Analysis of emissions and residue from methods to improve efficiency of at-sea, in situ oil spill burns

Abstract

The combustion efficiency of simulated at-sea surface oil burns (in situ burns) was determined in a 63 m³ tank while testing varied boom configurations and air-assist nozzles in the presence and absence of waves. Combustion efficiencies of Alaska North Slope oil based on unburned carbon in the plume emissions ranged from 85% to 93% while values based on oil mass loss ranged from 89% to 99%. A four-fold variation in PM_2.5 emission factors was observed from the test conditions. The most effective burns in terms of reduced emissions and post-burn residue concentration of total petroleum hydrocarbons were those that had high length to width boom ratios resulting in higher flame front surface area exposure to ambient air. The amount of oil mass lost was not related to any combustion efficiency parameters measured in the plume, representing a potential tradeoff between unburnt oil and air pollution.

Graphical Abstract

INTRODUCTION

At-sea, or in situ, oil spills are often treated by herding and concentrating surface oils followed by purposeful ignition at the apex of U-shaped boom systems. These controlled combustion, in situ burns (ISB) are employed to mitigate the potential environmental impact of floating and sinking oil (Allen et al., 2011). One undesirable consequence of the burns is the black plume formed by the combustion by-products. The plumes consist of major pollutants including nitrogen oxides (NOx), volatile organic compounds (VOCs), sulfur dioxide (SO₂), polycyclic aromatic hydrocarbons (PAHs), carbon monoxide (CO), and particulate matter (PM). Historical and more recent efforts have been undertaken to characterize these emissions including, for example, the comprehensive at-sea, Newfoundland offshore burn experiments (Canada, 1997), experimental pan burns (Benner et al., 1990), mesoscale tank burns (Fingas et al., 1996), at-sea sampling during the Deepwater Horizon spill (Aurell and Gullett, 2010; Perring et al., 2011; Middlebrook et al., 2012; Gullett et al., 2016), and laboratory simulations of sea surface oil burns (Gullett et al., 2017).

Emissions from burning oil are comprised of fine particulate matter (nominally <2.5 μm aerodynamic diameter, PM_2.5) and gaseous components. Measurements during the Deepwater Horizon disaster in the Gulf of Mexico found that 7.5% of the oil mass burned was emitted as fine particle mass into the atmosphere (Aurell and Gullett, 2010). These particles were 82% elemental carbon (EC) to which were bound PAHs at concentrations of 68 μg/g particle mass (Gullett et al., 2016). Trace amounts of polychlorinated dibenzodioxins and dibenzofurans (PCDDs/PCDFs), whose formation is commonly linked to combustion inefficiencies, were also found (Aurell and Gullett, 2010; Gullett et al., 2016).

A recent laboratory study (Gullett et al., 2017) with Bayou Sweet crude simulated the Deepwater Horizon burns, finding ~6% of the oil by mass became a particle emission of aerodynamic diameter 1 μm, over 90% of which was light-absorbing black carbon (BC). The gas phase modified combustion efficiency declined throughout the burns to 97.8%; incorporation of the carbon within the particle emissions and unburnt residue (in this case, 29% by mass) significantly lowers this efficiency.

Efforts to improve the efficiency of ISB aim to reduce the amount of PM and trace pollutants released in the plume, alleviating concerns related to inhalation and environmental exposure, as well as reducing the unburnt oil residues (Buist et al., 2013). Methods of improving ISB efficiency include reducing the loss of heat from the burning oil layer to the water below and increasing oxygen penetration and heat transfer to the burning oil plume. Previous research has investigated and measured the effect of oil burn conditions and combustion enhancements to reduce emissions (Rangwala et al., 2017).

The oil consumption and PM production during ISB has been related to pool size and slick thickness. Koseki and Mulholland (1991), showed that smoke yield (mass of PM/mass of oil burned) varied four-fold, from 0.05 to 0.2 kg smoke particles/kg oil burned. Smoke yield decreased with oil pool size from 3 m to 0.6 m in diameter due to the ability of the convective, rising plume to entrain combustion air into its core, enhancing combustion. Mechanical aeration to supply air into the oxygen-limited core of the plume has been explored through use of fire whirls (Franken et al., 1992) and air injection (Nordvik et al., 1995). Much of this work is summarized in Buist et al. (2013). The most successful approach was the use of compressed air nozzles injecting air into the flame which increased the oil slick burn rate from 3.5 mm/min to 5 mm/min (Nordvik et al., 1995). Heavy industrial blowers were used to provide up to 150% of the stoichiometric air requirement, passing across the oil surface. This reduced smoke yield, increasing the burn rate and flame temperature, but still resulted in an oxygen-starved core. These observations suggest that alteration of the boom geometry to affect oil thickness and plume dimensions as well as pneumatically-induced aeration may improve combustion efficiency by aiding air penetration of the plume and turbulent mixing resulting in fewer emissions.

In the work reported here, a linear boom geometry was tested for its ability to enhance combustion presumably by shortening the oil plume core thickness to allow greater air penetration and radiant heat feedback to the burning oil. This geometry, as well as addition of nozzle-supplied compressed air at the boom/oil interface, formed the basis of this technology test. Three boom geometries and air-assist nozzles were tested in an outdoor water tank equipped with a wave system. The presence or absence of waves were tested for their effect on the ISB. Oil consumption, combustion efficiency, residue characterization, and emissions were measured. Results were assessed through comparison of emission factors, or the mass amount of a pollutant per amount of oil burned as well as the residue constituent per mass of oil burned. These tests will be useful in optimizing performance of ISB technologies to remediate potential future at-sea spills. Such information can inform decision-making by Unified Command and/or Regional Response Teams during emergency response ISB operations.

