Water-column Measurements and Observations from the Deepwater Horizon Oil Spill Natural Resource Damage Assessment

The 20 April 2010 blowout, explosion, and fire on the mobile drilling rig, Deepwater Horizon, resulted in the tragic loss of 11 personnel and initiated the largest oil spill in U.S. history. Located offshore approximately 50 miles (80 km) southeast of the Mississippi River Delta in water nearly one mile (1.5 km) deep, the Macondo well blowout continued for 87 days and released 3,190,000 barrels (134,000,000 gallons) of oil (US District Court 2015; French McCay et al., 2015; Spaulding et al., 2015). Before the well was capped on 14 July 2010, the total volume was 12 times larger than the 1989 Exxon Valdez oil spill in Prince William Sound Alaska (equivalent to one Exxon Valdez tanker spill every 7 days), and the cumulative area of detectable oil measured using satellite-directed synthetic aperture radar exceeded 43,300 miles² (112,100 km²), an area about the size of Virginia (NOAA 2016). In response to the spill, over 1,840,000 gallons of dispersant (mostly Corexit 9500) were applied: ~1,070,000 gallons sprayed at the surface from boats and aircraft and ~772,000 gallons injected directly into the oil plume as it escaped the wellhead at depth (Lehr et al., 2010).

Herein, we draw upon several reports from the NRDA Administrative Record (Payne and Driskell, 2015a, 2015b, 2015c, and 2015d) and the Final Programmatic Damage Assessment and Restoration Plan (PDARP) and Final Programmatic Environmental Impact Statement (PEIS) (available at http://www.gulfspillrestoration.noaa.gov/restoration-planning/gulf-plan/). Detailed interpretations of our work regarding the final oil fate and transport in the deepwater plume have been submitted to Marine Pollution Bulletin (Payne and Driskell, 2017; Driskell and Payne, 2017).

Due to the unprecedented depths and scale of the Deepwater Horizon (DWH) blowout, it was necessary to develop a number of adaptive sampling strategies to document the oil’s fate and transport (Payne and Driskell, 2015a). Even in the initial planning stages, the sampling design was not intended to survey the entire spill impact. The scale of the event alone precluded efforts to sample the impacted region (eventually hundreds of km) in any comprehensive or statistically-designed manner. Instead, the offshore collections (primarily by ocean-going vessels at or beyond the shelf break) focused on sampling to understand the changing composition and fate of the oil in the rising plume above the blowout and in tracking the entrapped deep subsurface plume for hundreds of km SW across the Gulf (Payne and Driskell, 2016). Forensic assessments of each sample were intended to eventually serve as confirmation of spill impact models rather than as unbiased statistical estimates. Sampling approaches evolved early in the spill and eventually included:

• Commonly used rosette samplers and later, remotely-operated-vehicles (ROVs) (Fig. 1) were instrumented with real-time DO, fluorometry and CTD sensors to detect, track and sample within the deep plume (rather than collecting samples randomly or systematically at pre-assigned depths) (Fig. 2).

• ROVs were also equipped with real-time video and cameras plus 670 kHz sonar systems to document the subsurface plumes and measure droplet size distributions and concentration (Figs. 3 & 4). These and dedicated holographic-camera data were utilized by Li, et al., (2017) for developing an oil droplet size distribution model.

• Using ROV’s live video for precisely sampling near-bottom water and identifying and collecting floc samples, burn residues, and sediments without disturbing the ephemeral oil layer at the sediment-water interface (Figs. 5 and 6 – Details are provided in Stout and Payne 2016a, 2016b; Stout et al., 2016a, 2016b); and

• Filtered, phase-separated water samples were collected in the field (Payne et al., 1999; Payne and Driskell 2015a, 2015b, 2015c) to use in later parsing out phase information (dissolved versus particulate) in the routinely collected, unfiltered whole-water samples (Figs. 7 and 8).

