Rwanda has a high density of landslides, heavy precipitation events and a shortage of
resources to study them, making it an excellent candidate for study using satellite-based remote
sensing data. To assess landslide hazards countrywide, I first built a landslide inventory of 254
landslides and used a statistical methodology. Using logistic regression on 24 test variables, I
determined that slope and population density are statistically most relevant to landslide
occurrence in Rwanda. A preliminary predictive hazard map for Rwanda was produced, with an
overall predictive accuracy of 79.6%. Second, I worked to define a relationship between
precipitation intensity and landslide density for a landslide-prone study area in western Rwanda.
In the 1180 km2
study area, I mapped 577 landslides, using CNES/Astrium and WorldView
satellite imagery in Google Earth over a study period of 2000 to 2015. One 400 km2
part of the
study area has a high landslide density of 1.4 landslides/km2
, while another 780 km2
part with
identical geology, soils, land-use, and vegetation has a much lower landslide density. To identify
possible triggering events for these landslides, I analyzed a 16 year record of TRMM (Tropical
Rainfall Measuring Mission) satellite precipitation data. The high landslide density region and
the low landslide density region were not notably different in rainfall, as quantified by recurrence
interval analysis. A relationship between precipitation and landslide density could therefore not
be developed, and the null hypothesis cannot be ruled out. This apparent lack of connection
could result from a variety of factors including TRMM grid size, satellite imagery temporal
resolution, antecedent soil moisture, or vegetation regrowth rates.