MATERIALS AND METHODS

Test Location and Test Set-Up

Testing took place at the U.S. Army Corps of Engineers, Cold Regions Research and Engineering Laboratory (CRREL) located in Hanover, New Hampshire. The facility used for the testing was the CRREL above-ground tank shown (Figure 1). The interior dimensions of the wave tank are 14.3 m x 2.4 m x 2.4 m. The beach end is tapered and the operating water level is 2 m for an operating volume of roughly 63 m³. To simulate at-sea, ISB, 35 L of crude oil (Alaska North Slope, ANS) was floated on the surface of the water-filled tank and ignited (Figure 2). Tests were conducted from November 5–9, 2018 during mostly sunny days with temperatures ranging from 2°C to 15°C. Non-saline, fresh water was used for the tests.

Figure 1.

Overhead schematic of burn tank in 9:1 boom ratio configuration.

Figure 2.

U.S. Army Corps of Engineers wave tank. Crane-supported sampling apparatus (left) and variable geometry boom in wave tank showing 9:1 configuration (right).

The mass of added oil and recovered oil residue were measured to allow burn removal efficiency to be calculated. Burns were videotaped and timed to allow estimates of burn rate and smoke quality. The burn plume was sampled using an instrument system (Aurell and Gullett, 2013) suspended from a crane and maneuvered into the plume. Slight movements of the crane boom angle, rotational boom position, and cable length kept the instruments in the plume during wind shifts. The sampling system was mounted on an aluminum skid as shown in Figure 3.

Figure 3.

Sampling platform and samplers. A SUMMA canister is flanked by two large, gray filters after sample collection.

Different boom configurations and nozzle locations were tested, with replicates of one test configuration (9:1 boom ratio, waves, 45° air injection), resulting in a total of sixteen burns. Equal-area (3.3–3.4 m²) linear boom configurations were constructed with length to width boom ratios of 1:1 to 4:1 to 9:1 (Figure 1). The boom was a modified form of the commercially-available DESMI PyroBoom (Norresundby, Denmark) with additional structural components to maintain rectangular configurations and canted compress air nozzles. Twelve air-assist nozzles were affixed to the boom and angled directly across the oil slick (90°), angled up over the oil slick (+45°), and angled down toward the oil (−45°). High pressure air (123–205 scfm) was supplied to the nozzles at a rate (100 g air/s) about ten times the anticipated burn rate requirement (10 g air/g oil burned). In addition, the presence or absence of waves (12 cm amplitude, 1.5 s period) was tested. Photographs of the tank and typical burns are shown in Figure 2. Further details on the tank, boom, and test setup are available in (Lamie et al., 2019).

Emission sampling instruments were placed on a wire mesh, aluminum frame skid, about 1.5 m long x 1 m deep x 0.5 m high (Figure 3). An extension boom crane was used to maneuver the instrument skid into the plume while adjusting for wind shifts (Figure 2). Positioning of the skid in the plume was aided by radio and hand-signal communication between ground personnel and the crane operator. A telemetry system on the emission instruments provided real time CO₂ concentrations to personnel on the ground indicating whether the sampler was within the combustion plume or not.

Targeted emissions and their sampling methods are listed in Table 1. Extensive details on instrumentation and methods are covered elsewhere (Gullett et al., 2017); brief descriptions are provided herein. All of the sampling instruments were aboard the skid with the exception of a ground-based measurements of PM concentration (DustTrak), size distribution (Electrical Low Pressure Impactor, ELPI), and black carbon (Microaethalometer, MA350), all of which were supplied by an extractive sampling probe located on the skid and a sample line to the ground.

Table 1.

Oil Burn Emission and Residue Target Compounds

Analyte	Measurement/Instrumentation	Frequency	Method Reference
CO2	LiCOR-820, NDIR	Continuous	U.S. EPA Method 3A calibration (2017b)
CO	Electrochemical cell	Continuous	U.S. EPA Method 3A calibration (2017b)
PM2.5	Impactor/filter/gravimetric	Batch	40 CFR Part 50, Appendix L (1987)
PM1, PM2.5, PM10	TSI DustTrak DRX	Continuous	Factory calibration to respirable fraction.
PM size distribution	Electrical Low-Pressure Impactor	Continuous	Factory calibration of aerodynamic size resolved aerosol charge concentration
PM size distribution	TSI Engine Exhaust Particle Sizer, DustTrak		Laser optical, factory calibration
PCDD/PCDF	Quartz filter/PUF, HRMS	Batch	Modified U.S. EPA Method TO-9A (1999b)
PAHs	Quartz filter/PUF/XAD/PUF, HRMS	Batch	Modified U.S. EPA Method TO-9A (1999b)
VOCs	SUMMA cannister	Batch	U.S. EPA Method TO-15 (1999c) and modified U.S. EPA Method 25C (CO, CO2, CH4) (2017a)
BC, AAE	Aethlabs Microaethalometer, MA350	Continuous	Filter-based light attenuation, factory calibration
OC/EC/TC	Quartz filter	Batch	Modified NIOSH Method 5040 as described in Khan et al. (Khan et al., 2012)
Carbonyls	DNPH cartridges	Batch	U.S. EPA Method TO-11A (1999a)
Oil and residue metals	Inductively Coupled Plasma (ICP)/Optical Emission Spectroscopy (OES) and Mass spectroscopy (MS)	Batch	EPA 6010 (2000) ICP, U.S. EPA Method 6020 (1994) ICPMS, Prep., U.S. EPA Method EPA 3050 (1996)
Residue TPH	GC/FID	Batch, throughout project	U.S. EPA Method 8015D (2003)
Oil and Residue BTEX	GC/MS	Batch	U.S. EPA Method 524.3 Modified (2009)
Oil and Residue Alkanes and PAHs	GC/MS	Batch	U.S. EPA Method 8270D Modified (2014)

