Earlier today I started a discussion about landslide susceptibility and risk in eastern Canada here: landslide susceptibility and risk in eastern Canada. This post will extend that discussion and (I think) almost wrap the topic up. The final piece will be to subdivide risk by geographic boundaries (i.e. counties) as an aid to prioritization of study efforts.
In the earlier post we saw a map showing landslide susceptibility across the study area. We can re-frame landslide susceptibility from the qualitative, descriptive categories (e.g. “low,” “high” etc) into the following numeric categories:
1 – Low – landslide spatial frequency = 0.22 times the study area average value
2 – Low to moderate = 4.3 times study area average
3 – Moderate to high = 8.3 times
4 – High = 40 times
(these values come from statistical analyses embedded in a client report and will be detailed in a paper, but not elaborated further here)
Now I haven’t elaborated on risk concepts at great length, and in formal risk work there is a particular jargon with some very specific meanings. However, “risk” is a measure of some expected loss or harm (which can be an annual cost of damage or loss of life, for example), and can be generally written as being a product (or some combination) of “hazard” and “consequence,” where “hazard” is a measure of the probability of a potentially harmful event (say a large landslide) and “consequence” is the possible outcome of that hazard occurring, which would consider both the possibility of potentially affected receptors being present, and their vulnerability to the hazard given their presence.
I propose that on the regional scale, the probable presence of people or valuable infrastructure over time can be modelled approximately by population density. The following map shows inferred population density, as estimated from calculated road density according to Population density = 28 * (road density)^2; where population density is in people per square km, and road density is in km per square km, and this relationship comes from the “best fit” interpretation reported in prior posts noted in today’s earlier post. The map shows log10(pop density) thus showing inferred population on a logarithmic scale.
From these two maps, we have five categories of inferred population density (< 10 persons per sq km, < 100, < 1000, < 10000 and > 10000) and four categories of landslide susceptibility, or relative landslide frequency (0.22, 4.3, 8.3, 40). If we combine these two maps, we can obtain 20 different categories as follows:
Here we have multiplied the expected spatial frequency of landslides within the different susceptibility category zones with population density to yield a relative (numeric) scale for anticipated likelihood of spatial intersection between future large landslide occurrence and presence of people or infrastructure. This numeric scale has been broken into three broad categories of “relative risk” based on the value being < 1000, < 10,000, and > 10,000 as shown.
The precise qualitative meaning of these risk categories is not especially important; the knowledge that the product of landslide frequency by population density increases by an order of magnitude indicates that the probability of loss increases approximately by an order of magnitude from category 1 to 2, and again by another order of magnitude from category 2 to 3. If we choose to ignore the lowest risk category in order to focus our attention to the highest risk areas, we can generate a qualitative risk map for large landslides in sensitive clay affecting eastern Canada as follows:
We will have a closer look at landslide risk in specific subsections of the study area in one further post on the topic, when the analysis is complete.