Preparing for the "Big Daddy" of Earthquakes:
Vancouver's 9.0 Megathrust
Methodology
To begin, we found statistical information for the dissemination areas (DAs) in Vancouver from the 2006 census. With this data, we normalized the population over the area of the DAs and represented this new class on our map to get a sense of the relative densities of Vancouver, and to help us guide our focus on particularly dense areas. Doing so converted this extensive variable into an intensive variable, and thus made our map more reflective of the actual distribution of people in Vancouver as opposed to just using nominal data. The classification scheme used throughout was Jenks Natural Breaks, as this displayed the data most effectively, while using an appropriate number of classes (4). This data is visible in our final product in which population density was used to determine where lots of people were located and thus where more emergency relief centres should be located.
Creating a TIN and DEM
We then obtained elevation points from the Open Data Catalogue and, with that, created a TIN
and then a DEM. This was necessary because both slope and elevation are important factors when creating a risk analysis of an earthquake and thus determining the safest areas in Vancouver. Elevation data is necessary because areas that are less than 8 metres above sea level are at risk during a large earthquake due to potential tsunami or rise in sea level. Slope is also important because areas that have a higher slope are dangerous due to their lack of stability. Once we created the DEM from the TIN, we had the elevation data and could create the slope data. We used the spatial analyst tool “slope”.
We normalised the slope so that a lower slope is considered better. The closer the slope is to zero, the safer it is in terms of stability. In reality, a surface is considered unsafe once it exceeds 30 degrees. This limited our results as there were areas within the City of Vancouver with slopes up to 69.68 degrees, thus being very unstable in the event of an earthquake.
Surficial Geology
We then obtained surficial geology data for British Columbia from BC Mines & Energy, and created a new field in this and assigned them ranks based on their stability in the event of an earthquake. These rankings would be used during the next step of the multi-criteria evaluation. Once clipping this to the City of Vancouver, we realised that the surficial geology was generally quite uniform, however we carried on anyway as geology is, in fact, quite important.
Soil liquefaction is more intense in water-saturated, granular materials such as sand, which have high porosity and thus are prone to lose their strength and act more like a liquid under high stress events such as an earthquake. Soil liquefaction may lead to sinking buildings, bridges, uneven ground surfaces. For this reason, we assigned values to the sand-like types of surficial geology (sandstone, silt, sand, sand & silt) as being highly unstable, and thus a value of 0.
Rock has about 13x the cohesive strength as silt, and 67x the cohesive strength as clay. Within rock, granitic rock has a very high tensile strength, closely followed by volcanic and till. The tensile strength and relative stability of till varies widely depending on the kind of till and its origin, however these are details that aren’t available in the surficial geology dataset, so we assumed the till was mainly comprised of glacial till, thus making it quite stable, although not as stable as granitic or volcanic rock (which is why we ranked it as being 3rd most stable in our Multi-Criteria Analysis).
Performing a MCE (Multi-Criteria Evaluation)
We then performed a Multi-Criteria Analysis, following similar steps to the ones taken in our second lab. We identified the main critical factors for our projection: surficial geology, elevation, and slope. We normalized them so that a ranking of 1 for geology (representative of volcanic rock) was the best, and a ranking of 2 (till) was the worst. We used the fuzzy membership tool to accomplish this, similar to lab 2. A ranking of 1 for slope represented a low slope, which is good, and a ranking of 0 represented a generally unsafe, higher slope. This was also done similarly to lab 2, in which the raster calculator was used. In our MCE, we excluded elevations that were less than 8 metres above sea level, as this is considered unsafe in the event of tsunami post-earthquake according to Emergency Management BC, and used the fuzzy membership to normalize this data as well. Our weighting for the MCE was decided to be 40% for both the geology and slope, as those are both much more important than the elevation, which we weighted at 20%.
A sensitivity analysis was conducted where all three input criterion (elevation, slope and surficial geology) were equally weighted and run through the Weighted Sum tool. This was done to compare the results using a different set of weights and to test the model's response to changes in its parameters. When using equal weights for all criteria, the resulting map displayed a larger proportion of "best" suited land in the event of a megathrust earthquake, however it was not significant enough to create significant uncertainty that would nullify the use of the weighted MCE. Ultimately, we decided to use weights more reflective of the relative importance of the criteria upon choosing our MCE final map. Elevation above 8m above sea level is important, but not as important as surficial geology or slope. Thus, after clipping off all areas of Vancouver with elevations below 7m above sea level, we decided our map was more informative when using weights of 40% each for slope and surficial geology, and 20% for elevation.
Object (Building) Data
After this, we started to manipulate the object data which included buildings that we would suggest act as relief centres (community centres, schools, libraries). We did some calculations on the data in terms of assigning each type of building a capacity, and summing those capacities depending on how many buildings were in each DA. This required performing multiple joins based on spatial location, all results becoming linked to the DA attribute table so we could compare the relative densities of people and how much relief space could potentially be provided to them.
We then used a shapefile of the roads in Vancouver, and created a 50 meter buffer around the main streets. We did this because we assumed that there would be chaos following an earthquake and we wanted to avoid relief centres being directly on those streets so they are kept as cleared as possible for emergency services.
Finally, we used the MCE, the building age, and the population density to narrow down the community centres, public libraries, and schools. We only chose the buildings that were in the best suitability decided by the multi-criteria analysis. From that, we chose to highlight the buildings that were in dense areas and in DAs that had an average older building age because these would be the most in need of emergency relief shelters. And that is how we created the final project.