Introduction

A diverse and complex array of environmental and ecological factors potentially governs ecosystem resiliency (defined as the time required to rebound from a disturbance) following large-scale environmental perturbations such as fires (Thomas and Waring 2015), tsunamis (Whanpetch et al. 2010), eutrophication (Fertig et al. 2014), and climate shifts (Kovach et al. 2016). For oil spills, the chemical composition and intensity of oiling are key influences on initial impacts as well as long-term recovery (Michel and Rutherford 2014). However, exposure to petroleum hydrocarbons is influenced by numerous factors, e.g., soil structure, organic carbon content, and oil burial, degradation, and desorption rates (Duan et al. 2017) that change over time, and exposure may eventually decrease below critical thresholds. Oiling also degrades soil environmental quality by affecting physical, geochemical, and biological properties such as shading, oxygen supply (Levine et al. 2017), microbial transformations (Mills and McNeal 2014), and belowground plant structure (Culbertson et al. 2007). Impacts on soil quality may be very long lived following a spill (years to decades, Culbertson et al. 2007; Okoro et al. 2011), and soil quality in affected areas may not become equivalent to that in reference sites even after remediation (Trujillo-Narcia et al. 2012). Furthermore, indirect effects of oil exposure (e.g., toxicity differentially expressed on predators or competitors, Brunner et al. 2013) can influence species interactions and community dynamics that also play an important role in pollution effects and ecosystem recovery (Fleeger et al. 2003). For example, food supply may indirectly influence consumer recovery if species at the base of the food chain recover slowly from a pollution event, causing the flow of energy to decrease, slowing recovery at higher trophic levels (Thorne and Thomas 2008). As a result of these influences, the factors that promote resiliency after a spill can vary greatly within (Mendelssohn et al. 2012) and among habitats (Peterson et al. 2003) as well as over time. A better understanding of the relative importance of the factors that influence recovery is necessary to formulate effective mitigation and remediation strategies in the aftermath of oil spills.

In 2010, the Deepwater Horizon (DWH) oil spill released over three million barrels (0.51 × 106 m3) of crude oil (US District Court 2015), exposing the Gulf of Mexico to an unprecedented potential for environmental damage. Approximately 2100 km of US Gulf Coast shoreline was oiled, including more than 1000 km of sensitive coastal wetlands within the Mississippi River Delta, the nation’s largest and most productive wetland-estuarine environment (Nixon et al. 2016). The salt marshes of northern Barataria Bay, LA, were among the areas that received the most severe oiling and have been the focus of numerous studies describing the initial impacts of the DWH on the marsh ecosystem (reviewed by Rabalais and Turner 2016). The initial impacts in heavily oiled areas include nearly 100% plant mortality (Lin and Mendelssohn 2012; Hester et al. 2016) and significant reductions to benthic microalgae (Fleeger et al. 2015), insects (McCall and Pennings 2012; Pennings et al. 2014), infauna (Brunner et al. 2013; Fleeger et al. 2015; Husseneder et al. 2016), and macroinvertebrates (Silliman et al. 2012; Zengel et al. 2015; Pennings et al. 2016; Zengel et al. 2016a, 2016b; Deis et al. 2017). Furthermore, some studies are of sufficient duration to detect improvements in environmental quality and the initial stages of biological recovery. Although no studies have yet indicated complete recovery, environmental quality began to improve and biological recovery was initiated at oiled sites within ~ 2–3 years after the spill (Lin et al. 2016; Deis et al. 2017; Fleeger et al. 2018). Studies that examine the factors that influence recovery from large-scale disturbances, including oil spills (Peterson et al. 2003), over even longer time periods (> 2–3 years) are rare in general (e.g., Balthis et al. 2017), but are needed to fully quantify resiliency.

Here, we focus on the roles of foundation species, soil quality, and a food resource (i.e., benthic microalgae) in long-term recovery of the salt marsh infaunal community impacted by the DWH. Foundation species enhance biotic diversity in salt marshes by habitat modification that ameliorates harsh environmental conditions and by supporting consumer trophic function (van der Zee et al. 2016). Certainly, foundation-species function is impaired when marshes are denuded of plant structure. Further, the two co-dominant foundation species in Louisiana coastal wetlands (Spartina alterniflora and Juncus roemerianus, hereafter Spartina and Juncus) differ in their tolerance to oiling and are recovering at different rates from the DWH (Lin and Mendelssohn 2012; Lin et al. 2016). Benthic microalgae serve as a principal food resource of many salt marsh invertebrates (Galván et al. 2011) and are susceptible to oiling (Fleeger et al. 2015). Although soil quality may be assessed by many geochemical and biological measures, live belowground root and rhizome biomass and bulk density are significant indicators of soil quality for infauna in constructed marshes (Levin and Talley 2002; Craft and Sacco 2003) where benthic consumers require years to decades before densities become equivalent to reference sites (Warren et al. 2002; Craft et al. 2003). These same factors may influence recovery following oil spills (Culbertson et al. 2007).

