Inventory of landslides triggered by the 2020 Puerto Rico earthquake sequence

Here we present an inventory of remotely and field-observed landslides triggered by 2019-2020 Puerto Rico earthquake sequence. The inventory was mapped using pre- and post-event satellite imagery (PR_landslide_inventory_imagery.csv), an extensive collection of field observations (https://doi.org/10.5066/P96QNFMB) and using pre-earthquake lidar as guidance for mapping polygons with more precise locations and geometries (2015 - 2017 USGS Lidar DEM: Puerto Rico dataset). The inventory consists of a shapefile of 309 polygons (PR_landslide_inventory_pts.shp) outlining the source area and deposits together. It also includes a point inventory (PR_landslide_inventory_pts.shp) marking 170 individual displaced boulders that were outside of areas that could be mapped as polygons and 28 points that indicate rock falls for which we did not have sufficient information from imagery or photos to map as polygons. The individual boulders and rock fall points are differentiated from each other by the “Type” attribute field. Most individual boulders were mapped from imagery, while most rock fall points were mapped based on field observations alone (e.g. notes about occurrences without photos) because they were not readily visible in imagery or captured in photos. Most landslides were triggered initially by the largest earthquake, a M6.4 on 7 Jan 2020, and we include an attribute named “Trigger” to differentiate whether we think each landslide was triggered during the mainshock, an aftershock, a foreshock, or whether the trigger is unknown given the data available to us. The trigger attribute field is uncertain because in some instances, smaller rock falls and rock fall areas that did not completely strip the vegetation from the slope were at times obscured by vegetation immediately post-earthquake and were only visible in later satellite imagery once some vegetation began to die. These rock fall runout areas were mapped on imagery from several months after the earthquake (10 April 2020) because their spatial extent was far easier to see once the vegetation had died back sufficiently. We did not have sufficient information on many landslides to classify landslide types accurately, so this is not included as an attribute, however the vast majority were interpreted as rock falls. Only three landslides that we documented occurred solely in soil. More than a third of the rock falls we mapped occurred on artificially altered slopes like road cuts (noted by the attribute field “Cutslope”). The “Massive” attribute indicates whether the landslide appeared to involve a detachment of most of the source area at once, as observed in field photos where possible and imagery where photos were not available. Massive is set to false if the slope failure involves many individual rock detachments while other parts of the source area remained intact. Some large polygons actually represent runout areas over which many individual rocks rolled without involving failure of the entire face of the source area (Massive=False) and should not be interpreted as a single large landslide. While lateral spreading was triggered by this earthquake sequence, we do not include it in this inventory and instead group it with the liquefaction inventory for this event (https://doi.org/10.5066/P9JEN3H2).