The potential of implementing superblocks for multifunctional street use in cities

Case-study selection

The case studies are chosen to include different cities around the globe that are commonly used for urban studies⁵⁰. From the 18 selected case-study cities, 12 cities are part of the C40 cities initiative, which aims to lead the way in urban sustainability. The selection ranges from large cities strongly following a grid street plan (for example, Mexico City and Tokyo) to smaller cities that are not dominated by a grid-like urban layout (for example, Zürich). This selection is not exhaustive, and comparatively more European cities are analysed. Barcelona is considered to test the methodology for the city where the superblock concept originates. The presented analysis can, however, be applied to further cities and is merely constrained by data availability.

Street network data processing

The case-study extent for each city is 25 km², for which the street network is downloaded with the help of Overpass API⁵¹ and represented as graphs with edges and nodes. For the graph-based algorithms and geometric processing steps, the python packages networkX⁵² and shapely⁵³ are used. The downloaded raw data are processed in several steps to obtain more accurate street length estimations. First, network nodes that are within a 15 m distance of each other are clustered to obtain a simplified network. This street network abstraction particularly reduces the complexity of street intersections. Second, closed detached rings (loops) and isolated subgraphs are removed from the street network as well as long (>300 m) tunnels and streets intersecting buildings. Further network cleaning is conceivable, depending on the case-study context, such as removing elevated streets. Third, very small network edges not forming part of longer streets are removed as they typically represent driveways (Supplementary Information section 2). The street network was not re-designed by extension. For example, the street network connectivity was not increased by adding new streets, which could potentially help to design more superblocks or miniblocks.

Before searching for potential superblocks or miniblocks on the street network, the street network nodes are classified into higher- and lower-level nodes. Whenever a node forms part of an intersection (degree ≥3), the node is classified as a higher-level node. A lower-level node forms part of a street (edge) between two higher-level nodes. Typically, the street between two higher-level nodes consists of multiple lower-level nodes and edges as the streets are typically not perfectly straight. The lower-level nodes are considered for calculating distances on the street network and creating the blocks. However, when checking for the network topology criteria to identify superblocks and miniblocks, only the higher-level street network is considered. This differentiation is necessary as otherwise, superblocks and miniblocks often would not be identified on the street network due to the lower-level nodes (Supplementary Information section 1).

Street hierarchy

The complexities surrounding superblock implementation are higher for streets that are essential to urban mobility. Contributors to OpenStreetMap can classify streets and assign different attribute labels to define a street network hierarchy. On the basis of the provided street hierarchy, the assumption is made that streets labelled as primary, secondary and trunk streets are not suitable for superblock design. Similarly, streets that form part of a trolleybus or tram route are excluded as potential superblock locations. For this analysis, all footways and private streets are ignored. Streets categorized as pedestrian or living streets are considered to be already not centred on car-based mobility and are ignored as the focus here is on transforming streets that are currently focused around car-based mobility.

Calculation of population density and building coverage

To go beyond relying on street network characteristics for detecting superblock design, density values are calculated for the entire street network. For each network node, average density values are calculated considering a radius of 100 m and then averaged per edge by averaging the density values from the starting and end nodes. Alternatively, the population density could be calculated by buffering the street network edges. The population density calculation is based on population data provided by the Center for International Earth Science Information Network and Meta⁵⁴, which are population estimates based on satellite data and census data at a ~30 m resolution. With the help of OpenStreetMap building footprints, the building footprint coverage is calculated first for every node on the street network considering the same radius of 100 m. Second, the average building coverage per edge is calculated by averaging the value from the start and end node of each edge.