Calculations

Emission factor values were calculated using the carbon balance method. This method uses the ratio of the sampled pollutant mass (corrected for background concentrations) to the sampled carbon mass. The latter is determined from continuous gas phase CO + CO₂ measurements and the carbon measurements of the batch-collected particles. Minor carbon mass contributions from PAHs and trace volatiles have negligible effect on the emission factor determinations. The sampled carbon mass is related to the fuel using the carbon percentage of the fuel (85%) to determine an emission factor expressed as mass of pollutant per mass of oil combusted by pyrolytic or oxidative reaction (pollutant mass/kg oil_c). To account for oil that was not burned, this emission factor is multiplied by the fraction of initial oil present that was consumed through combustion to obtain an emission factor based on the initial oil amount (pollutant mass/kg oil_i). With less than complete oil consumption, this will decrease the value of the emission factor but is not reflective of the oil mass loss efficiency. In most cases we have presented these emission factors in terms of the initial oil mass as the oil consumption is condition-dependent and future scenarios and projections will have more sense of the spill amount than the efficiency of mass loss.

Results are evaluated in terms of modified combustion efficiency (MCE), both gas phase molar ratio (CO₂/(CO₂+CO), MCE_g) and gas plus particles carbon mass ratio (CO₂/(CO₂+CO+TC), MCE_T). The latter incorporates measurements of total particulate carbon (TC) as well as gas phase carbon species of CO and CO₂ and is most relevant for combustion comparisons with particle emissions. MCE_g was evaluated continuously throughout the burn and, while a commonly cited measurement of combustion efficiency, does not account for emissions of uncombusted carbon in particle form. MCE_T was based on two particle samples taken in series during each burn that were combined with the CO and CO₂ measurements during the same time period. Due to the dissimilar time periods of sample collection MCE_T is not always less than MCE_g despite the constant presence of particulate carbon.

Residues were characterized by comparing the initial oil concentrations with the collected residue to determine the percentage decrease in concentration. The total mass loss percentage of the residue components was calculated with the mass and concentration of the initial oil and residual oil.

Measurement Methods

CO₂ is continuously measured using a non-dispersive infrared (NDIR) instrument (LI-820 model, LI-COR Biosciences, Lincoln, USA. The LI-820 was calibrated on a daily basis in accordance with U.S. EPA Method 3A (2017b) using National Institute of Standards and Technology (NIST) certified standards. An electrochemical sensor (e2V EC4–500-CO, SGX Sensortech, Essex, United Kingdom) measured CO concentration through reactive oxidation and the changing impedance. Daily calibrations using U.S. EPA Method 3A (2017b) from 0 to 100 ppm prior to measurements resulted in a ± 2 ppm error. The CO₂ and CO concentrations are recorded with an onboard computer using a LabView generated data acquisition and control program. More details are available elsewhere (Gullett et al., 2017).

VOCs were collected for analysis using 6-L SUMMA canisters via U.S. EPA Method TO-15 (1999c). The canisters were fit with a manual valve, metal filter (frit), pressure gauge, pressure transducer, and an electronic solenoid valve. The VOC analysis also included CO, CO₂, and methane (CH₄) analyses via U.S. EPA Method 25C (2017a) allowing for determination of volatile emission factors.

PCDD/PCDF were sampled onto a quartz microfiber filter (20.3×25.4 cm) followed by a polyurethane foam (PUF) plug with a low voltage, Windjammer brushless direct current blower (AMETEK Inc., Berwyn, USA) with a nominal sampling rate of 0.85 m³/min following EPA Method TO-9A (1999b). The filter and PUF were extracted using methylene chloride and toluene, cleaned, and analyzed using an isotope dilution method based on U.S. EPA Method 23 (1991). Concentrations were determined using high resolution gas chromatography/high resolution mass spectrometry (HRGC/HRMS) with a Hewlett-Packard gas chromatograph 6890 Series coupled to a Micromass Premier mass spectrometer (Waters Corp., Milford, MA, USA). PCDD/PCDF were analyzed using U.S. EPA Method 8290A (2007) with standards (CIL Cambridge Isotope Laboratories Inc., U.S.A.). containing tetra- to octa-PCDD/PCDF native and ¹³C-labeled congeners designed for modified U.S. EPA Method 23 (1991). Results are expressed with any non-detectable toxic congeners (n = 17) set to 0 (ND = 0) and at the limit of detection (LOD).

PAHs were sampled using a PUF/XAD-2/PUF sorbent preceded by a quartz microfiber filter with sampling rate of 0.005 m³/min (Leland Legacy pump (SKC Inc., Eighty Four, PA, USA)). The Leland Legacy/Gilian 5000 sample pump was calibrated with a Gilibrator Air Flow Calibration System (Sensidyne LP, St. Petersburg FL, USA). Target compounds were analyzed using a modified U.S. EPA Method 8270D (2014). Modifications to the method include use of a pre-sampling, pre-extraction, and pre-analysis spike with labeled standards.

PM_2.5 was sampled with a Leland Legacy pump onto Teflon filters through an impactor (SKC Inc., Eighty Four, PA, USA) at 10 L/min and measured gravimetrically following procedures described in 40 CFR Part 50 Appendix L (1987).

Particle organic/elemental/total carbon (OC/EC/TC) was sampled with an SKC PM_2.5 impactor using 37 mm quartz filter via a Leland Legacy sample pump (SKC Inc., USA) with a constant airflow of 3 L/min. The OC/EC/TC was analyzed via a modified thermal-optical analysis (TOA) using NIOSH Method 5040 (1999d) and Khan et al. (2012).

Carbonyls were sampled at 1 L/min into 2,4-dinitrophenylhydrazine (DNPH) coated silica cartridges (Sigma-Aldrich (PN 505323)) as per U.S. EPA Method TO-11A (1999a). DNPH cartridges were extracted and analyzed by high-pressure liquid chromatography (HPLC) on an Agilent 1100 HPLC with a Diode Array Detector as per U.S. EPA Method TO-11A (1999a).

