LANDSLIDE SUSCEPTIBILITY ANALYSIS

Problem

Landslides can cause loss of life and property. In this exercise, I used the statistical learning capabilities of
R to develop a model that predicts where landslides are likely to occur.

The study area is a mountainous region of southern Ecuador. Geologists familiar with the area have determined
that the incidence of landslides is related to:

slope angle,
curvature with respect to the horizontal plane,
curvature with respect to the vertical plane,
the extent of area whose runoff flows towards a central location (catchment area), and
the degree of vegetative land cover.

Data concerning land cover was unavailable, therefore, I used elevation as a proxy for vegetation type
and precipitation rates.

Geologists logged the locations of 175 landslide points. My model compared these locations with 175 randomly
chosen spots where no landslides were observed.

Data Source

I was provided with a table with the locations of 175 landslide points and 175 non-landslide points. A 10 meter resolution digital elevation model (DEM) of the study area was provided as well.

Analysis

My job was to build a model which stakeholders could use to assess the risk of landslides throughout the study area.

I had a list of spots where landslides are known to happen, and a group of randomly selected sites where landslides have not happened. Within R, I used elevation data to calculate slope, concavity, and catchment extent for the study area. Then, I compared the terrain characteristics of landslide-prone areas to those of stable areas. R developed a linear regression model and assigned a coefficient to each terrain characteristic. Using the linear regression model, I could then assign a landslide susceptibility score to every spot in the study area.

Results

Each 100 square meter tile was given a rating between 0 and 1 representing the degree to which the spot's terrain bears similarity to places where landslides have happened. As expected, the areas most susceptible to landslides lie in valleys and steeply sloping ravines where the topography concentrates water accumulation.

Tools Used

I calculated the slope and terrain curvature data in R. I used QGIS to create the hillshade image and assemble the maps. The 3D visualization was prepared in QGIS as well.

This demonstration was based on an exercise from the book Geocomputation with R by Robin Lovelace, Jakub Nowosad, and Jannes Muenchow.