Page 7 - i1052-5173-29-6
P. 7

areas (Fig. S2 [see footnote 1]). Collecting   Additionally, the antecedent soil moisture   Lorenzo Formation granodiorite (Ksl) had
          accurate rainfall data during hurricanes is   in LLD and HLD areas was 13% and 11%   no HLD areas and lacked landslides in
          intrinsically difficult due to high winds and   above average, respectively. SMAP data   53% of its area, yet the formation is essen-
          the inability to measure sideways rainfall.   showed that prior to the passage of   tially equivalent to the Ku granodiorite
          During Hurricane Maria, these complica-  Hurricane Irma, the average island-wide   (Bawiec, 1998), which displayed the
          tions were compounded by damage to rain   soil moisture was already higher than    greatest areal percentage of HLD (63%).
          gages (at least 14 of the U.S. Geological   normal, and rainfall from Irma only   To the extent of the resolution of our
          Survey’s 24 gaging stations; U.S. Geo-  caused a slight increase in island-wide soil   data, these variations in landslide density
          logical Survey, 2017b) and the failure of   moisture (Fig. 2C). The increase in soil   cannot be attributed to differences in rain-
          two Federal Aviation Administration   moisture caused by heavy rainfall during   fall or slope. Instead, it appears that rela-
          Doppler radars (Buchanan, 2017; National   Hurricane Maria was unprecedented   tions between geologic formation and
          Weather Service, 2017b). Additionally,   (41%–48% above average) during the time   landslide density were largely due to soil
          localized effects, such as orographically   period for which SMAP data are available.   moisture variability. For example, Ksl
          enhanced rainfall, may have produced   However, increased antecedent soil mois-  occurs in the eastern quarter of the island
          pockets of heavy rain that were not accu-  ture prior to Hurricane Maria evidently   where soil moisture was relatively low
          rately represented by any of the existing   influenced the susceptibility to land-   (0.25–0.26 m /m ), whereas Ku occurs
                                                                                             3
                                                                                          3
          data sets.                         sliding in LLD and HLD areas. This    where soil moisture was relatively high
            Soil moisture is a better proxy for pore-  finding, along with the physical relation   (Fig. S5 [see footnote 1]). Individual for-
          water pressure than rainfall because it is a   between increased pore-water pressure   mations with large HLD areas (Fig. 3) also
          subsurface hydrologic response variable.   and decreased effective material strength   reveal trends between landsliding and soil
          We found that soil moisture measurements   (Terzaghi, 1950), may make soil moisture   moisture (Fig. S5) when considering for-
          from NASA’s Soil Moisture Active   a useful tool for estimating differential   mations that cover large areas of Puerto
          Passive (SMAP) mission (see [A] in the   susceptibility to landsliding prior to future   Rico (necessitated by the gross resolution
          GSA Data Repository [see footnote 1])   storms with predicted heavy and wide-  of SMAP data). The three formations that
          differentiated between LLD and HLD   spread rainfall. Although the resolution of   are most widespread from east to west
          areas more effectively than did rainfall   soil moisture data is coarse (9 km × 9 km),   (basaltic breccia and basalt lava [Kln],
          data. We compared landslide density to   and the data for Puerto Rico are primarily   hornblende quartz diorite porphyry [Thp],
          SMAP’s 9 km × 9 km gridded global esti-  derived from a land-surface model with-  and diorite [TKdi]; spanning east-west
          mates of root zone (0–100 cm) volumetric   out local calibration, our preliminary   distances of ~60, 70, and 160 km, respec-
          water content before (9:30 a.m. AST on 18   assessments indicate that SMAP data   tively) display strong positive correlations
          September 2017) and after (9:30 p.m. AST   could be a useful component of landslide   between soil moisture and landslide den-
          on 21 September 2017) Hurricane Maria   forecasting across widespread areas prior   sity (Fig. S5). We conclude that landslide
          (Reichle et al., 2018; Fig S3 [see footnote   to future hurricanes.   densities were not strongly controlled by
          1]). These estimates rely on remotely   An examination of mapped rock forma-  variable susceptibility to landsliding
          sensed measurements of brightness tem-  tions in mountainous areas indicated that   among the geologic formations, but by
          peratures and the solution of a water and   geologic material did not consistently cor-  variable soil moisture.
          energy balance (see [A] in the GSA Data   relate with landslide density. Twelve of the   Although we have identified spatially
          Repository for details [see footnote 1]).   island’s 145 rock formations (Bawiec,   averaged correlations between landslide
          Both antecedent and post-event root zone   1998) had >10% HLD by area. Landslides   density and soil moisture, the resolution
          SMAP measurements showed that soil   were observed nearly everywhere in half   of our island-wide data sets and our
          moisture was 10%–23% greater than the   of these 12 formations and had HLD clas-  method of grid analysis do not allow
          island-wide average in areas where land-  sification in ≥11% of their areas. In con-  for differentiation of contributing factors
          slides occurred (LLD and HLD areas) and   trast, the other six formations had HLD    at a scale sufficient for detailed analyses.
          5%–7% less than the island-wide average   in ≥14% of their areas but also had no   For example, our analysis of island-wide
          in NLS areas (Fig. 2A). Additionally, the   observed landslides in ≥11% of their areas.   geology (1:100,000 scale) does not
          difference in deviation from the mean   Similar variability is also apparent when   account for localized variations in geo-
          between LLD and HLD areas for anteced-  grouping geologic formations by terrane   logic units that may affect rock weather-
          ent and post-event root zone soil moisture   (Fig. S4 [see footnote 1]). For example, the   ing rates and soil formation, and therefore
          measurements (11%) was nearly twice as   intrusive igneous rock terrane had the two   landslide susceptibility. The resolution of
          large as the spread between LLD and   formations with the highest proportion of   our 2 km × 2 km grid framework is com-
          HLD areas for rainfall (3%–6% increase   HLD area, both of which also had land-  patible with island-wide rainfall and soil
          and 7% decrease).                  slides across nearly their entire areas   moisture data sets but is likely less useful
            SMAP measurements from the two   (granodiorite-quartz diorite of the Utuado   for comparison with higher resolution
          years prior to Hurricane Maria (Fig. 2B)   batholith [Ku] and hornblende quartz dio-  data sets (<4 km ). While our rapidly pro-
                                                                                             2
          show that soil moisture across Puerto Rico   rite porphyry [Thp]; Fig. 3). However, two   duced landslide density map allowed for
          immediately before Maria (18 September   similar formations both lacked landslides   the analysis of contributing factors on a
          2017) was 10% above the average island-  and had large areas of high density land-  regional scale, a detailed inventory would
          wide soil moisture during the period of   slides (diorite [TKdi] and rhyodacite por-  be invaluable for more localized assess-
          record (April 2015–December 2017).   phyry [Trhp]; Fig. 3). Finally, the San   ments of landslide susceptibility.

                                                                                       www.geosociety.org/gsatoday  7
   2   3   4   5   6   7   8   9   10   11   12