Walkability in Greater Vancouver Region

Discussion

    This project successfully merged a number of spatial datasets in GIS to generate measurements of the built environment which can predict walkability in the Metro Vancouver region. The comprehensive approach suggested in this project (“Walkability Index IV”) has proven to fit with the results achieved with previously developed indices. But additionally, it provides some differentiation, especially in the high walkability categories. The final smoothed map of this walkability index reflects the settlement structure of the GVRD with highly walkable urban cores in Vancouver and Burnaby and car-depended residential 'suburbs'. Inversely to the spatial distribution of walkability, CTs with higher socioeconomic status are located rather in the outer suburban areas. Thus, a negative correlation between walkability and socioeconomic status is shown for the GVRD. However, the global regression cannot fully explain the spatial distribution of walkability scores. This can be caused by the nature of a global linear regression and the data itself but it might also result from limitations to the approach proposed here.

    In the process of constructing and executing our project, we came across numerous uncertainties in our data and methods which limit the validity of our results.

Datalimitations

Land use data: categorization    

    As already described above, the creation of a measurement of land use mix turned out to be very complex. This is partly caused by the classification scheme used in the GVRD area. There are no separate categories for entertainment areas or office areas. Rather, areas mainly covered by office buildings (e.g. downtown Vancouver) fall under the category “Resource and Industrial”. As this category is also applied to other areas that do not contain destinations for pedestrians or are not walkable at all (e.g. factories), there might be some bias in the land use measure.

Land use data: up-to-date?

    The land use data used is from 2001, while the other datasets are from 2006-2010. It is likely that there have been some land use changes between these time periods which might have changed our results (e.g. the changing land use around False Creek due to the development of the Olympic Village 2010).

Missing data: Net retail area

    The net retail area has been employed in previous walkability indices as this measurement provides information about possible destinations for pedestrians (shops, but also places of local employment). Unfortunately, we were not able to obtain this data.

Methodological limitations

CTs as smallest spatial unit
    Census tract-level data was chosen as the most appropriate geographical scale in order to take advantage of the socioeconomic data available at this geographic level. In the 2006 Census, there are 2,500 to 8,000 people living in each Census Track (Census Mapping). This population based approach results in differences in the spatial extend of the CTs. While the urban, densely populated CTs are rather small, the rural, sparsely populated CTs are larger. These differences are to consider when calculating densities or proportions, as they are mainly dependent on the area of a CT. In order to provide an overview over the whole GVRD area, we included all CTs in the calculation of our walkability index, but excluded CTs with a population of less than 200 per square kilometre from the socioeconomic analysis. However, as the larger CTs dominate the visual impression even in the overview maps, they can distort the information provided by the maps.

Adequate measurements of walkability variables?
    Although the entropy score has been applied as measurement for land use mixture in many studies, there are some problems connected with this method: The entropy score only considers the distribution of land use types in an area but not the different land use categories itself. Brown et al. (2009) claim that the presence of walkable land use categories is more important for the walkability of a neighbourhood than the proportional distribution of different land use categories in that neighbourhood.

    The relative quantile classification scheme does not always reflect the nature of the data because it sometimes emphasizes the extreme values at the ends of the data range too much. Besides, the results allow only intra-urban comparison, no comparison with other cities.

Further Research

    The walkability index described above considers five environmental characteristics. However, there are many other factors of the built environment that affect walkability. As most important factor, the presence and the condition of sidewalks should be considered. Moreover, the walkability indices presented here focus on walking as a means of daily transport (to work or school or to do shopping), not on walking as a recreational activity where one has to consider other factors. The limited access to data, but also the extent of this student project prevented us from including more factors. Besides, most of the additional factors need to be studied into greater detail before they can be applied in a walkability index. Thus, the approach proposed in this project can be a starting point to a more detailed and comprehensive measure of walkability in the Metro Vancouver region. A measure that includes additional elements and is built upon the actual residential locations may overcome the limitations of our approach and provide a more robust prediction of the walkability in a neighbourhood.

    Such studies will be of growing importance in the future as their results inform decision makers about efficient methods to implement a sustainable development, to promote active forms of transportation and thus, to mitigate adverse public health problems.