*3.1. Vertiport Networks Based on Commuting Trends*

Air mobility operations may be differentiated between urban air mobility inside city limits, sub-urban air mobility connecting city and surrounding metropolitan areas (trip exceeds 20 miles (32 km)), and regional air mobility providing city to city transport [81]. Depending on the operation type, different repercussions on vertiport location, size, resource provision and operating concept may be expected. Historic commuting behavior can be used as a starting reference to evaluate where and to what extend air mobility may serve mobility needs. Once the need and potential demand is evaluated, a suitable location has to be defined for each vertiport of the network; on the one hand a vertiport needs to be conveniently reachable, on the other hand the amount of vertiports should be reduced to the most needed.

Developing theoretically a vertiport network may consider "uncapacitated" and "capacitated" facilities. The use of "uncapacitated" facilities makes sure that individual vertiports are not causing any operational bottlenecks during analysis and, therefore, are able to serve unlimited demand (see e.g., [81]). Instead, "capacitated" vertiports only serve limited demand (see e.g., [81,83]).

For the U.S areas *San Francisco Bay area*, and *Salt Lake City-Provo-Orem*, *Dallas-Fort Worth* and *Washington-Baltimore-Arlington*, the UAM market potential was investigated considering a multi-modal transportation network in which UAM provides single legs of a commuting trip [83,84], respectively. Further, ref. [81] analyzes a sub-urban air mobility vertiport network setup in *Miami (U.S.)* based on work-home trip data-sets. A data driven optimization framework for defining and solving the Mixed-Integer-Programming based network problem was used while targeting to minimize the vertiport network setup costs. Lastly, ref. [85] established a six-piece vertiport network in *Islamabad (Pakistan)* focusing on vertiport site selections next to frequently used commute routes and places where traffic congestion is faced.

In order to reflect different time saving requirements and to develop resulting vertiport performance constraints, ref. [83] proposes to cluster commuting travellers into long distance commuters and short distance commuters. For long distance commuters a time saving of 25%, and at least 50% for short distance commuters is required due to their

different value of time in order to switch to the UAM mode. The demand and vertiport distribution problem is formulated as an uncapacitated facility location problem which uses k-means algorithm for clustering.

This k-means approach was also used by [86], who investigated a vertiport network of 10, 40 and 100 vertiports in the metropolitan area of *Seoul (Korea)*. Areas like Han River Park, highway intersections, rooftops of parking lots and existing helipads on skyscraper rooftops have been utilized for vertiports. In order to evaluate how well the data is clustered, the silhouette technique is performed. Final vertiport locations are selected by re-positioning them to the appropriate sites near the centroid of the cluster to comply with geographical conditions. This caused frequent challenges due to most of the clusters are being residential areas.

Another "clustering approach" was defined by [81] who introduces the concept of a "catchment area" (3 miles (4.8 km) radius) where vertiport locations are paired up. The resulting time saving based on different numbers of work-home blocks and vertiport pairs is analyzed for the operating area of *Miami (U.S.)*. The larger the catchment area the bigger is the number of potential vertiport locations and routing options which then requires less vertiport pairs to satisfy the demand. On the contrary, larger catchment areas impose longer egress and access legs for the customer.

Since a change of transport modes is inevitable when considering a multi-modal transportation network, increasing overall time savings always asks for optimized transfer times between subsequent modes of transport.

Transfer times of 5, 10 and 15 min and varying numbers of vertiports (1 to 30) are considered for the *San Francisco Bay area (U.S.)* by [83]. The direct haversine between the origin and destination of each trip is computed and compared to the travel time on ground based on different ground traffic congestion levels extracted from the *Mobiliti* simulation by [87] and *Google Maps'* API. Focusing on short distance commuters, even if high transfer times of 15 min and high ground congestion are assumed, 45% of the short distance commuters in the *San Francisco Bay Area (U.S.)* will benefit from switching to UAM. However, it requires a rather large network of 30 vertiports in the east and 24 in the west. This benefited commuting share drops significantly to 3% if uncongested traffic and 10 min transfer time is assumed. A smaller network of 29 vertiports in the east and seven in the west is required instead. By contrast, no benefit is created if transfer times of 15 min and uncongested traffic are assumed. Additional time-saving and efficiency analyses about choosing UAM instead of ground taxis were conducted e.g., for New York City (U.S.) and Hamburg (Germany), and parameters affecting UAM mode choice were analyzed for the city of Munich (Germany) by [88–90], respectively.

Potential vertiports in the U.S. cities *Salt Lake City-Provo-Orem*, *Dallas-Fort Worth* and *Washington-Baltimore-Arlington* were examined by [84] and resulted into potential vertiport network sizes of 38, 407 and 207 vertiports, respectively. Census data and tracts are used to approximate the vertiport location in the centroid of census block groups. Those networks generated by different heuristic methods such as elimination heuristic, maximal edgeweighted subgraph heuristic, greedy heuristic, greedy heuristic with updates are compared. 1200 different cases are explored differing in input variables such as location, network type, battery range, number of vertiports and vehicle speed. Overall, the two greedy algorithms with update steps concluded as best-performing algorithms and produced solution networks with 91% of the optimal value. When selecting optimal vertiports the interdependence of vehicle attributes, potential locations, and desired network structure was considered.

Rather uniquely in this set of vertiport-network-publications, a noise analysis around the UAM route is performed on the basis of the day-evening average sound levels for the vertiport network in *Seoul (Korea)* [86]. To measure the percentage of the population affected by noise, a curve fitting function of the Shultz curve is used. By dividing the area of Gangseo-gu into hexagonal tiles, according to [86], noise will affect roughly 400,000 people in the 41.6 km<sup>2</sup> area. Due to the lack of eVTOL noise data, noise maps are created by using an aviation environmental design tool and by assuming noise characteristics of a five-seat helicopter. A noise priority scenario defined as a flight along the least populated area was compared to a business scenario describing a flight following the shortest distance; the number of affected people decreased by 76.9% for the noise priority scenario.
