(1) Limited data availability
Unfortunately, we had limited access to data from Singapore. Even if some sort of spatial data was available, they are often "locked" as they come in the form of images or PDFs. We had to georeference these images and digitise them to convert them into machine-readable formats in ArcGIS. While all efforts were made to ensure that the georeferencing and digitisation were done accurately, it is inevitable that some uncertainty has arisen due to these processes.
In addition, the distribution of waste across planning areas (used in Part 2) was reliant on many assumptions. While they were made in order for us to proceed with the analysis, our project would have been more credible if an accurate distribution of waste in Singapore had been available.
We also recognise that there may be other factors - known or unknown - that are not considered in this model. For example, in our cost analysis we have assumed that regulators can effectively coordinate and instruct waste collection operators to send their trucks to specific incineration plants according to the least cost scenario. This may not be feasible in reality. Another example would be that we have not considered the possibility of having waste receiving stations around the island, where waste collection operators can store and compact waste before sending them to an incineration plant. In essence, their operational behaviour will greatly affect whether - and to what extent - the cost savings described in Part 2 can be reaped.
Due to computational limitations, we did not allow our Solver model to run fully. We have used an intermediate result to prepare our final map, which may not be the ideal combination yet. Since other solutions to overcome this problem, such as running the model on a cloud, were not feasible to us at the point of writing, we have made do with this intermediate result. If the model was allowed to run fully, we can say with more confidence that we have obtained an optimal result that has really minimised costs.
Moreover, we have initially set out to perform a Part 3 of our project, to look at other impacts of the incineration plant, such as the impacts of additional pollution generated and proximity to roads and substations. These could have been taken into account through a GIS analysis too. However, due to time constraints we have limited our project to two parts.
Lastly, there may be unknown factors that we currently have failed to identify. If you wish to provide any feedback or pointers to us, please feel free to contact us.
(3) Other factors not considered in model
discussion
This page details the limitations of our analysis that we have identified. It is hoped that these limitations can be improved on in future projects, and also serve as useful pointers for future students taking this course to be mindful of.
We recognise that there is a possibility of a Modifiable Areal Unit Problem (MAUP) arising from this project. We have worked with planning areas, which are aggregated units of smaller areal units called subregions. The waste statistics are thus summed up into these larger areal units, which may mask more detailed distributions within each planning area. Perhaps using more granular units - that is, subregions - would have given us a more accurate representation of reality and hence a more reliable cost distance analysis.
Another possible error related to using a larger areal unit for analysis is that we may have overgeneralised the unit when we converted the polygons into point features. These centroids may not be the ideal points from which we have calculated the least cost paths.
