Map depicting susceptibility to landslides triggered by intense rainfall, Puerto Rico 
By K. Stephen Hughes and William H. Schulz
2020
Open-File Report 2020-1022

Landslides in Puerto Rico present significant hazards to human safety and the built environment, as illustrated by the more than 70,000 landslides triggered by Hurricane Marķa during September 2017 and numerous previous widespread landslide events, which commonly are triggered by intense rainfall. Maps depicting relative susceptibility to future rainfall-induced landslides may therefore be useful for planning, zoning, and emergency-response preparation. The accompanying map and geospatial files portray the relative susceptibility to rainfall-induced landsliding across the main island of Puerto Rico. These data were developed following a bivariate statistical approach considering a complete-island landslide inventory created from imagery obtained after Hurricane Marķa, slope inclination, land-surface curvature, soil type, geologic terrane, mean annual precipitation, land use, soil moisture, and distance to roadways and streams.  Details describing data development are provided in the accompanying pamphlet. 

This Open-File Report is provided separately in English and Puerto Rican Spanish; contents of the pamphlets and maps (one for each language) are identical. The maps are provided as printable .pdf files that may be opened using the freely available software Adobe Reader (https://get.adobe.com/reader/). The map also is provided as georeferenced .pdf, .jpg, and .kml files that may be opened using a variety of programs, including nearly any geographic information system program. Nearly all computers and mobile devices are capable of opening .jpg files using pre-installed software, although georeferencing of the files is not always retained. Google Earth (https://www.google.com/earth/versions/) is a freely available geographic information system program that can open .kml files through an internet browser or by using computer or mobile device versions of the program. Because of the level of detail provided by the map and its associated large file size, georeferenced versions of the map are best suited for use on computers running a geographic information system, including Google Earth.

Use and limitations of the landslide-susceptibility map
The map and related geospatial files depict estimated relative susceptibility to landslide occurrence during or soon after intense rainfall, such as produced during tropical storms including hurricanes. They are intended to illustrate relative potential landslide hazard to help guide site-specific landslide susceptibility assessments, plan for future development, and plan for future widespread landslide events. For example, locations with higher landslide susceptibility could (1) be excluded from future development; (2) require slope-stability studies prior to development and grading, foundation, and drainage design and construction that considers potential slope instability; (3) be targeted for evaluation and mitigation of potentially adverse conditions; (4) be monitored for signs of incipient landsliding, such as cracks in roadways and structures, leaning utility poles and trees, and broken water or sewer lines; and (5) be considered for evacuation prior to likely landslide-inducing rainfall and for emergency response activities after such rainfall. The map and related files are not a substitute for site-specific slope-stability investigation by licensed geologists and engineers. Additionally, many landslides in Puerto Rico move far from the locations where they originate and can destroy structures and injure and kill people as they move downslope; the map and related files do not depict this significant hazard. Finally, locations of landslides triggered by earthquakes or prolonged, relatively low-intensity rainfall may significantly differ from high/very high/extremely high susceptibility locations shown on the map and related files. 

For best use during emergency response to specific rainfall events, the map and related files should be evaluated alongside event-specific rainfall conditions, with landslides being more likely where larger amounts of rainfall occur in locations with higher landslide susceptibility. As described in the accompanying pamphlet, root-zone soil moisture estimated by the NASA SMAP mission correlated well with landslide locations during Hurricane Marķa, and professionals may utilize event-specific soil moisture for more quantitative assessment of potential locations of landslides that result from future rainfall. Such use requires expertise in GIS data manipulation. Interested professionals may obtain root-zone soil moisture data from the NASA SMAP mission (https://smap.jpl.nasa.gov/, last accessed 26 December 2019), ensure that the data projection matches that of the SI_raster_for_SMAP file available with this report, reclassify the NASA SMAP data to the appropriate SI values provided in Table 3 of the accompanying pamphlet, sum the SI_raster_for_SMAP and reclassified NASA SMAP data, and reclassify the sum by area percentile to obtain susceptibility rankings as provided in Table 5 of the accompanying pamphlet. 

Most landslides in Puerto Rico triggered by intense rainfall are shallow (to several meters deep) and occur in unconsolidated material (soil and saprolite) overlying rock (e.g., Jibson, 1989; Larsen and Torres Sanchez, 1992; Bessette-Kirton et al., 2019). These are the types of landslides most well represented by the landslide susceptibility map because it was produced using an inventory of landslides triggered by Hurricane Marķa (Hughes et al., 2019). However, intense rainfall may also trigger very large, deep landslides that occur in rock, such as the tragic landslide that occurred at Mameyes in Ponce municipality during 1985 and which killed more than 129 people (Silva-Tulla, 1986). Landslides such as this are poorly represented by the susceptibility map because very few were triggered by Hurricane Marķa so did not provide statistically robust measures with which similar landslides could be forecast. Additionally, many landslides worldwide are triggered by human activities, such as hillslope grading and redirection of drainage. Local activities such as these cannot be accounted for on the map. 

Common to nearly all modeling efforts, accuracy of the landslide susceptibility map relies upon accuracy and completeness of input data utilized for its development. Among others, it is expected that the following omissions and conditions detract from accuracy of the map: (1) variability within given geological terranes of bedrock strength and hydrologic properties, discontinuity orientation and condition; (2) variability within given soil types of soil strength and hydrologic properties; (3) consideration of only part of the road network, including omission of most undocumented farm roads, paths, and trails; and (4) use of a 2015-2016 topographic model that pre-dates landslides triggered by Hurricane Marķa and other topographic changes.