The Deepwater Horizon Trustees (2016) reviewed numerous individual studies conducted in estuarine, coastal, and deep-sea ecosystems to produce a broad synthesis of the effects of the DWH. However, it is important to examine integrated biotic components to comprehensively assess ecosystem recovery within a habitat. Studies by Lin and Mendelssohn (2012), Lin et al. (2016), Deis et al. (2017), and Fleeger et al. (2015, 2018), conducted jointly at the same sites along a gradient of oiling intensity, have amassed data on a diversity of chemical and physical factors (e.g., total petroleum hydrocarbon concentrations decreased from ~ 500 to 200 mg g−1 dry soil at heavily oiled sites by 2-year post spill), as well as biological factors (e.g., plant community composition, belowground biomass of roots and rhizomes, and macroinvertebrate abundance) that may interact to influence salt marsh recovery. Some of these ecosystem components recovered quickly, e.g., live aboveground biomass of Spartina alterniflora recovered by 2–3 years (Lin et al. 2016) and benthic microalgae rebounded to reference values by 2 years (Fleeger et al. 2015). Other components (e.g., belowground plant biomass and some infaunal and epifaunal taxa) have remained impacted over the duration of the study (Fleeger et al. 2015, 2018; Deis et al. 2017). Fleeger et al. (2018) examined the responses of individual infaunal taxa, but not the community, after the DWH along with a limited number of these biological and environmental factors to assess their importance to recovery. Their findings suggested that Spartina recovery over the first 2 years after the spill aided the recovery of some infauna but that soil quality became more important thereafter. Here, through the use of multivariate statistics, we quantify the contributions of a broader range of environmental factors to variation in the infaunal community as recovery progressed from 2 to 6.5 years after the spill. We hypothesize that foundation species, soil quality, and benthic microalgae promote infaunal resiliency and that the importance of these factors to recovery of the community varies across time.

Methods

Infaunal Community

Two datasets from reference oiled sites were examined to generate a regionally comprehensive analysis of the effects and recovery of the metazoan infaunal community. “Short-term sites” were located at coordinates 29.2650° N, 89.5310° W and consisted of replicated (n = 4), 50 m2 plots located along a linear shoreline distance of ~ 800 m (Zengel et al. 2015). Data from two treatments were used in our analysis: reference (as designated by Zengel et al. 2015 but designated lightly oiled by Fleeger et al. 2015) and heavily oiled plots. Plots were not cleaned of oil prior to or following establishment, and those designated as heavily oiled remained among the most severely impacted areas in the region (Zengel et al. 2015). The second dataset (“long-term sites”) was generated from samples collected at 21 sites located over an area 8 km × 5 km, between coordinates N 29.44060°and 29.47459°, W 89.88492° and 89.94647°. Long-term sites consisted of replicated (n = 7) salt marsh shorelines designated as “reference” (RF, no visible oiling), “moderately oiled” (MD), or “heavily oiled” (HV) (see Lin et al. 2016). Because oil was primarily transported into the estuary by south and southeasterly winds, heavily oiled sites generally occurred on south- and southeast-facing shorelines, while moderately oiled sites generally occurred on adjacent tangential shorelines (Lin et al. 2016). No oil cleanup treatments were applied at these sites. Reference sites were located along north- and south-facing shorelines located between ca. 0.5 and 4 km from oiled stations. A map locating all sampling sites is available (Online Resource 1).

To quantify the infaunal community, soil cores were collected from each plot or site on various sampling occasions (Fleeger et al. 2015, 2018). Samples from the short-term sites were collected 13, 22, and 29 months after the spill (i.e., May, 2011, February, 2012, and September, 2012), for a total of 24 samples. At the long-term sites, samples were taken at each site at approximately 6-month intervals on 12 occasions 1.5–6.5 years after the spill (i.e., October, 2011, April and November, 2012, April, September, and November, 2013, April and October, 2014, June and October, 2015, May and October, 2016). Four of the seven sites from each oiling category were randomly selected on each sampling date for processing, and a total of 150 samples were examined. Core samples were processed through 0.5- and 0.032-mm sieves, and specimens retained on the finer sieve were later identified and enumerated to higher taxonomic rank (e.g., nematodes, annelids) while simultaneously enumerating the sabellid polychaete Manayunkia aestuarina. Species composition of some higher taxa was examined in detail from a subset of individuals from the samples. All tanaids examined from the subsets were identified as Hargeria rapax, and all kinorhynchs were identified as Echinoderes coulli; species names rather than the name of the higher taxon were used in all analyses here.

