A vertiport is a compact airfield designed for aircraft that take off and land vertically—such as electric air taxis. But what happens when a city has population, road, and airspace data yet lacks reliable information about actual travel patterns? How can we determine the best locations for vertiports under such constraints?
This is precisely the situation in Lagos, Nigeria, and many other rapidly urbanizing cities. The most straightforward approach is to use population density as a proxy for demand, identify the busiest areas, and propose vertiport sites there. To test this hypothesis, I used a population raster—a grid of squares, each with an estimated population count—converted each square to a point weighted by population, and applied K-means clustering, which groups nearby points and places a central point in each cluster.
The initial results spread across the city's most densely populated areas. At first glance, the output appeared reasonable. However, a closer inspection revealed significant flaws. In the left panel of the figure below, the population-only result is shown; the right panel displays the refined version after additional screening rules were applied.
The model was not failing randomly—it executed exactly as instructed, optimizing for population concentration. The core issue was that population density alone is an incomplete objective. It fails to account for whether a location is accessible, buildable, or safe for aircraft operations.
The initial result is not a usable answer. When I checked the proposed sites against two simple rules, some fell within a safety buffer around Murtala Muhammed International Airport, while others landed on water or wetlands. The model identified clusters of people, not viable sites. This gap—between where people are and where a vertiport can actually be built—is the central challenge this article addresses.
A Repeatable Methodology
This article presents a repeatable method that starts with population as a proxy for travel demand, given the absence of actual trip data. The process then incorporates proximity to roads and transit hubs, excludes locations that fail defined screening criteria (such as water bodies, wetlands, mangroves, and airport or military buffers), clusters the remaining points using real-world distances rather than raw map coordinates, and verifies each final site after any spacing adjustments. The output is not a construction blueprint but a ranked shortlist that readers can reproduce and investigate further.
Lagos serves as a valuable case study due to its data limitations. There is no comprehensive public dataset showing inter-neighborhood travel patterns, so this article demonstrates how to build a practical first-pass model without conflating population counts with travel demand. The airport and military buffers used here are illustrative, not official airspace maps; a real deployment study would replace them with geometry provided by aviation authorities.
The Short Version
The approach follows four key decisions:
- Where are the people? Use a population map as a baseline estimate of potential demand.
- Which places are easier to reach? Apply a small boost to locations near major roads and transit hubs—this is the accessibility score, a simple measure of transport connectivity.
- Which places should be ruled out? Exclude water, wetlands, mangroves, and locations within configured airport or military buffers.
- How do the remaining places get grouped and ranked? Use K-means clustering, as defined above, then evaluate and rank each resulting site.
The order of these steps is critical: clustering algorithms excel at finding concentration but are blind to constraints like airports, land ownership, flood risk, or other operational factors. By embedding these rules before clustering, we ensure the final recommendations are both data-driven and practically viable. This methodology offers a scalable template for cities worldwide facing similar data gaps.