Discussion

While we are happy with the outcome of our model and believe it to be a relatively accurate representation of transportation risks, there are possible issues. Areas of concern in our project can be divided in two categories, sources of error and limitations of the project itself.

Sources of error include uncertainties in the datasets used, unintentional introduced uncertainty by Arcmap operations, and complications from raster resolution. While our datasets are from reputable sources and are up to date, there is always the possibility of error and that have been changes of significance since the datas was released. Also, the processing of our data may have had unintended effects. Use of the focal statistics tool and normalization may have diluted population densities in the more rural areas of Canada, ignoring smaller towns in the waste's path. Focal statistics computes the average of the surrounding 20km, reducing the presence of rural towns in our model, and limiting their effects on the final cost surface. Computing power limited the resolutions of the raster we were able to use in the model. We would have like to have used a finer resolution, but space limitations and length of processing times made this unfeasible. The cell size of 100m, given that track is rarely wider than 3 meters, likely led to some over-generalization of features. The conversion from vector to raster, in some places, allows the cost path to bypass risk features (see map 10 ). In order to properly add the layers together, we had to buffer the track types to ensure overlap. This many have led to over-representation of smaller features.

We attempted to examine the effects of raster cell size on the model by creating an additional equally weighted cost surface with a higher resolution of 25m (See map 11). The results are similar as those of the 100m cell, though not identical. Spanish is still the most desirable destination for spent fuel (see map 12) and Northwest Ontario is still the most desirable region. This shows that the model is robust and produces similar results when the cell size of the cost surface is decreased. Using a small cell size for the analysis would however likely have yielded a more accurate cost surface, in that the higher resolution would have detected smaller and more closely packed roads and bridges, possibly ignored in the lower resolution. Although 25m cells would be a better choice, operations in this resolution are very time consuming, and the total rendering time for the 25m layer was 33 hours. With more computing power at our disposal, we would have been better able to use the high resolution of our data to our advantage. With more time we could have calculated a weighted cost surface at 25m to compare with 100m, but this proved to be unfeasible given timing and available hardware.

Sources of Error and Limitations

The scope of our project is limited in that we only made use of only existing rail lines, employed fairly subjective variables, and assigned arbitrary weights. We chose to look exclusively at railroads to narrow the amount of data we would have to process, and because we figured it to be a safer means of transporting waste to the sites. It is important to note, however, that unmarked trucks are also commonly employed for moving nuclear waste. Our model does not consider the possibility that it would be safer to transport the waste by truck. Many of the sites, such as Pinehouse and English River First nation, are not in close proximity to the railroad, and therefore carry the additional risk of transporting the waste from the existing track to the town itself (see map 1). The risk variables we chose are based on review of nuclear safety reports, however there may be other regional factors that were not included. It is also important to note that a number of social and cultural factors are not addressed in this model. For track types, weights are in the right order, but their magnitude is fairly arbitrary (see table 2).

It is also important to note that the federal government, under no circumstances, will release information on the routes of waste transport. Routing is kept highly classified, and the unmarked containers are routed such that they cannot be predictably intercepted. We therefore cannot check the accuracy of our results with official routes.

Our results show that the proposed sites in Northwest Ontario, notably Spanish and Hornerpayne, pose the least of risk to the transport of nuclear waste from Pickering based on the infrastructure and population variables we include in our analysis.

This analysis considers 3 factors risk factors: distance, infrastructure and population. The risk associated with distance in our model is low, and is only really a factor when sites are extremely distant from Pickering. Longer distances generally result in higher risk not because of absolute distance, but of increased exposure to infrastructure weaknesses along the way.

The relationship between infrastructure and population density is the major factor that influenced risk in this project. This is most clearly illustrated by the the high risk associated with Walkerton and South Bruce, despite their proximity to Pickering. This is due to route constraint which force them to travel through Toronto (see maps 7 & 9). While population density and infrastructure features weights are normalized, the restricted nature of population densities in Canada generally means it is better to avoid population centres as surrounding areas have much smaller populations and therefore much lower risk. This is shown both in the weighted and non-weighted results. Avoiding populated areas is also somewhat useful for reducing risk cost with infrastructure but not to the same extent. While infrastructure is more concentrated in major centres it is relatively evenly dispersed along the rail lines in rural regions. It is not, however, always possible to avoid major centres. All the cost paths for the proposed sites in Saskatchewan were channeled through Winnipeg, as this is the only route through this region (see map 9).

Map 9: Population and Infrastructure

Map 10: Raster Path Complications

Conclusion

The weighted model is the best representation of safest cost paths because it best protects population from the harmful effects of a spill or incident. The NWMO would not route trains through downtown Toronto just because the bridges are new and there are few tunnels, and would instead seek to limit contact between waste an populations as much as possible. We therefore conclude that the best site for the repository, when considering waste transportation, is Hornerpayne Ontario (see map 8). The next step would be to wait and see if we are correct as the project continues to progress.

Map 11: Cost Paths 25m cells, equal weights

Map 12: Least Cost Path 25m cells, equal weights

Source: DMTI Spatial, StatsCan

Source: DMTI Spatial

Source: DMTI Spatial, StatsCan