Environmental Factors

We analyzed various physical and biological measures taken simultaneously with infauna on each sampling event from the long-term sites with the goal of identifying factors most important to recovery. Although a large number of environmental variables were measured, visual exploration of environmental data plotted over time and by oiling level and preliminary multivariate analysis suggested that many had little influence on or correlation with the benthic community. Those not selected for detailed analysis include soil redox at 2- and 15-cm depths, the photosynthetic pigments zeaxanthin and fucoxanthin, and eight ions (Ca2+, Cu2+, Mg2+, Na+, S, Zn2+, K+, and P) present in the soil. The soil shear strength was also measured but because belowground biomass and Spartina biomass may control the strength and stability of soils after the DWH (and thus shear strength, Lin et al. 2016), we did not include it in further analyses to avoid using variables that were highly correlated. Eight environmental measures were selected for additional study: total petroleum hydrocarbons (hereafter TPH), live aboveground biomass of Spartina and Juncus (hereafter referred to as Spartina and Juncus biomass), dead aboveground plant biomass (hereafter dead biomass), belowground live biomass (hereafter belowground biomass), soil pH, soil bulk density, and a proxy for the biomass of soil-dwelling microalgae (chlorophyll a concentration, hereafter Chl a). TPH were measured gravimetrically after extraction with a solvent (Lin et al. 2016) and expressed as mg TPH g−1 dry soil. We determined plant aboveground biomass (expressed as g m−2) at each site as all plants rooted within a quadrat were clipped to the ground surface, separated into live and dead components by species and weighed after drying (Lin et al. 2016). We determined live belowground biomass (expressed as g m−2) from the upper 6 cm from soil cores by washing soil over a 2-mm mesh sieve, separating live roots and rhizomes, and weighing them after drying to a constant mass at 60 °C. The soil pH level was determined by mixing 10 g of dry soil from the top 15-cm depth with 10 mL of deionized water for a 2-h equilibration period, and then measuring with a pH meter. The soil bulk density was determined for the upper 15 cm of soil after collection with a 5-cm-diameter semi-cylinder peat corer. Cores were sealed in pre-weighed watertight plastic bags, kept on ice in the field, and transported to the laboratory. Each core was dried at 60° C in a forced-air oven to constant weight, and density was expressed as g cm−3. Finally, photosynthetic pigments were extracted from core samples, and high-pressure liquid chromatography (HPLC) was used to quantify Chl a concentration (in μg cm2) following Buffan-Dubau and Carman (2000). Chl a was not measured in a few MD and HV replicates because the high volume of oil might have damaged the HPLC column. One or two replicates were unavailable at HV on 5, and at MD on 4, of the 12 sampling events.

Statistical Analysis

We used multivariate analysis to examine the response of the infauna to the spill. The community (consisting of nematodes, adult and juvenile copepods, copepod nauplii, E. coulli, ostracods, M. aestuarina, annelids excluding M. aestuarina, amphipods, H. rapax, isopods, juvenile bivalves, and juvenile gastropods) was analyzed by non-parametric multi-dimensional scaling (nMDS) after square-root transformation and with the Bray Curtis similarity index (PRIMER Version 6, Clarke and Gorley 2006). Two tests were conducted; the first combined short- and long-term datasets to examine trends across a broader range of conditions from 13 month–6.5 years after the spill, and in the second, the long-term sites were analyzed from 2 to 6.5 years after the spill. In addition, PERMANOVA (permutational multivariate analysis of variance) (Anderson et al. 2008) was used to test for variation in the community at the long-term sites from 2 to 6.5 years after the spill associated with oiling category (RF, MD, and HV), sampling time, and their interaction (α = 0.05).