Black carbon (BC), PM concentration, and PM size distributions were measured from a diluted sample line. A two-stage dilution process, starting with the skid-mounted porous dilution probe (Gullett et al., 2017) and ejector diluter (DI-1000, Dekati Ltd., Kangasala, Finland), used N₂ as the diluent. The sample was conveyed to the ground-based through a 30-m long, 1.03 cm inner diameter carbon-impregnated Teflon hose wrapped in a stainless steel braid (Detroit Flex Defense, Troy, MI). The dilution was calculated by the CO₂ concentration at the inlet and outlet and varied from 30:1 to 73:1 but was constant throughout each burn. Aerosol particle losses in the sampling system were calculated using the United Technologies Research Center particle loss calculator (NASA, 2011).

BC was also measured with an MA350 microaethalometer (AethLabs, San Francisco, CA) from the diluted sample line, measuring light attenuation at five wavelengths. The variation of attenuation with wavelength can be used to calculate the absorption angstrom exponent (AAE), which can be used to quantify brown carbon and distinguish between combustion sources.

Continuous PM size fractions (PM₁, PM_2.5, PM₁₀) were sampled with DustTrak DRX (TSI, Inc., Shoreview, MN) on a diluted sample line, measuring light scattering by aerosols as they intercept a laser diode. An enclosed, 37-mm pre-weighed filter cassette provided a simultaneous total suspended particle (TSP) gravimetric sample for determining sample-specific, custom photometric calibration factor (PCF). Particle size distributions were also measured continuously with an electrical low pressure impactor (ELPI, Dekati, Kangasala, Finland) over 13 size stages at a constant airflow of 10 L/min. The ELPI provides a continuous measurement of aerosol size from 0.007 – 10 μm.

The post-burn residue solids were collected with pre-weighed oil absorbent pads (New PIG Corp., Tipton, PA, USA) to determine oil mass loss. The bottom of the tank was oil-free; minimal amounts of oil adhered to the stainless steel spheres on the boom. Pads and tank samples were analyzed for TPH using gas chromatography (GC) with flame ionization detection (FID) using modified U.S. EPA Method 8015D (2003). Concentrations of TPH were calculated using a six-point calibration curve generated with the ANS, unweathered oil and expressed as mass TPH per mass residue. GC/MS analyses were conducted for monoaromatic hydrocarbons (i.e., benzene, toluene, ethylbenzene and xylene; “BTEX”, using modified U.S. EPA Method 524.3 (2009), alkanes and PAHs (modified U.S. EPA Method 8270D (2014)). Alkanes (C10-C35 normal aliphatics, and branched alkanes [pristine and phytane]) and PAHs (including 2–4 ring compounds and their alkylated homologs (i.e., C0-C4 naphthalenes, C0-C4 phenanthrenes, C0-C3 fluorenes, C0-C4 dibenzothiophenes, C0-C4 napthobenzothiophenes, C0-C4 pyrenes and C0-C4 chrysenes) were summed to compute total alkane and PAH concentrations, respectively.

RESULTS AND DISCUSSION

Emission Factors

The 16 test conditions, resultant oil mass loss, combustion efficiencies, and primary emission factors (normalized to the initial oil mass, kg oil_i) are shown in Table 2 and a typical CO, CO₂, and MCE_g trace is shown in Figure 4. Environmental conditions included temperatures ranging from 2 to 15 °C, water temperatures from 6 to 12 °C, and wind speeds of 0 to 1.7 m/s. Burn durations ranged from 250 to 453 s and were determined by visual observation of flame extinction. One condition, “Control”, was replicated to provide an estimate of precision. The number of batch samples collected for each test configuration is shown in Supplemental Information (SI) Table S-1. The three Control tests resulted in a mean oil mass loss of 97.1% (relative standard deviation (RSD) 2.7%) and an MCE_T of 0.882 (RSD 3.3%). The PM_2.5 emission factors for the three replicate Control runs had an RSD of 8%, indicating the reproducibility of the tests and measurements.

Table 2.

Run Conditions and Primary Emission Results.

Burn No.	Test Condition	Boom Ratio	Nozzle Orientation	Mass oili (kg)	Mass loss (%)	MCEg	MCET	CO2 (g/kg oili)	CO (g/kg oili)	CH4* (g/kg oili)	PAH16 (mg/kg oili)	Benzene (mg/kg oili)
1	Control	1:1	None	31.0720	94.3	0.927	0.852	2132	27.8	2.0	271.13	787.08
2	Air	1:1	90°	31.0715	95.5	0.932	0.862	1846	18.2	0.9	114.14	357.22
3	Waves	1:1	None	31.1195	91.5	0.873	0.855	2258	29.8	1.6	388.36	806.32
4	Air and Waves	1:1	90°	31.0688	94.4	0.961	0.863	2540	28.2	0.6	97.22	589.51
5	Air and Waves	1:1	+45°	32.5236	95.0	0.941	0.881	2056	19.5	0.7	176.86	559.87
6	Control	9:1	None	31.0659	99.6	0.888	0.911	2666	38.7	2.0	438.76	1404.74
7	Air	9:1	+45°	31.1188	98.7	0.919	0.907	2745	31.3	0.8	341.35	590.19
8	Waves	9:1	None	31.0647	94.9	0.935	0.917	2465	31.1	0.7	401.20	27.12
9	Air and Waves	9:1	+45	31.0670	89.2	0.926	0.932	2519	23.9	0.9	311.16	565.20
10	Control	4:1	None	31.0710	97.1	0.904	0.882	2559	10.9	1.7	212.61	1060.51
11	Air	4:1	+45°	31.0679	96.4	0.927	0.893	2558	25.6	0.6	171.65	510.85
12	Waves	4:1	None	31.1109	89.2	0.924	0.877	2260	27.3	1.0	187.84	774.08
13	Air and Waves	4:1	+45°	31.1118	90.5	0.963	0.903	2312	22.1	0.6	194.60	620.70
14	Air and Waves	4:1	−45°	31.0640	88.8	0.976	0.881	2231	21.6	NS	116.43	NS
15	Air and Waves	9:1	+45°	31.0620	89.5	0.972	0.92	2358	22.5	0.7	244.02	462.75
16	Air and Waves	9:1	+45°	31.0850	91.8	0.972	0.922	2566	27.2	0.7	157.24	466.51
	Average				93.5	0.934	0.891	2380	25.4	1.0	239.04	638.84
	SD				3.5	0.030	0.026	241	6.4	0.5	108.42	313.97
	RSD (%)				3.7	3.2	2.9	10.1	25.3	51.6	45.4	49.1