Detecting superblocks and miniblocks on the street network

The street network forms the basis for locating potential superblock and miniblock candidates. Before searching potential candidates, the street network is characterized by street type, population density and building coverage as outlined in the previous two paragraphs for narrowing down the search. Then, the street network is first cleaned for cul-de-sac street elements, and the degree of all street network nodes is calculated. Second, all nodes with degree ≥3 are filtered as they could potentially be part of interior street intersections of superblocks or miniblocks. From this reduced filtered street network, all network cycles are identified as they could potentially be an interior street loop of a superblock. Isolated nodes with degree ≥3 not forming part of a network cycle or interior street loop of a superblock are further evaluated as a potential miniblock. Third, to find the exterior streets, all neighbouring nodes of the considered node(s) are identified, and a shortest-path algorithm is applied on the network where the considered nodes are removed to connect all the neighbouring nodes. If no path is found, the street network typology prevents the design of a miniblock or superblock. For miniblocks, all neighbours of the single node (single interior street intersection) are selected. Similarly, for superblocks, all neighbouring nodes not forming part of the interior street loop are connected along the shortest-path route to form the superblock polygon. If no path is found, or the path crosses a bridge, the node is not further considered. Finally, the geometric properties of the identified potential superblocks or miniblocks are checked to determine whether all the boundary conditions are fulfilled. In case a potential superblock does not fulfil the boundary conditions, the conditions for fulfilling a miniblock are tested. Table 1 lists all conditions, and the geometric scenario calculation is outlined in the next section.

Geometric scenario calculation

The geometry of the superblock is calculated on the basis of the assumed 400 m length of the Barcelona superblock (l_BCN) consisting of three individual blocks (block side length (l_{{mathrm{block}}} = frac{{l_{{mathrm{BCN}}}}}{3})). When identifying superblocks (S) or miniblocks (M), the following boundary conditions for the exterior street length (l_ext), interior street length (l_int) and length of the interior street loop or ring (r) (for superblocks only) need to be fulfilled:

where f denotes the deviation factor and z incorporates an uncertainty of ±20% of the Barcelona superblock and takes a value of 0.8 (min) or 1.2 (max). Example values for G₀ and G₁ are provided in Table 1.

Street NDI

The Edmonds–Karp algorithm is applied to assess the importance of each network edge concerning traffic flow across the street network. This approach is inspired by ref. ⁵⁵, where artificial ‘super’ sinks/sources are introduced. Super sinks/sources enable extending a network by adding a single node that feeds all the sources or drains all the sinks⁵⁶. The following steps are performed to assess the street network edge criticality. In a first step, four supernodes are projected in each direction to the network and placed outside the street network extent. Then, 100 helper nodes are equally distributed along a straight axis on each side of the network extent. An auxiliary linking edge is established between each helper node and the supernode. An additional linking edge is next added between each helper node and the closest node on the street network. A visualization and a more detailed explanation of this step are provided in the Supplementary Information section 3. In a second step, the Edmonds–Karp algorithm is run consecutively for each direction between opposing super sinks and super sources. The resulting flows of each simulation run are summed and averaged for each edge (i) to obtain the average flow ((f_i^{{mathrm{avg}}})) per edge. The average edge flow is then normalized ((f_i^{{mathrm{norm}}})) with the calculated maximum average flow value of the network:

To consider local as well as regional network impacts, these outlined steps are performed on a raster with a resolution of 2.5 km as well as for the entire case-study area. The calculations on the raster provide local flows ((f_i^{{mathrm{norm}}})) as outlined in the preceding. As the overall analysed city extent is 25 km², local flows are calculated across four regional cells across the city. In addition to these local flows, the same calculation is performed once for the entire case-study area using the 5 × 5 km² as the input for the flow calculations in equation (9). This calculation using the entire street network reflects flows at a higher geographical level and is termed (F_i^{{mathrm{norm}}}). In a third step, the two calculations are equally weighted and combined into a single indicator (NDI) to calculate the relative importance of each edge concerning traffic flow:

Edges with high NDI are edges that are critical to the street network; edges with low NDI have a low disruption potential as they do not form part of a critical network element and alternative paths exist for rerouting traffic. Calculating the NDI provides an approximate indication of the street disruption of a network towards urban mobility. The NDI is further used to derive classes indicating the street network disruption (low, middle, high). As identical geographical extents have been selected across our case studies, the obtained NDI values are comparable.

Source: Ecology - nature.com