After separating the residential layer from other layers such as industrial or commercial, the roads layer had to be removed from the residential areas in order to increase the accuracy of the layer. The buffers created around the roads were assigned in an arbitrary manner. This has the potential to influence the accuracy of the project. Or at the very least, a different set of assigned buffer values would produce slightly different results. Another point to consider is that highways and expressways will vary in their respective number of lanes during various segments. In order to remain consistent we have chosen to assign only one buffer value to each particular type of road network.

    In the process of editing the residential layer, removing the roadways created slivers as the polygons were split into smaller areas. If the slivers were not edited from the residential areas they would create artificial representations of city blocks (i.e. polygons in which no population is actually present). Aerial photos were placed beneath the residential layer in order to detect these slivers and delete them (Figure 18). We also sorted the area in ascending order in the table to find the very small slivers that were not visible in the previous method.

Figure 18

    Although residential areas were defined, within this area green space was still detectable. The residential layers were overlain with aerial photos of the same area in order to manually remove these areas from the existing residential layers. This improves the accuracy of the residential layers however a degree of error will still be present. Even after employing this technique, some areas of green may still have been overlooked or were simply considered insignificant and thus not removed. A more accurate method of removing the unclassified green space from the residential layer would have been to use a remote sensing software such as ER Mapper or PCI Geomatica to differentiate between vegetation and man-made structures. Unfortunately a limitation amongst the group members was that no one had enough experience in this area.

    After examining the postal code layer, it was determined that many postal codes contain a population of zero. These points were removed in order to increase the accuracy of the layer. After overlaying the residential layer with the postal code data, it was determined that we had more postal codes than polygons. Many postal code points did not fall within polygons as roads had been removed from the residential layer. This is another example which illustrates how error can be introduced as one attempts to create improved accuracy.

    Each postal code was assigned to a polygon based on proximity. A certain degree of error lies in this method as it is not possible to determine exactly which polygon each postal code belongs to. Upon examining the downtown area of Vancouver an abundance of postal codes lying outside polygons was noted. As a solution to this problem, we wanted to account for the mixed commercial and residential areas which exist in the downtown location. The GVRD residential layer contained this extra layer of information. Ideally one would not incorporate residential data from 1996 with residential data from 2001 however, the commercial residential layer in downtown is too important to be ignored. We did however choose to ignore the commercial residential layers which exist in other municipalities as these areas were minor.

    Another area in which we noted a discrepancy between postal codes and a lack of residential polygons was in the agricultural areas. The areas are classified as agricultural however people still reside in these areas. As a solution we chose to use the existing agricultural layer and attach our postal codes to these polygons. This still results in an accurate representation of population density as these are large areas which do not contain many inhabitants.

    One of the final stages in creating the project involved the creation of centroids. These centroids are created from the buffered/edited residential polygons. While the centroids do exhibit the population of the polygon, they are centered inside each polygon and do not represent where the postal codes are actually distributed within.

Methodology: Area of Study . Dataset . Modus Operandi . Cartographic Model . Map Discussion
                        Errors & Limitations

next ► Conclusion