Copyright@2023
8057 Zürich
In this vizualisation, we represent the accessibility to nearby green spaces using hexagons of 300 metres radius and colouring them into the Spatial Accessibility Score. This allows us to see which areas of the city have better accessibility to green spaces than others. Furthermore, based on the median of the Spatial Accessibility Score of the selected city, as well as of all cities that can be selected, a worldwide comparison can be made.
Disclaimer
For most cases, our map shows that areas with low spatial area index (SPAI) values are clustered, are further away from green spaces than areas with higher SPAI values. This makes sense considering distance is one of our factors of accessibility. However, this can some areas which appear to be close to green spaces are considered low in value. This can arise when areas of interest show high population densities which lead to higher competition between the population when it comes to green space capacity. A factor not considered are edge effect with border metropolitan areas. Therefore, there is a possibility of overestimation and or overestimation of green space accessibility of along city borders.
Due to its large natural reserves Brisbane City has a very high total area of green space available. Coupled with its big city area and therefore in comparison low population density this allows for an overall high SPAI. The borders and the city centre with their clusters of multiple smaller green spaces allows for clusters of high and very high SPAI values relative to the city’s median. In Brisbane, the port stands out with the large cluster of very low SPAI values and the airport next to it with very high SPAI values, with both having larger green spaces nearby.
Toronto overall shows strong patterns of low accessibility along the northern city borders and high accessibility within the city centre surrounding Sunnybrook Park. In addition, the green spaces in the city's interior are mainly smaller or tube-shaped. In the eastern part of the city, there are rather smaller green spaces, with smaller clusters of low SPAI values all around. In the western part of the city, the clusters of lower SPAI values are larger, with fewer small green spaces.
Osaka has large clusters of very low and very high SPAI values compared to the other two cities. Osaka has rather small green areas, although there are two larger green areas near the city centre and in the south. Otherwise, you can see in Osaka that around green spaces the SPAI values are very high and where there are no green spaces, there are very low SPAI values.
Using the local SPAI we can make statements on local distribution of green space accessibility. When comparing the median SPAI across three cities, differences in population density inherently create massive difference among each city in calculated SPAI. The median SPAI of the three cities is 0.025. Brisbane's median is 0.058, which is the highest median value in this city comparison. Toronto's value is slightly below the median value of the cities. Osaka stands out with a very small median. This is also reflected in the Spatial Accessibility vizualisation of Osaka in global view, where the clusters of low SPAI values are significantly larger than in the local view. Due to the rather similar medians of Toronto and Brisbane, compared to Osaka, no noticeable differences between the local and global views can be identified.
In this project, we developed a visualization of the accessibility to green spaces in three cities of similar population size and want to compare them with each other. The fundamental research question was: How do cities of similar population size compare in terms of accessibility to green space (based on SDG 11 target 11.7)?
We established two hypotheses, which can be shown in the visualization of the cities:
H1: There is uneven access to green space in cities.
H2: Population density affects the accessibility of green spaces.
The first hypothesis cannot be rejected due to the visualization, because in each of the cities, Brisbane, Osaka, and Toronto, it is not possible to show a uniform accessibility to green spaces. Nevertheless, there are clear differences between the three cities. Brisbane has very large green spaces and the median SPAI value of the city is the highest compared to the other cities. This could indicate that larger green spaces in or around the city on average favour the accessibility of green spaces for the population, while several smaller green spaces in the city tend to be worse. This is well illustrated in the comparison of Brisbane and Toronto. Osaka breaks out strongly into the lower range with accessibility. This city has visually fewer green spaces and has a very high population density compared to the global average. In Brisbane, the population density is rather low compared to the other cities, and the median value is very high. In this sense, the second hypothesis cannot be rejected. Here we can say that the population density of cities has an influence on the accessibility to green spaces. It may be that it is more difficult to plan for more green space in cities with high population density, because the space is mainly used for housing. To bring the benefits of green spaces to the population, other greening options of the city should be considered in the planning.
Our approach of assessing accessibility of green spaces comes with many limitations. We limit our assessment on what we consider the most influential and call for future work to improve upon them.
The biggest limitation of our project is the terminology and how they are defined. We argue that the way accessibility, green space and population is defined the results can be skewed. We argue that perception of green spaces – be it attractiveness and or capacity - varies by culture and geography and therefore suggest the inclusion of qualitative data in future studies to add semantic meaning to our mostly quantitative analysis of green spaces.
The issue of terminology propagates into the issue of data quality. As our definition of green space had to be fit to the tags used by OSM there is a bias as to what we consider green space and what we do not consider. This might lead to areas classified as green space which residents might not consider green spaces. We also must consider mappers bias and the possibility of incorrect tag use in OSM due to it being VGI data. As VGI has become more and more prominent it is getting more and more robust. However reassuring data quality is still hard[26].
Another big limitation is the lack of high-resolution census data for worldwide applications. As WorldPop uses a top-down method to interpolate the population across cells classified as settlements, areas that appear to be misclassified can lead to local anomalies in population density. This in connection with the MH3SFCA method can lead to over or underestimation for areas such as the airport area around Brisbane City.
As we did not consider a threshold when matching green spaces representative points to the nearest network node, green spaces can be matched to network nodes that exceed our defined walking distance of 300m. This can result in distances between a green space and a population node to be considered shorter.
In this analysis, a normalization of the great circle was used by means of the Gaussian curve, which adjusts accessibility as a function of distance. As a result, locations closer to a given point are weighted more heavily and have higher accessibility, while locations farther away have lower accessibility. The effect of this is that accessibility decreases linearly with distance, both within and outside the 300-meter distance. This leads to simplification and may overlook specific accessibility patterns.
Given the population data associated with the Great Circles, there is a risk of accessibility bias. The population data are used as weighting factors to account for the importance of the reachable places. If the size of the catchment areas is determined by the Great Circle, this can lead to bias. It may result in areas with low population density that are close to a densely populated area having higher accessibility.
In our analysis method, human behaviour was not considered in detail. For example, the 300-meter distance to green areas could already be too big for certain population groups, such as elderly people or people with an impairment. Furthermore, the choice of location is presented in a very simplified way. In this project, it is assumed that people will only use the services that are closest to them. Personal preferences are neglected here, whereby a person prefers to travel a longer distance to a green space that is better for him or her. In principle, this leads to an overestimation of the accessibility of certain green spaces.