When done properly, sampling the water column for hydrocarbons during or after an oil spill can be highly insightful, but the task is challenging with multiple opportunities for sample contamination, often without any feedback until weeks or months later when data come back from the lab. Detailed sampling protocols were formalized to ensure collection of uncontaminated oil and water samples (Payne and Driskell 2016) including:

• Sampling in cleaner areas first and working towards more contaminated areas whenever possible;

• Using GoFlo^® samplers that pass through the water surface as closed vessels to avoid surface slick contamination (then opening at ~ 10 m and descending through the water column open to allow complete flushing) vs. open-entry Niskin bottles;

• When necessary to sample through slicks (e.g., in more heavily oiled areas or when GoFlo Bottles^® were not available), using bow thrusters on the sampling vessels or a few drops of detergent (Dawn^®) in calm water to open surface slicks; and

• Returning a distilled or deionized (DI) rinse water sample, unused in its original container, along with normal equipment decontamination and field blanks to further assess potential contamination issues. Commercially available DI water is notoriously dirty at the ng/L (ppt) level (Payne and Driskell, 2016).

Due to expansion and breakage, water samples could not be frozen and thus had a maximum 14-day hold-time limit prior to extraction at the analytical lab. To meet these requirements offshore while generating thousands of samples, runner boats were used to intermittently pick up samples during the extended offshore cruises. During the year-and-a-half of sampling efforts after the Deepwater Horizon event, only 217 of 22,039 water samples (0.98%) were compromised by exceeding the maximum hold time.

The vast majority of the whole water, dissolved and particulate-oil fractions, and sediment/floc samples reviewed for the NRDA effort were analyzed by Alpha Analytical Laboratory (Mansfield, MA) for detailed hydrocarbon composition in accordance with the AQAP (NOAA 2014). Analyte lists and methods are detailed in Stout et al. (2016a) and the AQAP. All chemistry data were independently validated by EcoChem (Seattle, WA) as third-party validators. Publicly available online data (NOAA ERMA) are surrogate-recovery corrected; however, for the purpose of forensic analyses, the data were used uncorrected. Forensic methods are detailed in Payne and Driskell (2016).

For this publication, only the offshore cruise samples collected in 2010 (5,332) were considered, and of those, 4,189 were forensically characterized (Payne and Driskell, 2016 and 2017). Normally, forensic determinations for oiled matrices include match, indeterminate, or no-match categories (ASTM 2000). For water-column samples, “match” categories were further subdivided into phase assignments (i.e., dissolved, particulate, or unresolvable [unparseable]) generating seven categories to support oil fate and transport modelling (Table 1).

Forensic assessments covered 45 cruises with 1,766 matches (category 1–3) from 4,189 samples. Most MC252-matching samples were less than 1,000 ng/L (ppt) TPAH with category medians at 875, 177 and 30 ppt, for categories 1, 2 and 3, respectively. Extreme concentrations for these same three categories reached levels of 100,000, 10,000, and 5,000 ppt. Detailed descriptions of our forensic evaluations are in Payne and Driskell (2017), but in summary, our analysis of analytic data, field instruments, and observations led to the following findings:

Hydrocarbon chemistry and corroborating field data positively link Macondo oil to offshore oil in near-surface waters, rising through the water column near the wellhead, and entrained in an extensive deepwater oil plume (1,000–1,400m) advecting predominantly to the SW with occasional shallower lenses (Fig. 9). Benzene was largely removed by dissolution from rising droplets during ascent but remained at depth (~20–40 μg/L, ppb) out to 15 km while other BTEX components were detected at depth (10–80 ppb) to 20 km, and toluene was detected (< 5 ppb) up to 100 km from the wellhead (Payne and Driskell 2016 and 2017).