A series of analyses was conducted to evaluate the influence of environmental factors on the meiobenthic community at the long-term sites. PRIMER’s RELATE procedure was conducted to determine if variation in environmental variables correlated with variation in the community. The significant result of RELATE (see below) prompted us to evaluate which environmental attributes best explained community variation by applying a distance-based linear model (DISTLM) approach using PERMANOVA software (Anderson et al. 2008). The environmental variables were tested for collinearity by using Draftsman’s plots and with a Spearman’s correlation matrix; all correlation coefficients were < 0.45 and all attributes were included in the model. TPH data were transformed as log(x + 1) to reduce skewness, and environmental data were normalized before constructing a matrix of dissimilarities between samples based on Euclidean distances. Adjusted R2 was used as the selection procedure, and a forward selection criterion was adopted to test the importance of the environmental variables individually and together as a model; the procedure was run with 9999 permutations. Distance-based redundancy analysis (dbRDA) was used to examine the influence of these predictor environmental variables on the multi-dimensional distribution of samples across time and oiling category (Anderson et al. 2008). DISTLM and dbRDA were conducted in two ways to determine if the importance of environmental factors varied as recovery progressed; first, all data from the long-term sites (from 2 to 6.5 years after the spill) were analyzed and second, data from 5 to 6.5 years after the spill were analyzed. The time frame from 5 to 6.5 years was chosen because taxa continued to rebound over time, and the factors associated with recovery could thus explain an increasing amount of the community variation.

SPSS, version 23, was used to create a matrix of Kendall’s tau correlation coefficients calculated between individual taxa and environmental variables across all sampling events and oiling levels at the long-term sites (n = 123). A p value < 0.05 was considered significant in these tests; however, because many tests (88) were performed, Bonferroni’s correction factor was also calculated to better interpret individual correlation tests.

Data Availability

Data are publicly available through the Gulf of Mexico Research Initiative Information & Data Cooperative (GRIIDC) at https://data.gulfresearchinitiative.org (doi: https://doi.org/10.7266/N7P26W1S, doi:https://doi.org/10.7266/N7DN430F, doi: https://doi.org/10.7266/N7J9649F).

Results

Effects of and Recovery from Oiling on the Community

Data from all available infaunal samples taken from the region after the spill were examined using nMDS. The community at HV from the short-term sites 13–29 months after the spill and from the long-term sites at 18 months post spill was dissimilar from other HV samples as well as MD and RF samples taken over 6.5 years; these HV samples extended in a distinct arc to the lower left of the ordination plot (Fig. 1a). When only the long-term sites (from 2 to 6.5 years after the spill) were examined, RF samples were mostly positioned in the upper half of the nMDS plot while HV samples were mostly positioned in the lower half with MD samples intermediate (Fig. 1b). The community at HV also displayed a pattern of change over time as many samples early in the recovery period were to the right of the center of the plot and many samples later in the period extended to the left of the plot (Fig. 1b). PERMANOVA revealed that the community from 2 to 6.5 years post spill differed among oiling categories and sampling events but without interaction (for oiling category and time, p = 0.001 and for interaction, p = 0.251).

Fig. 1
figure 1

a, b nMDS plots of the meiobenthic community following the DWH. Numbers refer to months after the spill. The upper plot (a) is from 13 months to 6.5 years post spill at short- and long-term sites (short-term site samples are 13, 22, and 29 months after the spill), and the lower plot (b) is 2–6.5 year post spill at long-term sites. RF, reference; MD, moderately oiled; and HV, heavily oiled. The oval highlights HV samples taken early in the recovery period in the upper figure

Full size image

Patterns of Variation in Physical and Biological Attributes

Physical factors exhibited distinctly different temporal trends associated with oiling during the period of our study (from 2 to 6.5 years post spill, Fig. 2). TPH averaged 0.3 ± 0.22 at RF, 5.9 ± 14.5 at MD, and 95.9 ± 121.0 mg g−1 dry soil at HV across time. However, TPH did not decrease at HV after year 2, while TPH decreased by 66% at MD from years 2–4 to years 5–6.5. The mean Chl a varied relatively little (ranging from ~ 2 to 4 μg cm−2), except at HV sites during years 2–3 and after years 4–5 when means were ~ 5–12 μg cm−2. Spartina biomass began to rebound 2–3 years after the spill and continued to increase in years 3–4; means varied minimally thereafter. Spartina biomass at HV remained low over time (e.g., HV means were 19% lower than at RF after year 4). In contrast, Spartina biomass at MD was consistently higher than that at RF over time (e.g., ~ 22% higher after year 4). Juncus did not rebound at HV sites (where mean biomass remained near 0 g m−2 over the duration of the study), and means at MD were lower than those at RF by ~ 17% after year 4. However, Juncus biomass tended to increase over time at RF and MD. Dead biomass and belowground biomass had similar trends relative to oiling intensity; highest values were at RF and consistently lower values, even after year 4, were observed at MD (MD means were 32% and 18% lower than RF for dead biomass and belowground biomass respectively) and at HV sites (HV means were 80% and 70% lower than RF for dead biomass and belowground biomass respectively). However, dead biomass decreased over time at all sites while belowground biomass increased. Bulk density and soil pH level were lowest at RF for each sampling date. Bulk density remained higher at oiled sites over time, by 31% at MD and 35% at HV compared to RF, after year 4. Although highest values of pH values were at HV for years 2–4 after the spill, pH values at the three oiling levels became more similar after year 4.

