to define the subnational units. We contributed by assisting in developing a functional global road network utilising OpenStreetMap data.

Network Model World

To develop a globally navigable road network, we initially obtained the complete OpenStreetMap dataset in .pbf format from www.geofabrik.de for each continent. We imported these datasets into GeoDMS and converted the necessary attributes into the GeoDMS File System Storage (.fss) format. We then sifted through the ‘other_tags’ field to identify key details such as one-way roads, ferry routes, vehicle accessibility, and maximum speed limits. Using this enhanced dataset, we isolated the network applicable to automobiles. Our subsequent task was to refine this network by eliminating subnetworks, ensuring that all potential origins and destinations could be efficiently connected. To improve processing speed, we iteratively refined the network with a custom algorithm that removed unnecessary junctions, dead ends, duplicate elements, and other non-essential links and nodes.

Additionally, during this project phase, we resolved several other technical challenges, including issues related to coordinate conversion.

Implement models to project subnational built-up land, population and/or built-up area per capita change.

Marcello Schiavina calculated the projected subnational built-up land and population into the functional areas developed in Task 2.1. He provided these projections in an Excel sheet, which was then imported into the 2BURP model as claims (Figure 2).

These claim tables were then attached to geographical data of the functional areas to be used in the model and cast into relevant metric units (Figure 3).

Final subnational units set generation

Subnational units were created following the Functional Rural Area process. As origins, we select the centroids of settlements in 2020 according to the GHSL 2019 release. But only catchment areas were generated for settlements with at least 50,000 inhabitants that are the biggest within a 30-minute drive. Four levels of functional areas were generated. Results from these sets were chosen per country to limit the number of subdivisions in very populous countries, while ensuring that geographically very large countries with limited population are divided into multiple subnational units. The following criteria were used and selected:

  • 2M inhabitants target population, 120 minutes minimum, 240 minutes maximum driving time (used for 163 countries).
  • 4M inhabitants target population, 120 minutes minimum, 480 minutes maximum driving time (used for 54 countries, namely MAR, AFG, AGO, ARG, ARUS, AUS, BGD, CAN, CHL, CIV, CMR, COD, COL, DEU, DZA, EGY, ERUS, ESP, ETH, FRA, GBR, GHA, IDN, IRN, IRQ, ITA, JPN, KAZ, KEN, KOR, MDG, MEX, MMR, MOZ, MYS, NGA, NOR, PAK, PER, PHL, POL, SAU, SDN, THA, TUR, TZA, UGA, UKR, USA, UZB, VEN, VNM, YEM, ZAF).
  • 4M inhabitants target population, 120 minutes minimum, 240 minutes maximum driving time (used for 17 countries, namely BLR, BOL, MLI, MRT, NER, NPL, NZL, PNG, PRK, ROU, SEN, SOM, SSD, SWE, TCD, TKM, ZMB).
  • 4M inhabitants target population, 240 minutes minimum, 480 minutes maximum driving time (used for 3 countries, namely BRA, CHN, IND).

Five small countries with less than 2M inhabitants consisted of multiple, relatively small functional areas. These countries were COM, CPV, GNQ, LVA and ESH. The functional areas in these countries were grouped into single functional areas per country.

Finally, the functional area algorithm occasionally yielded erratic geographies and implausible exclaves, typically in low-density areas with capricious shorelines, islands and navigated rivers. Where the algorithm created such distortions, functional area boundaries have been adjusted manually. This has been done in the Amazon estuary in Brazil and the border of MEX and GTM. Similar adjustments have been made to the surroundings of Ouésso, Congo; the southern tip of the CAR; the northeastern border of ZAF with eSwatini; areas in SEN, MRT, and MAR. This has also affected areas in Bangladesh, India, Pakistan, Iran, Azerbaijan, China, Russia, Spain, Germany, Norway, the UK and the Czech Republic. In addition, a number of very small functional areas in Bangladesh and India have been merged with larger surrounding functional areas. For an overview of manual edits, cf versions fas_combined_20240627_54009.shp and fas_combined_20250110_54009b.shp (except for the manual changes in Brazil, which have been done much earlier in the process).