Following this morning’s post, here, landslide susceptibility and risk in eastern Canada part 3, I’ve done some further work on spatial orgaization of the risk map, looking for a meaningful and efficient way to draw priorities from it.
I previously discussed some of the challenges in interpreting priorities on the basis of risk ranked by county, either in terms of total risk (per county) or risk density. Since the counties are all different sizes, their area influences the analysis, with both representations (aggregate risk and risk density) giving unintuitive (and hard to interpret/understand) results.
I’ve therefore divided the study area into equal area grids, both 10 km by 10 km and 25 km by 25 km, and obtained the total risk within each box. Since each box has the same area, the values can be compared directly. Note that total risk is calculated by counting all pixels of each risk class, and weighing medium and high risk pixels by a factor of 10 and 100, respectively, to reflect that they each represent an order of magnitude higher risk than the lower category.
If we look at this on a 25 km grid with a stretched scale for relative risk, we obtain:
We can re-plot the grid squares to show the top 5, 10 and 20 % of risk as follows:
The same two plots for a 10 km grid follow:
It should be noted that some of the higher risk areas likely need relatively little investigation to rule out significant concern. For example, most of the Montreal area is shown with elevated risk, but this is a result of the susceptibility model overpredicting susceptibility in selected parts of the site. Concern in most of that area can probably be eliminated through focussed air photo study.
From these maps, one can generate priorities for further study. I will admit that I find the large grids a little jarring, and might prefer to work directly with the risk map in un-modified state, if it were my task to move ahead with further study.