NS = not sampled.

^*CH₄ emission factor derived from SUMMA canister. Data calculated from the entire burn.

Figure 4.

Representative burn profile of CO, CO₂, and MCE_g.

The 16-test matrix yielded oil mass loss from 88% to 99% (Table 2), averaging 93.5%, consistent with literature values above 90% (Fingas et al., 1994; Fingas et al., 1996; Canada, 1997; Potter and Buist, 2008) MCE_g ranged from 0.873 to 0.976 (0.934 mean) which can be compared with values of 0.978 (Gullett et al., 2017) for pan burning of Bayou Sweet Crude oil, 0.94 with Prudhoe crude oil (Evans et al., 2001), and 0.975 during the Deepwater Horizon disaster based on maximum observed CO and CO₂ values (Perring et al., 2011).

The 16-test PM_2.5 emission factors (Table 3) averaged 91.2 g/kg oil_i, RSD 28.2% (97.4 g/kg oil_c, RSD 27.6%, see Table S-11), or 9.74% of the original oil mass was emitted as PM_2.5. This value is higher than measured (70 g/kg oil_c, RSD 12%) with pan burns (Gullett et al., 2017) and during the Deepwater Horizon burns (75 g/kg oil_c, RSD 29%, (Aurell and Gullett, 2010)) with the same equipment and methods. Our values were on the low end of the range reported by Walton et al. (1993) of 94 to 152 g PM/kg oil and much lower than determined (148 to 152 g PM/kg oil) during the Newfoundland Offshore Burn Experiment (NOBE) tests (Canada, 1997).

Table 3.

PM-related Emission Factors.

Burn No.	Test Condition	Configuration		MCET*	PM2.5	OC	EC	TC
		Boom Ratio	Nozzle Location		g/kg oil initial
1	Control	1:1	No nozzle	0.840	145.2	33.9	103.0	138.0
2	Air	1:1	90°	0.860	118.2	24.7	107.1	119.7
3	Waves	1:1	No nozzle	0.851	122.4	24.1	112.8	133.5
4	Air and Waves	1:1	90°	0.861	102.9	71.2	57.3	122.0
5	Air and Waves	1:1	+45°	0.894	99.0	37.3	74.0	98.2
6	Control	9:1	No nozzle	0.917	101.3	12.4	52.5	58.7
7	Air	9:1	+45°	0.906	74.4	8.7	64.4	65.6
8	Waves	9:1	No nozzle	0.919	67.9	7.0	54.7	57.7
9	Air and Waves	9:1	+45°	0.935	58.9	6.0	47.6	51.3
10	Control	4:1	No nozzle	0.895	102.0	17.2	83.9	94.5
11	Air	4:1	+45°	0.900	92.0	13.0	77.9	87.2
12	Waves	4:1	No nozzle	0.878	102.9	25.1	88.8	107.7
13	Air and Waves	4:1	+45°	0.902	76.0	7.6	78.8	80.4
14	Air and Waves	4:1	−45°	0.879	86.8	16.0	97.5	111.9
15	Air and Waves	9:1	+45°	0.922	57.8	9.0	58.2	65.7
16	Air and Waves	9:1	+45°	0.924	52.1	8.7	53.3	60.1
				Avg.	91.2	20.1	75.7	90.8
			Stand. Dev.		25.7	16.7	21.4	29.1
			RSD (%)		28.2	83.3	28.3	32.0

^*MCE_T calculated only when PM samples were being collected. MCE_T, PM_2.5, OC, EC, and TC values

The concentration normalized size distributions for varying boom ratios are shown in Figure 5. There was minimal change in the size distribution across the different burn scenarios. The number concentration was dominated by the smallest particles (< 100 nm), but only particles smaller than 2.5 μm were measured due to the losses of larger particles in the sampling lines. Oil burns emit particles with a bimodal distribution, with a second mode larger than 2.5 μm contributing as much as 30% of the PM mass (Evans et al., 2001; Gullett et al., 2017) However, the number concentration of these larger particles is likely negligible and would not impact the number concentrations or size distributions measured here.

Figure 5.

Particle number distribution for control burns of varying boom ratio.

The average particle number, PN, and diameter are 1.83×10¹⁶ #/kg oil_i and 106 ± 24 nm, respectively (Table 4). This PN emission factor is nearly an order of magnitude larger than previously observed for ISB (3.5 – 5.5×10¹⁵ #/kg oil, (Ross et al., 1996; Middlebrook et al., 2012; Gullett et al., 2017), but direct comparison across studies is hampered by the differing measurement methods, which do not include all size fractions. The total PN emission factor was approximately constant with MCE_T, while the geometric standard deviation (σ_g), and median diameter (d_g) showed a slight decreasing trend with increasing MCE_T, i.e., smaller median size, and a more narrow distribution were emitted as the MCE_T increased (Table S-8). However, the correlations with MCE_T were low (PN emission factor R² = 0.089, d_g R² = 0.075, σ_g R² = 0.19, Figure S-5).