Fig. 2
figure 2

a–h Environmental variables plotted by years after the spill at the long-term sites presented as mean ± SD. Sample sizes ranged from 12 to 21 per oiling category across years. Oiling level designations as in Fig. 1

Full size image

Patterns of Variation in Infauna

Nematodes, copepods (and nauplii), and annelids (excluding M. aestuarina) at oiled sites increased to densities similar to or higher than those of RF by 2–4 years after the spill, indicating a potential for rapid recovery (Fig. 3). However, some of these taxa continued to increase at all oiling levels thereafter reaching very high densities after year 4 at RF with even higher densities at HV (e.g., nematodes were ~ 50% higher at HV than RF after year 4). Juvenile bivalves, amphipods, and H. rapax began to increase from very low densities after year 3 at all oiling levels (including RF), but also reached very high densities after year 4 at RF and even higher densities at HV (for example, amphipods averaged ~ 2× higher in density at HV than RF after year 4). Ostracods and juvenile gastropods increased somewhat in abundance; densities at MD were similar to those at RF by years 4–5 but remained much lower at HV than RF (25% lower for ostracods and 44% lower for gastropods) after year 4 (Fig. 4). Manayunkia aestuarina and E. coulli however proved to not be resilient as densities remained low over time at MD and HV (means were 61% lower at MD and 68% lower at HV for M. aestuarina and 52% lower at MD and 88% lower at HV for E. coulli after year 4).

Fig. 3
figure 3

a–f Density responses (mean ± SD) of selected meiobenthic taxa at the long-term sites to oiling category by years after the spill. Patterns of copepod nauplii (data not shown) closely paralleled adult and copepodite copepods. Sample sizes ranged from 12 to 21 per oiling category across years. Oiling level designations as in Fig. 1

Full size image
Fig. 4
figure 4

a–d Density responses (mean ± SD) of selected meiobenthic taxa at the long-term sites to oiling category by years after the spill. Sample sizes ranged from 12 to 21 per oiling category across years. Oiling level designations as in Fig. 1

Full size image

Linking Community Variation with Environmental Attributes

RELATE indicated that variation associated with the environmental variables tested was significantly correlated with variation in the infaunal community from 2 to 6.5 years after the spill (Spearman correlation = 0.116; p = 0.003).

DISTLM for the period from 2 to 6.5 years post spill period explained 27.3% of the variation in the meiobenthic community, and contributions from seven of the eight environmental variables achieved the best solution (Table 1). All measures of aboveground plant biomass contributed and were significant; Spartina biomass explained 8.3%, Juncus biomass 5.3%, and dead biomass 3.2% of the total variation. Chl a was also significant and explained the second highest (7.7%) fraction of the variation. Bulk density, belowground biomass, and TPH contributed to the model (and together explained ~ 2% of the variation) but were not significant. pH level was not included in the best solution of the model.

Table 1 DISTLM sequential test results listing factors that significantly explain meiobenthic community variation from 2 to 6.5 years post spill. Spartina, Juncus, and dead refer to aboveground plant biomass, BG refers to belowground live biomass, and BD refers to bulk density. Proportion = proportion of variation of the community explained by each variable
Full size table

DISTLM conducted for the period from 5 to 6.5 years post spill explained 83.2% of the variation in the meiobenthic community, and contributions from seven environmental variables achieved the best solution (Table 2). Belowground biomass explained the greatest proportion (37.7%) of community variation. All measures of aboveground plant biomass contributed and were significant; Spartina biomass explained 18.1%, Juncus biomass explained 11.7%, and dead biomass explained 4.7% of the community variation. TPH and Chl a were also significant and explained 6.3 and 3.9% of the variation, respectively. Bulk density explained ~ 1% of the variation but was not significant. pH level was not included in the best solution of the model.

