Accelerated marsh erosion following the Deepwater Horizon oil spill confirmed, ameliorated by planting

Abstract

Multiple studies have examined the effects of the Deepwater Horizon oil spill on coastal marsh shoreline erosion. Most studies have concluded that the spill increased shoreline erosion (linear retreat) in oiled marshes by ~ 100-200% for at least 2-3 years. However, two studies have called much of this prior research into question, due to potential study design flaws and confounding factors, primarily tropical cyclone influences and differential wave exposure between oiled (impact) and unoiled (reference) sites. Here we confirm that marsh erosion in our field experiment was substantially increased (112-233%) for 2 years in heavily oiled marsh after the spill, likely due to vegetation impacts and reduced soil shear strength attributed to the spill, rather than the influences of hurricanes or wave exposure variation. We discuss how our findings reinforce prior studies, including a wider-scale remote sensing analysis with similar study approach. We also show differences in the degree of erosion among oil spill cleanup treatments. Most importantly, we show that marsh restoration planting can drastically reduce oiled marsh erosion, and that the positive influences of planting can extend beyond the immediate impact of the spill.

Figure 1

Field experiment map with study sites by marsh oiling/treatment class, including reference. The study area is salt marsh dominated by Spartina alterniflora, located on Barataria Bay, part of the Mississippi River Delta, Southeast Louisiana, USA. Inset map includes tracks for Hurricane Katrina (2005, east of study site) and Hurricane Isaac (2012, south and west of study site). Figure generated by the authors for this research using ArcGIS v10.8.1 (https://www.esri.com/en-us/arcgis/products/arcgis-desktop/overview).

Figure 2

Field measured marsh shoreline erosion rates 2010–2016 (m yr⁻¹). Data are means with 90% confidence intervals, n = 5 for Reference, 9 for Oiled-Untreated, 5 for Oiled-Manual, and 2–6 for Oiled-Mechanical treatments (n = 14–20 for Oiled sites combined) depending on year. Due to missing values, the desire to use as much data as possible, and the lack of clear differences among cleanup treatments, we pooled the oiled site data for statistical analysis. Marsh erosion rates differed among Reference and Oiled sites (F_1,17 = 9.751, p = 0.006); among years (F_2.32,39.40 = 2.703, p = 0.072); and for the interaction of oiling and year (F_2.32,39.40 = 2.648, p = 0.076). Pairwise differences (Tukey’s test) among Reference and Oiled sites were observed for 2010–2011 (p = 0.003) and 2011–2012 (p = 0.001), but not for other years. See Supplementary Table S1 for detailed two-way mixed ANOVA results.

Figure 3

Field measured marsh soil shear strength 2011–2012 (KPa). Data are means with 90% confidence intervals, n = 4 for Reference and 13 for Oiled sites. Soil shear strength differences were observed among the Reference and Oiled sites (t_6.29 = − 3.877, p = 0.007, Welch’s t-test). See Supplementary Table S2 for further details.

Figure 4

Remote sensing measured marsh shoreline erosion rates 1956–2018 (m yr⁻¹). Data are means with 90% confidence intervals, n = 5 for Reference, 9 for Oiled-Untreated, 5 for Oiled-Manual, 9 for Oiled-Mechanical, and 5 for Oiled-Mechanical-Planted treatments. Marsh erosion rates differed among oiling/treatment categories (F_4,28 = 9.368, p = 0.000); among time-periods (F_2.48,69.47 = 62.210, p = 0.000); and for the interaction of oiling/treatment and time-period (F_9.92,69.47 = 4.593, p = 0.000). In the post-Katrina/pre-spill period (2005–2010) erosion differences were not observed among any oiling/treatment classes, including Reference (p = 0.868 to 1.000). In the spill impact period (2010–2012) erosion differences were observed between: Reference versus all oiled classes (p = 0.000 to 0.001) except Planted (p = 0.410); between the Manual and Mechanical treatments (p = 0.063); and between Planted versus all other oiled classes (p = 0.000 to 0.065). In the post-spill period (2012–2018) erosion differences were observed between: Planted versus Untreated and Manual treatments (p = 0.066 and 0.072). Between the post-Katrina/pre-spill and spill impact time-periods, erosion differences were not observed within the Reference (p = 0.989) or Planted (p = 0.340) classes; however, differences were observed between these time-periods within each of the other oiled classes (p = 0.000 in all cases). See Supplementary Table S3 for detailed two-way mixed ANOVA results. Tukey’s test was used for all pairwise comparisons after ANOVA.

Figure 5

Modeled mean wave power in oiled marsh versus reference sites 1956–2018 (W m⁻¹). Data are means with 90% confidence intervals, n = 5 for Reference, 9 for Oiled-Untreated, 5 for Oiled-Manual, 9 for Oiled-Mechanical, and 5 for Oiled-Mechanical-Planted treatments. Mean wave power differed among oiling/treatment categories (F_4,28 = 52.166 p = 0.000); among time-periods (F_1.13,31.71 = 396.386, p = 0.000); and for the interaction of oiling/treatment and time-period (F_4.53,31.71 = 15.503, p = 0.000). Within each time-period mean wave power differences were observed between Reference versus each oiled class (p = 0.000) but were not observed among any of the oiled categories (p = 0.153–1.000). Between the post-Katrina/pre-spill (2005–2010) and spill impact (2010–2012) time-periods, erosion differences were observed within each of the oiled/treatment classes including reference (p = 0.000 to 0.004). See Supplementary Table S5 for detailed two-way mixed ANOVA results. Tukey’s test was used for all pairwise comparisons after ANOVA.