Table 4.

Particle characteristics.

							BC	AAE
Burn No.	Test Condition	Configuration		PN	Dg	σγ	(880 nm)
		Boom Ratio	Nozzle Location	#/kg oil initial	nm	nm	g/kg oil initial	unitless
1	Control	1:1	No nozzle	2.08□1016	89	3.45	155.7	0.57
2	Air	1:1	90°	1.26□1016	135	3.01	110.4	0.37
3	Waves	1:1	No nozzle	2.03□1016	132	2.90	117.7	0.44
4	Air and Waves	1:1	90°
5	Air and Waves	1:1	+45°	1.503□1016	126	2.74	97.6	0.56
6	Control	9:1	No nozzle	2.02□1016	112	2.74	67.4	0.78
7	Air	9:1	+45°	2.06□1016	115	2.79	65.4	0.69
8	Waves	9:1	No nozzle	1.62□1016	145	2.61	59.4	0.83
9	Air and Waves	9:1	+45°	1.43□1016	126	2.80	41.0	0.79
10	Control	4:1	No nozzle	2.50□1016	78	2.90	68.2	0.63
11	Air	4:1	+45°	2.60□1016	81	2.93	60.5	0.59
12	Waves	4:1	No nozzle	1.81□1016	92	3.05	49.0	0.67
13	Air and Waves	4:1	+45°	2.08□1016	90	3.00	47.6	0.65
14	Air and Waves	4:1	−45°	1.19□1016	111	3.57	36.6	0.42
15	Air and Waves	9:1	+45°	2.14□1016	74	3.24	42.1	0.85
16	Air and Waves	9:1	+45°	1.71□1016	74	3.22	32.8	0.79
			Average	1.87×1016	105	3.00	70.1	0.64
			SD	4.74×1015	24	0.27	35.2	0.15
			RSD (%)	22.3	22.9	9.1	50.2	23.7

The continuous PM_2.5 concentration measured by the DustTrak was strongly correlated with the filter based PM_2.5 (R² = 0.91). The BC values (full data are shown in Table S-13) were also moderately correlated with the EC concentrations (R² = 0.48). The poorer correlation may indicate varying optical properties as BC is an optical-based measurement and EC is a thermal-based measurement. The AAE calculated from the multiwavelength BC data can be used to distinguish among combustion sources (Sandradewi et al., 2008). The AAE measured here was less than one, which has been observed before from crude oil burns (Gullett et al., 2017), indicating that the particles emitted from oil burning are aggregates of larger spherules, which have been associated with larger diameter pool fires (Evans et al., 2001). The AAE increases with MCE_T (Figure S-6, R² = 0.63), suggesting that particle morphology is also impacted by combustion efficiency. This change in AAE trends with the MCE throughout the burn as shown in Figure 6, with the most efficient combustion occuring at the very beginning of the burn and dropping off as the burn progresses.

Figure 6.

A typical PM_2.5, black carbon (BC), and number concentration trace (data frequency 1 second) with corresponding MCE (10 seconds moving average) and absorption angstrom exponent (AAE) during air and waves 9:1 +45°, Burn No. 16., MCE calculated using CO₂ and C.

Analysis of collected particles (see Table S-12 for burn data) show that the TC average was 90.8 g/kg oil_i, (RSD 32.0%) and 84.8 g/kg oil_c (RSD 31.8%) of which EC was 75.7 g/kg oil_i (RSD 28.3%) and OC was 20.1 g/kg oil_i (RSD 83.3%). Both the EC/TC ratio and EC/PM ratios averaged 0.83 which compare to 0.82 and 0.77, respectively, from the Deepwater Horizon ISB (Gullett et al., 2016). The closeness of these values, despite different oil types and burn conditions, indicate that the 16-tests reported here were a reasonable simulation of at-sea burns for emitted particle characteristics.

MCE_T, calculated from CO, CO₂, and TC, ranged from 0.852 to 0.932 (0.891 mean), with the generally lower values, compared to MCE_g,_, accounting for the inclusion of particulate carbon (Table 2). The CO emission factor, 25.4 g/kg oil_i was about 1% of the emission factor and the CH₄ was about 3% of the CO value (measured from the SUMMA can analysis).

The emission factor for PAH₁₆ (mean 239.04 mg/kg oil_i) reflects both gas phase and particle bound PAHs (Table 2). This value is lower than that found in pan burning of the Bayou Sweet Crude (697 mg/kg oil_i, (Gullett et al., 2017). The PAH₁₆ emission factor in this current work is 50 times higher than found for particle-only PAH analysis of collected solids from the Deepwater Horizon ISB (Gullett et al., 2016). This suggests that the majority of PAHs from oil burns are volatile and/or that extensive loss of volatile PAHs (blow off) can occur during extended duration sampling of particles (Gullett et al., 2016). The proportion of volatile to total PAHs (including particle phase PAHs) sampled was over 3/1 during at sea oil burn simulations (Gullett et al., 2017). In this current work the dominant PAH was naphthalene, 46% of the total, followed by acenaphthylene and phenanthrene at 14% and 10%, respectively, consistent with previous work (Gullett et al., 2017). This PAH emission factor value is more than twice the value from open burning of domestic waste (Aurell et al., 2012). The 16-run PAH toxic equivalency (TEQ) based upon weighting factors relative to B[a]P_eq (Larsen and Larsen, 1998) was 7.9 mg TEQ/kg oil_i (Table S-3), lower than found previously (20.6 mg TEQ/kg oil_i) for oil pan burns (Gullett et al., 2017). For comparison, values for open air burning of JP-5 fuel were 1.2 mg TEQ/kg fuel combusted (Aurell et al., 2017) where the difference in the denominators of the two values (fuel initial versus fuel combusted) would be expected to result in less than a 5% difference. A complete list of the PAH emission factors is included in SI, Tables S-2 and S-3.

