PROXIMITY ANALYSIS OF RAINFALL INDUCED LANDSLIDE SITES TO ROAD SURFACES AND FLUVIAL CHANNELS DURING HURRICANE MARIA IN PUERTO RICO

Puerto Rico (PR) is prone to landslides due to abundant steep slopes and moist soil conditions associated with tropical rainfall. Hurricane Maria impacted PR as a category 4 storm on September 20, 2017. Heavy rains caused landslides in the mountainous regions of the island, leading the US Geological Survey (USGS) to develop a preliminary map of mass wasting density. At the Univ. of PR Dept. of Geology, aerial imagery from FEMA, NOAA, and Digital Globe were used to catalog individual landslides in order to create a complete whole-island inventory which yielded over 40,000 landslides.

Past studies have highlighted the importance of roads and channels in relation to landslide occurrence. In this study, the Frequency Ratio statistical method was used in combination with geological and geographic data to analyze and quantify the importance of site distances from roads and rivers/streams. The Euclidean method was used to determine the degree to which river networks and roads influence landslides.

Results show that proximity to river has little impact on landsliding. River flowlines were those defined by in the USGS National Hydrography Dataset. Distances ranging from 0-25m and above 400m have low correlations with landslide sites. Sites that are over 400m from a fluvial channel are mostly within the island’s karst region where there were notably fewer landslides caused by Hurricane Maria. A positive correlation exists at distances ranging from 25-400m. It is important to note that many landslides may have occurred along or near small sub-channels that are not designated as fluvial channels in the dataset. Many of the highest slope and highest susceptible areas likely lie away from the principal channels used in the analysis.

As previously studied in eastern PR, our results confirm that roadways across the island are highly important with respect to landslide sites. The distance from roads in the US Census Bureau dataset shows a consistent correlation with landslides. There is an extremely low correlation of landslide sites that are more than 400m from a roadway. These results reinforce the interpretation that human activities, such as road construction, can be extremely influential in increasing the vulnerability of natural sites. These factors should be considered carefully when developing landslide susceptibility models.