Table 2 DISTLM sequential test results listing factors that significantly explain meiobenthic community variation from 5 to 6.5 years post spill. Spartina, Juncus, and dead refer to aboveground plant biomass, BG refers to belowground live biomass, and BD refers to bulk density. Proportion = proportion of variation of the community explained by each variable
Full size table

On the dbRDA plot for 2–6.5 years post spill (Fig. 5), axis 1 explained 68% of the fitted and 18.6% of the total variation, respectively, while axis 2 explained 24.2% and 6.6% of the fitted and total variation respectively. Vectors for Spartina biomass and Chl a were closely aligned and in the same direction, almost parallel with axis 1. RF, MD, and HV samples were, however, scattered across similar distances along axis 1. Vectors for Juncus biomass and belowground biomass were closely aligned and in the same direction, almost parallel to axis 2. The vector for bulk density was also similar in orientation to axis 2 but was opposed to Juncus biomass and belowground biomass. Vectors for TPH and dead biomass were oriented in opposite directions and not closely aligned with either axis. RF, MD, and HV samples were strongly separated along axis 2 with HV forming a tight grouping extending in a line parallel to axis 1. MD and RF samples were partially intermingled and scattered almost equally along the dimensions of axes 1 and 2.

Fig. 5
figure 5

Distance-based RDA ordination of the meiobenthic community based on abundance data as predicted by a DISTLM for the long-term sites from 2 to 6.5 years post spill. The environmental variables that best explain the variation in meiofaunal abundance at the three oiling categories are shown. Vectors indicate direction of the effect of quantitative variables in the ordination plot. Sp, Spartina biomass; Ju, Juncus biomass; Chl, Chl a concentration; dead, dead biomass; BD, bulk density, BG, belowground biomass

Full size image

On the dbRDA plot for 5–6.5 years post spill (Fig. 6), axis 1 explained 56.2% of the fitted and 48.4% of the total variation, respectively, while axis 2 explained 25.2% and 21.7% of the fitted and total variation respectively. Vectors for belowground biomass, Juncus biomass, and dead biomass were oriented in the same direction along axis 1. Vectors for TPH and Chl a were similar in orientation with dead biomass but in the opposing direction. The vector for Spartina biomass was almost parallel to axis 2. HV samples were distinct from MD and RF along axis 1, and were in a tight line parallel to axis 2. MD and RF samples were partially intermingled and scattered almost equally along axis 1 and 2.

Fig. 6
figure 6

Distance-based RDA ordination of the meiobenthic community based on abundance data as predicted by a DISTLM for the long-term sites for the last 18 months of the study. The environmental variables that best explain the variation in meiofaunal abundance at the three oiling categories are shown. Vectors indicate direction of the effect of quantitative variables in the ordination plot. Sp, Spartina biomass; Ju, Juncus biomass; Chl, Chl a concentration; dead, dead biomass, BD, bulk density; BG, belowground biomass

Full size image

Nematodes, copepods, nauplii, and H. rapax were positively correlated with both Spartina biomass and Chl a (Table 3). Amphipods, bivalves, and gastropods were also positively correlated with Spartina biomass. Taxa that were positively correlated with Spartina biomass (except gastropods and H. rapax) were negatively correlated with dead biomass but with few other variables. Echinoderes coulli and M. aestuarina were negatively correlated with TPH, pH level, and bulk density but positively correlated with Juncus biomass, dead biomass, and belowground biomass. Bonferroni’s correction factor reduced the accepted p value for significance to p < 0.0006, and the number of significant correlation tests was decreased from 32 to 15, although some tests remained significant for each environmental factor and for most taxa.

Table 3 Kendal's tau correlation matrix, including p values, comparing the density of various taxa and simultaneously collected environmental factors, p < 0.05 in bold font and n = 123. Bonferroni’s correction factor indicates significant values for p < 0.0006. Spartina, Juncus, and dead refer to aboveground plant biomass, BG refers to belowground live biomass, and BD refers to bulk density. Annelids excludes M. aestuarina
Full size table

Discussion

The salt marsh meiobenthic community in northern Barataria Bay was strongly affected by oiling from the Deepwater Horizon spill. The intensity of oiling affected both the severity of initial effects and the rate of recovery. For example, our short-term sites experienced heavier and more persistent oiling than long-term sites (Fleeger et al. 2015; Zengel et al. 2015), and the infaunal community at oiled short-term sites remained distinct from reference samples 29 months post spill, while oiled long-term sites began to resemble reference sites by 24 months post spill (Fig. 1a). However, despite this relatively rapid initial rebound, the community at heavily oiled sites remained distinct from moderately oiled and reference sites 6.5 years after the spill (Fig. 1b and PERMANOVA). Extrapolation of the observed pace of recovery indicates that complete recovery at moderately and heavily oiled sites will likely take much longer than a decade, which is slower than reported in many previous studies of oil spill impacts on meiofauna (Giere 2009). However, oil spills in salt marshes can negatively affect benthic invertebrates for over 4 decades (e.g., Culbertson et al. 2008). Further, chronic exposure to petroleum hydrocarbons over decades can influence the sensitivity and resilience of benthic invertebrates to future exposure events (Carman et al. 2000).