MC252 oil was identified in the entrapped deep plume as particulate-phase hydrocarbons (Category 1) up to 155 km from the wellhead, and as dissolved-phase positive matches (Category 2) as far as 267 km SW from the wellhead (Payne and Driskell 2015b, 2017). With increasing time and distance from the wellhead, the plume’s PAH signal eventually became phase-ambiguous (Category 3) and then diluted until it was no longer detectible using SIM GC/MS methods. And yet, even in the far field, the plume was still visible in the corroborating data (depth, fluorometry signals, DO sags, and persistent dispersant constituents). By these methods, the plume was detected out to 412 km from the wellhead but with no relevant PAH signal. We hypothesize that lack of oil signatures in these samples still exhibiting corroborating evidence for the oil’s presence may be, at least in part, attributable to oxygenated and polar microbial degradation products not detectible using SIM GC/MS methods (Aeppli et al., 2012, McKenna et al., 2013; Gutierrez et al, 2013a, 2013b; Bacosa et al., 2015; Bagby et al., 2016). Helping define plume boundaries and demonstrating an essentially clean background, non-matching or clean samples (the white symbols in Fig. 9B) were taken from various depths and directions up to 530 km from the wellhead.

Generally, near-surface oil samples showed evidence of substantial dissolution weathering as the oil droplets rose through the water column, as well as enhanced evaporative losses of lower-molecular-weight n-alkanes and aromatic hydrocarbons. Oil that had reached the surface also showed evidence of photo-oxidation of alkylated chrysenes (Stout and Payne, 2016a; Payne and Driskell, 2017). Typical of surface oil dynamics, near-surface increases in dissolved and particulate-oil fractions were observed as a result of wind-induced entrainment of surface films and dispersant effects (Driskell and Payne, 2017).

Effects of dispersant treatments, both in surface applications and injected at the wellhead, were seen in the oil profiles as unique, enhanced weathering patterns. The patterns suggested that as dispersants created micro-droplets, the increased surface area accelerated dissolution of less alkylated homologues of lighter-molecular-weight PAH (Figure 10). In this process, the PAH appeared to weather atypically faster than the SHC. These results also imply that dispersants were a functionally effective mediation treatment in this event (Driskell and Payne, 2017). For forensic characterization, dispersant-mediated hydrocarbon profiles were distinguished by:

• Presence of dispersant (or indicators), often at very high concentrations; DOSS, glycol ethers (GE), or 2-butoxyethanol (2BE);

• A very weathered PAH profile with accelerated loss of light ends up through dibenzothiophenes;

• A very distinctive water-washed chrysene pattern with C0<C1<C2<C3<C4;

• Significant SHC often present in samples within 5 km of the wellhead that, based on n-C₁₇/pristane and n-C₁₈/phytane ratios, were not completely microbially degraded;

• Dibenzothiophene/phenanthrene (D/P) ratios on atypical low-trend path in double-ratio plots (D2/P2 vs D3/P3).

• Decalins, the dominant “pseudo-PAH” in Corexit, may be present and sometimes high, and

• Dissolved-phase samples (the associated complements of filtered particulate-phase samples) usually contain excess fluorenes relative to phenanthrenes.

Optimal sampling approaches included: Combining multiple real-time sensors (DO, fluorometry, CTD, particle-size-analyzers, and video) with ROV and rosette-based sampling systems to detect and collect oil at depth; collecting filtered, phase-separated water samples to aid in parsing out dissolved versus particulate signatures in unfiltered samples; and using ROVs for sampling near-bottom water and identifying and collecting floc samples, burn residues, and sediments without disturbing the ephemeral oil layer at the sediment-water interface.

Benzene was largely removed by dissolution from rising droplets during ascent but remained in the dissolved phase at depth out to 15 km while other BTEX components were detected at depth up to 100 km from the wellhead. MC252 oil was identified in subsurface water samples as particulate-phase hydrocarbons up to 155 km from the wellhead, and as dissolved-phase as far as 267 km from the wellhead. Furthermore, based primarily on dispersant indicators, fluorescence and DO features, the presence of the plume was detected 412 km from the wellhead. Oxygenated hydrocarbons and other polar derivatives not detectible using SIM GC/MS methods may still be present in unknown concentrations in this distant plume.

Dispersant application at depth resulted in significantly enhanced dissolution of lower-and intermediate-molecular-weight PAH. The presence of dispersant indicators, measured for the first time in field-collected, particulate-phase oil samples at depth, plus the enhanced dissolution effect, documented the utility of dispersant injections at the wellhead.