VOC and carbonyl analyses (Tables S-4, S-5 and S-6) showed that benzene was the dominant VOC averaging over 600 mg/kg oil_i compared with previous studies at 1,120 mg/kg oil_i (Gullett et al., 2017). The next most prevalent VOCs in order were toluene, propene, styrene, and acrolein.

Seven PCDD/PCDF samples were taken (Table S-7) and emission factors are reported using TEQs from Van den Berg et al. (2006). In most cases these were composite samples from multiple, same-condition tests to better ensure against the possibility of non-detects for the 2,3,7,8-Cl-substituted congeners. PCDD/PCDF emission factors ranged from 0.20 to 3.1 ng TEQ/kg oil_c, bracketing the value found at sea during the Deepwater Horizon ISB of 2.0 ng TEQ/kg oil_c (Aurell and Gullett, 2010). Since this current study was done with freshwater, the similarity of the PCDD/PCDF emission factors suggests that chlorine from seawater is not necessary for PCDD/PCDF formation, with the chlorine possibly originating from trace chlorine in the oil. These PCDD/PCDF emission factors overlap the range found for field grass and forest burns (0.8 – 1.9 ng TEQ/kg) despite significant differences in fuel types and a 4- to 7-fold higher PM_2.5 emission factor with the oil burns (Aurell et al., 2015).

Residues

Oil residues were characteracterized for BTEX, alkanes, PAHs, and TPH and compared with the initial oil concentrations (Tables S-14 to S-17). All four measures resulted in reduced concentrations. The decrease in concentration of the highly volatile BTEX compounds averaged 98.9% (Table S-17) from the raw oil to the residues. These values for the alkanes, PAHs, and TPH were 60.3, 50.4, and 69.1%, respectively. The lighter, lower-boiling-point compounds were preferentially removed at a higher rate than the heavier, higher-boiling-point compounds. In comparison to the source ANS crude oil, alkanes with 10 to 18 carbons (denoted as nC10–18) decreased in concentration by 83% on average post-burn, while nC19–27 decreased by 19.8% and nC28–35 increased by 0.1%. Similarly, 2-ring PAHs decreased by 73.8% on average post-burn, while the 3-ring PAHs decreased by 29.7% and the 4-ring PAHs increased by 2.6%.

While TPH analysis cannot differentiate individual hydrocarbon compounds, it can comprehensively measure resolvable volatiles, semi-volatiles, and unresolvable complex mixture (Morales-McDevitt et al., 2020). The percent concentration loss of TPH ranged from 47.1% to 92.5% with an average of 69% (the TPH RSD for the replicate runs was 7% which indicates good reproducibility of the conducted test condition). The percent loss of hydrocarbon concentration in the post-burn residue generally agree with previous studies (84.9% (Buist, 1997), 84.2% (Ross et al., 1996), 90% (Potter and Buist, 2008), 94% (Fingas et al., 1996), 84–99% (Fingas, 2010), 95–98% (Day et al., 1979), 98–99% (Walton et al., 1993) and 96–98%, (Evans et al., 2001)). When the residual oil mass for each burn was applied to the residue concentrations, 100% of the BTEX compounds were lost (within rounding error). Over 96% of the alkane, PAHs, and TPH mass in the original oil was lost during the ISB.

Effect of Test Conditions

To discern the influence of the test parameters (Air, Waves, and Boom Ratio) on emission and residue metrics, stepwise linear regressions were analyzed to examine the effect upon four combustion efficiency metrics: oil mass loss, MCE_T values, PM_2.5, and TPH. These analyses are summarized in Table 5 and graphically shown in Fig. 7. Statistically significant effects are noted for the presence of Waves which decreased the oil mass loss and the amount of BTEX, alkanes, PAHs, and TPH consumed. Increases in Boom Ratio decrease the PM_2.5 emission factor and improve the MCE_T combustion efficiency. No significant effect of Air in our test configurations was noted on any of the four combustion efficiency metrics. Pairwise correlations for the non-categorical, principal dependent parameters are examined in SI and Table S-18. Understanding the interplay of combustion efficiency metrics is helpful for minimizing burn residues and emission particulate formation, assessing environmental tradeoffs (Fingas, 2017), and allocation of cleanup resources by the Unified Command and Regional Response Teams (NRT, 1995, NRT, 1997). The higher Boom Ratio increases the flame front area exposed to ambient air, reducing the PM_2.5 emission factor and increasing the burn times. The burn times for all six tests with a Boom Ratio of 9:1 exceeded the burn durations of the ten 1:1 and 4:1 tests. Analysis of the linear boom frontal circumference (7.3 m, 9.1 m, and 11.1 m for the respective Boom Ratios of 1:1, 4:1, and 9:1) shows a slightly greater correlation (R² = 0.69) with PM_2.5 than Boom Ratio (R² = 0.68). Both analyses confirm that improved combustion quality is directly related to the area of flame/air exposure, consistent with diffusion flame rate limitations. The higher exposure of the flame front to ambient air increases the availability of oxygen by reducing the distance through which oxygen would have to diffuse or convect into the fuel-rich zone above the volatilizing oil pool. The extent of combustion would then be reflected by the flame/air interfacial surface and the distance (or volume) through which oxygen would have to be transported through the fuel-rich zone. In these tests with equal oil area configurations and assuming a constant height of the fuel-rich volatiles, the volatile plumes’ surface area/volume ratios are proportional to the boom circumference length. Tests to understand the effect of plume surface area/volume ratios on combustion quality would verify this concept. This enhanced oxidation reduces partially oxidized C (CO) or incompletely oxidized C (particulate carbon), resulting in higher MCE_T. The combined influence of and PM_2.5 is shown in Fig. 8 for all of the tests, indicating the strong relationship between higher combustion efficiency and less particulate emissions.

