Method

Data Collection and Preprocessing

The first step of our project was to look for fitting literature and accompanying data. The selection of fitting literature, influence factors for living space selection and user groups was already thematised in the chapter 'Background ' . In the following text, the datasets that were chosen for each factor are shortly presented. To facilitate the index implementation, all factors were classified as negative or positive.
As ground data for the implementation the geometries of Zurichs municipalities were used (Amt für Raumentwicklung Stadt Zürich, 2023).
The link to each dataset can be found in the ' Literature ' section.

National Traffic Model (NPVM)

Source: Bundesamt für Raumentwicklung (2023)

The first dataset stems from the National Traffic Model (dt. Nationales Personenverkehrsmodell). The dataset provided ground data for 3 of 5 variables (- ) and therefore had a high influence on the outcome of this project. The dataset is made up of traffic zones, that were calculated from the density of work places and housing possibilities (for more information please consult the cited report).
For each zone the travel time to the closest centre of national importance by street transport and by public transport (), the travel time to the closest agglomeration or central municipality (regional centre) by street transport () and the travel time to the closest regional centre by public transport () are given.

For , the minimal travel time to the closest centre of national importance was chosen. Then the dataset was split into those three attributes, of which each was cropped to the extent of the canton of Zurich and rasterized for further calculations. - were treated as negative influence variables.

Visualization of minimal travel time to the closest national centre with
either public transport or by street (Bundesamt für Raumentwicklung, 2023).

Tranquility map of the swiss midlands

Source: Leeb et al. (2020)

The tranquility map was created by researchers at the ETH Zurich and shows how tranquil a place is. The researchers therefore considered several variables like noise exposure, closeness to infrastructure or landscape quality. The raster was used for calculations regarding . The pre-processing limited itself to cropping the raster to the extent of the canton of Zurich.
Unfortunately, two municipalities already lie in the pre-alpine region of Switzerland and therefore were excluded from our results, as no tranquility values for those municipalities were provided. was treated as a positive influence variable.

Visualization of tranquility in the canton of Zurich (Leeb et al., 2020).

Share of unemployed residents

Source: Statistical Office of the Canton of Zurich (2023)

The last dataset used provided the share of unemployed citizens per municipality. The dataset only consisted of a CSV, which then was linked to the municipality's geometries by a join. Other than that, no further pre-processing was necessary. was treated as negative influence variable.
The simple map on the right shows the data after being linked to the geometries.

Share of unemployed citizens in % per municipality (Statistical Office of the
canton of Zurich, 2023)

Data Processing

In the next step the layers were created using QGIS and ArcGIS Pro. This section briefly describes all steps that were taken for the index implementation.

1) After the rasterization of the datasets, all of the datasets below were normalised between 0-100. Thereby the score 100 was given to the highest values in the dataset, and 0 to the lowest.
Example: The pixel with the furthest distance to a national centre was classified as 100 in the negativ influence variable 'Distance from a centre with national significance'.

2) The next step included the calculation of the index for each of the user groups, by multiplying each factor by its weight (positive influence variables) or by its negative weight (negative influence variables). All of the layers were summed up and divided by the number of weights. This resulted in nine raster layers with the index numbers for each group.

3) As the goal was to have an index for each municipality, the raster layers were aggregated to the municipalities. Each municipality feature received the mean of the index numbers within its area. From this step nine vector layers, one for each user group, resulted.

4) The calculated index numbers are not yet interpretable, as they are still absolute. The final index is meant to be a relative assessment of a municipality's suitability as a living space for a certain user group, respectiv to all other entities in the target area. Therefore, the numbers had to be classified. While Fahrländer et al. (2014) decided on six classes, in the here presented approch only five classes were chosen. Three classes did not seem to display all necessary nuances. As a neutral class seemed to be useful, five classes appeared to be the best choice. As classification method 'equal interval' was chosen. From this step the first set of result layers with a Living Space Quality Index (LSQI) per municipality and user group emerged. The maps are visible in the 'Map ' section.

5) In the last step, the Diversity Map was created. It consists of the mean LSQI of all user groups and each municipality. The Diversity Map is visible as the last layer in the dashboard of the Map section.

Design

Parallel to the Data Processing, the design of the results was initialised. The main tasks were the creation of the dashboard and the website design. The dashboard (see under 'Map' ) was created using the dashboard builder of ArcGIS online. For the initial website design planning manual HTML and CSS scripts were set up. In a later state of the project and for the creation of the final webpage, the website creation was mitigated to 'teleporthq.io' (a HTML builder)