Of the eight environmental factors included in the DISTLM, seven contributed to the best solution of the model (Tables 1 and 2). Spartina biomass (8.3%) and benthic algal biomass (7.7%) explained the highest proportions of community variation over the duration of the study and varied in a manner that suggested a direct and positive influence on faunal recovery. Recovery of Spartina and microalgae at oiled sites began 2–3 years post spill, and the densities of nematodes, copepods (nauplii, copepodites, and adults), H. rapax, juvenile bivalves, and amphipods began to rebound with these factors either simultaneously or within ~ 2 years. These positive correlations remained throughout the study period; microalgal biomass reached much higher levels after years 4–5 at heavily oiled sites when nematodes, copepods, amphipods, and H. rapax reached their highest densities (Fig. 3). However, a different pattern was observed for other environmental factors and taxa. Belowground biomass, total petroleum hydrocarbon concentrations, bulk density, Juncus biomass, pH level, and dead biomass at oiled sites either did not rebound or began recovery much more slowly, especially at heavily oiled sites, and generally did not reach values equivalent to those at reference sites. A similar lack of resiliency was observed in E. coulli and M. aestuarina (Fig. 4) as well as for ostracods and juvenile gastropods (Fleeger et al. 2018). Failure of these environmental factors to rebound indicates that environmental quality had not sufficiently recovered from oiling to support the recovery of the infaunal community as a whole.

The recovery of Spartina and benthic microalgae both directly contributed to the recovery of infauna but likely by different mechanisms. Spartina alters the habitat by slowing the flow of water, adding oxygen, enhancing basal food resources by adding dissolved and particulate organic matter to the soil, increasing soil shear strength, and by providing shelter as a foundational species. The removal of Spartina (e.g., by clipping) in the absence of oiling greatly affects soil-dwelling microflora and infauna (Whipple et al. 1981; DeLaune et al. 1984; Whitcraft and Levin 2007). On the other hand, benthic microalgae serve as a food resource for many estuarine infauna and epifauna. Taxa that correlated with Chl a (i.e., nematodes, copepods, ostracods, and tanaids) are often characterized as grazers of benthic microalgae (Carman et al. 1997; Pascal et al. 2013), while the diet of M. aestuarina (which did not correlate with Chl a) consists primarily of phytoplankton (Galván et al. 2008).

Previous investigation linked the earliest stages of faunal recovery (< 2 years after the spill) with the recovery of Spartina at our sites (Fleeger et al. 2015), supporting the general conclusion that ecosystem recovery tracks foundation species recovery (e.g., McGlathery et al. 2012). However, the manner in which Spartina and Juncus responded to the spill, and how infauna interacted with each, differed greatly over the longer term. Taxa that were positively correlated with Spartina biomass did not correlate with Juncus biomass or vice-versa. Juncus is intolerant of oiling (Lin and Mendelssohn 2012) and has not recovered. Juncus biomass was consistently much higher at reference than oiled sites, as were the densities of E. coulli and M. aestuarina, attesting to the positive correlation between Juncus and these two species and the large contribution (11.7%) of Juncus to explain community variation during the last 18 months of the study. However, M. aestuarina and E. coulli are both abundant residents in Spartina-dominated marshes (Higgins and Fleeger 1980; Johnson et al. 2007), suggesting that factors other than Juncus biomass control their recovery (e.g., poor soil quality may have slowed the recovery of infauna and Juncus).

Factors associated with the soil environment (e.g., belowground biomass and bulk density) have been implicated as important to the reestablishment of infauna in constructed salt marshes, which can require years, or even decades, before densities (Craft et al. 2003) and food webs (Nordstrom et al. 2015) become equivalent to reference marshes. Belowground biomass explained the greatest proportion of community variation (37%) during the last 18 months of our study. In general, roots and rhizomes improve environmental quality by releasing oxygen and dissolved organic matter, increasing sediment volume, and by providing physical structure that binds sediment. Variation in bulk density in restored marshes has been linked with plant recovery; an increasing biomass of roots and rhizomes decreases bulk density (Craft et al. 1999). The density of M. aestuarina was previously related to belowground biomass and bulk density during marsh restoration (Craft and Sacco 2003). Furthermore, Craft and Sacco (2003) estimated that 500 g m−2 of live belowground biomass is necessary to support infauna, including M. aestuarina, at densities comparable to reference sites, and this level was reached after ~ 4 years at our moderately oiled sites but was not reached even after 6.5 years at heavily oiled sites. These observations suggest that longer-term recovery for infauna is similar in nature and timing to that associated with reestablishment in constructed marshes and that the same factors control recovery after an oil spill. Thus, soil quality may be most important for long-term infaunal recovery when an oil spill is intense enough to cause extensive plant mortality.