Table 5.

Summary of Test Parameters on ISB Efficiency.

Metric	Effect of Parameter on ISB Efficiency Metric
	Waves	Boom Ratio	Air Injection
Mass Loss	Decrease	No Significant Effect	No Significant Effect
	C = −2.7, (R2 = 0.60, F = 20.9, Prob>|t| = 0.0004	C = +0.02 (R2 = 0.00, F = 0.0, Prob>|t| = 0.94	C = +0.7, (R2 = 0.04, F = 0.6, Prob>|t| = 0.44
PM2.5	No Significant Effect	Decrease	No Significant Effect
	C = +11.4, R2 = 0.20, F = 3.4, Prob>|t| = 0.08	C = −5.9, R2 = 0.63, F = 24.5, Prob>|t| = 0.0002	C = +12.6, R2 = 0.24, F = 4.4, Prob>|t| = 0.05
MCET	No Significant Effect	Increase	No Significant Effect
	C = −0.01, R2 = 0.04, F = 0.6, Prob>|t| = 0.4	C = +0.01, R2 = 0.86, F = 85.1, Prob>|t| = <0.0001	C = −0.01, R2 = 0.08, F = 1.1, Prob>|t| = 0.3
TPH concentration reduction	Decrease	No Significant Effect	No Significant Effect
	C = −0.09, R2 = 0.60, F = 21.2, Prob>|t| = 0.0004	C = +0.01, R2 = 0.10, F = 1.6, Prob>|t| = 0.22	C = −0.04, R2 = 0.12, F = 1.9, Prob>|t| = 0.18
TPH mass loss	Decrease	No Significant Effect	No Significant Effect
	C = −0.32, R2 = 0.58, F = 19.4, Prob>|t| = 0.0006	C = −0.09, R2 = 0.03, F = 0.4, Prob>|t| = 0.54	C = 0.52, R2 = 0.08, F = 01.2, Prob>|t| = 0.28

Figure 7.

Boom Aspect Ratio versus a) modified combustion efficiency (MCE_T), b) PM_2.5 emission factors, c) oil mass loss percentage, and d) total petroleum hydrocarbon concentration (TPH).

Figure 8.

The effect of MCE_T on PM_2.5 emission factors for all of the test runs (two particle samples were collected in series per test run). Common colors indicate similar Boom Ratios.

The presence of waves decreases the oil mass loss, possibly due to the wave action causing the fire to extinguish faster or due to enhanced heat transfer from the flaming oil to the water below. Introduction of other test variables has an insignificant effect on explaining the variance. This is seen in Figure 7 where all Waves tests have lower oil mass loss than the Control. The lack of an Air effect indicates that the enhanced turbulence and supply of oxygen just above the oil/water interface is not the rate limiting step for oil consumption.

Oil mass loss was unrelated to any combustion emission measures, such as MCE_T and CO₂, nor to the duration of the flaming fire. This indicates that the extent of the burn, or the consumption of oil, was not a function of how well the combustion emissions were controlled. This finding presents a challenge for optimizing ISB, indicating a potential tradeoff between emissions and residues.

Previous studies have noted that waves lower the burning rate (Buist et al., 2003) due to increased heat loss to the water beneath a slick (Walavalkar and Kulkarni, 1996). As the burning efficiency and burning time decreases, the residue percentage will increase as well (Fritt-Rasmussen, 2010). This could occur if waves stretch the oil slick making the slick become thinner and more prone to heat loss (Brandvik et al., 2006), or if choppy and steep waves prevent the oil slick from igniting (Buist et al., 2003). The latter may also aid in transferring hydrocarbons from the slick to the water beneath, thus making them unavailable to burn.

The change in residue TPH concentration is an indicator of burn quality with respect to the original oil composition (Figure S-9). Higher oil mass loss is associated with lower TPH concentrations in the residue (higher reductions of the TPH concentration in the original oil) as well as lower BTEX, alkane, and PAH concentrations. The mass loss of these residue components are also negatively affected by the presence of waves. No other test parameter is statistically significant. The addition of waves may result in the oil slick oscillating lower into the water, resulting in greater heat loss to the water and less combustion efficiency. Waves may also enhance the dissolution of hydrocarbons from the slick into the water below.

Table 5 highlights the important effects found in this effort. The presence of waves decreases the oil mass loss and the reduction of the oil’s TPH concentration (as well as the TPH mass loss). Higher Boom Ratios increase the combustion efficiency, MCE_T, and reduce the PM_2.5 emissions. Air injection had no significant effect on mass loss, MCE_T, TPH, or PM_2.5.

CONCLUSIONS

Mesoscale simulations of at-sea, in situ oil burns conducted in a wave tank resulted in extensive characterization of emissions. Where historical emissions data overlapped, the wave tank results compare favorably, indicating the representativeness of these burns. Oil combustion efficiency based on plume emission measurements was improved when the interfacial area between the flame front and the ambient air was increased at higher Boom Ratios. For ISB applications, this concept would be reflected by increases in the flame front surface area (boom length) to volume ratio (spill diameter). A three-fold range in emissions related to combustion efficiency (MCE_T) was observed, indicating that improvements in combustion efficiency can make a significant difference in ISB emission levels. In these limited-number tests, measures to improve combustion efficiency (Boom Ratio, Waves, Air injection) that reduced air pollution were not statistically associated with higher oil mass consumption, presenting a potential tradeoff for ISB decisions. The presence of waves decreased the amount of oil mass lost during combustion likely due to enhanced heat transfer from the oil to the underlying water, encouraging flame extinction. These tests provide insight into operational strategies that can be applied during ISB response and monitoring efforts.