Page 6 - i1052-5173-29-6
P. 6

may have occurred during Hurricane Irma,   north-central and northwestern reaches of   ASSESSMENT OF
          but an absence of imagery during the   the Cordillera Central. Landsliding in the   CONTRIBUTING FACTORS
          interim precluded us from differentiating   municipalities of Utuado and Naranjito was   Our evaluation of landslide distribution
          between landslides that occurred during   particularly severe, with ~38% and 45% of   shows that while landslides occurred
          each hurricane. We divided the island into    each municipality, respectively, classified   throughout most of Puerto Rico’s moun-
          a 2 km × 2 km grid and classified each grid   as having HLD. All five HLD clusters were   tainous interior, landsliding was particu-
          cell as either having no landslides (NLS),    located ~10–20 km north of the Cordillera   larly severe in five distinct areas (Fig. 1B).
          1–25 landslides/km  (low landslide density,   Central divide (Fig. 1). Areas along the   We examined rainfall, soil moisture, and
                         2
          LLD) or more than 25 landslides/km  (high   southern flank of the Cordillera Central,   geology in NLS, LLD, and HLD areas as a
                                     2
          landslide density, HLD). With the intent of   which have some of the highest relief and   first-order attempt to understand the dif-
          rapidly preparing a product to aid emer-  steepest slopes on the island, generally had   ferences between severely impacted areas
          gency response agencies, we visually   lower landslide densities.     and neighboring areas in which landslides
          examined each grid, but did not map indi-  During field work, we observed a vari-  were less spatially dense.
          vidual landslides. Landslide scars were   ety of landslide failure modes and material   Elevated pore-water pressure from rain-
          readily visible in imagery because of defo-  types and properties. Most landslides were   fall is the most common trigger for land-
          liation from strong winds during Hurricane   shallow, translational failures in soil or   slides (Terzaghi, 1950). Because subsurface-
          Maria, and because of the sharp color con-  saprolite, generally measuring decimeters   water pressure is not commonly measured,
          trast between exposed soil and rock and the   to a few meters deep (e.g., Fig. S1A [see   rainfall is often used as a proxy. During
          remaining vegetation (Fig. S1 [see footnote   footnote 1]). We also observed deeper (up   Hurricane Maria, rainfall amounts
          1]). See Bessette-Kirton et al. (2017) for   to ~30 m) complex failures (e.g., Fig. S1B)   reported by the NHC and NCEP data sets
          further description of the mapping proce-  in soil, saprolite, and rock, and rock falls   for NLS areas were 7%–12% less than the
          dure. We validated and updated our pre-  and rock slides. Many landslides transi-  average island-wide rainfall (Fig. 2). The
          liminary density map (Bessette-Kirton et   tioned into debris flows (e.g., Figs. S1C   average island-wide rainfall reported by
          al., 2017) by helicopter and on the ground,   and S1D), and coalescence and subsequent   the PERSIANN-CCS data set was about
          covering a distance of 1950 km between    channelization of debris flows (e.g., Fig.   equal to rainfall in NLS areas (Fig. 2).
          26 October and 6 November 2017.    S1D) was common. Landslides that partly   All three data sets showed that rainfall in
            From our mapping, we estimated that   reactivated preexisting landslides were   LLD and HLD areas was greater than
          more than 40,000 landslides resulted from   also common. Crosscutting relationships   average (4%–19% increase; Fig. 2), but
          Hurricane Maria. Landslides occurred in at   between debris flow and flood deposits   differences in the spatial distribution of
          least 59 of Puerto Rico’s 78 municipalities   indicated that landsliding occurred before,   rainfall resulted in inconclusive differ-
          (Fig. 1B). Many of the other 19 municipali-  during, and after extreme flooding (e.g.,   ences between LLD and HLD areas.
          ties have such low relief that landsliding is   Fig. S1C). Landslides continued to occur   Variability between rainfall data sets
          unlikely. Five HLD clusters, ranging in    during storms in the days and weeks    calls into question the validity of using
          size from 12 to 132 km , occurred in the   following Maria.           island-wide rainfall estimates for localized
                           2

                     A                                                                                 B
                                                             C























                 Figure 2. (A) Deviation from average island-wide rainfall and soil moisture in areas with no, low, and high landslide densities. (B)
                 Soil Moisture Active Passive (SMAP) root zone soil moisture (Reichle et al., 2018) measurements from April 2015–December 2017
                 (Hurricane Maria shown with stars). (C) Antecedent SMAP root zone soil moisture measurements during the two weeks preceding
                 Hurricane Maria, including the passage of Hurricane Irma. SMAP data were averaged across the entire island and in areas with
                 low landslide density (LLD) and high landslide density (HLD). CCS—PERSIANN-Cloud Classification System; NCEP—National
                 Centers for Environmental Prediction; NHC—National Hurricane Center.

       6  GSA Today  |  June 2019
   1   2   3   4   5   6   7   8   9   10   11