Total petroleum hydrocarbon concentration was a significant but minor contributor (< 1%) to variation in the infaunal community over the duration of the study. Total hydrocarbon concentration at heavily oiled sites decreased from 9 to 24 months after the spill (Lin et al. 2016), but remained constant thereafter. Even though concentrations remained elevated relative to reference sites 24 months post spill, hydrocarbon toxicity likely declined at heavily oiled sites (Duan et al. 2017). Such a decline in toxicity is supported by the observed recovery pattern of amphipods which are known to be highly sensitive to oiling (Gesteira and Dauvin 2000); amphipod abundance was very low at all sites for 3–4 years post spill and then increased by ~ 10× at heavily oiled compared to reference sites after year 5.

Species loss and changes in species composition or functional characteristics after an environmental disturbance may strongly influence ecosystem properties including ecological services (Hooper et al. 2005). The activities of meiofauna modify physical, chemical, and biological sediment properties that affect ecosystem services including sediment stabilization and biochemical cycling, and meiofauna serve as important intermediates in benthic-based food webs (Schratzberger and Ingels 2017). Fleeger et al. (2018) reported that the density of total meiofauna recovered within 6.5 years of the DWH, although changes in infaunal species composition and relative abundance as noted here indicate a modified community, especially at heavily oiled sites. At least some of the value of infauna as food for higher trophic levels (such as shrimp and fish) was likely restored at oiled sites in the form of copepods and amphipods (Fry et al. 2003). Recovery of the density of total meiofauna suggests that at least some of the ecological services associated with sediment stabilization and biochemical cycling also recovered. However, M. aestuarina did not recover, and this species potentially contributes to food webs and sediment-related services in a unique fashion. This tube dweller comprises about 5% of the total meiofauna abundance (one of the most abundant single species) at reference sites and functions as a surface-dwelling suspension feeder that grows into the macroinfaunal size range (Galván et al. 2008). As such, it likely interacts strongly with the soil environment. Furthermore, surface-dwelling macroinfauna are important prey of crustaceans and fish (Bell 1980; Bonsdorff et al. 1995; Hentschel et al. 2017). The near absence of this species could indicate significant alteration of ecological function at heavily oiled sites as has been found in soil-based ecosystems (Wagg et al. 2014), including salt marshes (McCann et al. 2017) when species diversity decreases.

Foundation species, sediment quality, and a food resource (a proxy for benthic microalgal biomass) were all significantly linked with infaunal recovery. But of all the factors studied, the recovery of Spartina provided the greatest positive impact to infaunal recovery (Table 1). In the short term, habitat modifications (perhaps most closely associated with the recovery of Spartina stems, Fleeger et al. 2018) were likely most important to the reestablishment of infaunal populations. In a test of this hypothesis conducted as an experimental companion study to our long-term monitoring, Johnson et al. (2018) found that planting Spartina in bare soil that had been heavily oiled stimulated the initial colonization of infauna by increasing recruitment of the opportunistic polychaete Capitella spp.; nutrient enrichment increased benthic microalgal biomass but did not affect infaunal colonization. Over the long term, however, the continued recovery of Spartina likely also improves soil quality as belowground organic matter accumulates. Roots and rhizomes not only improve the soil environment by affecting soil structure (e.g., decreasing bulk density) but also add to the detrital pool that improves food supply and food web support, functions also associated with a foundation-species effect (van der Zee et al. 2016). Therefore, we recommend that future spill mitigation strategies include planting foundation species in bare soil to improve the initial colonization of infauna, and thus enhance the ecosystem services they provide. However, neither Spartina aboveground biomass nor belowground biomass fully recovered at our heavily oiled sites, suggesting that restoration methods that increase Spartina biomass over the long term should be considered. One possible method would be to fertilize initial plantings. In our companion planting experiment conducted on bare soil that was initially heavily oiled (Johnson et al. 2018; Lin, unpublished), fertilization increased Spartina above- and belowground biomass for the duration of a 2.5-year study. Graham and Mendelssohn (2016) concluded that nutrient enrichment can enhance the development of functional processes following a disturbance by facilitating plant establishment and by stimulating initial belowground growth. However, nutrient enrichment has the potential to negatively impact soil organic matter content as plants in mature communities respond to excess nutrient availability by allocating more resources to above- and less to belowground structure (Deegan et al. 2012; Graham and Mendelssohn 2016).