Subsea Dispersant Injection (SSDI) - Summary Findings from a Multi-Year Research and Development Industry Initiative

In 2010, the American Petroleum Institute (API) launched a Joint Industry Task Force to coordinate funding of cooperative research that would answer key questions regarding oil spill response options. The API focused on studying the effectiveness of subsea dispersants injection (SSDI) because of the novelty of the technique and the lack of prior research. SSDI is still under scientific and regulatory review despite the fact that there was substantial evidence from responders and the regulatory agencies during DWH indicating that SSDI was critical in reducing surfacing oil above the well (Nedwed et al. 2012). However, the nature of field evidence collected during an emergency response does not completely allow the scientific rigor necessarily to unambiguously demonstrate the effectiveness of SSDI. Also, because of the novelty of SSDI, there were no validated theories or models to guide SSDI application for future contingency planning. The SSDI knowledge gaps were recognized by API and the so-called D3 committee was formed to formulate and execute a comprehensive research program that would improve our fundamental understanding of SSDI effectiveness. This paper summarizes that research which provided critical data and information considering that SSDI efficacy during DWH was initially challenged (Peterson et al, 2012, Paris et al. 2012, Amman et al., 2015).

A basic conceptual understanding of what happens to oil and gas in a deepwater blow-out was summarized by the National Academy of Sciences (NAS, 2003). In its basic form there are 3 phases: an initial jet phase extending a few meters above the blowout source (or 5–10 times the release diameter) where oil droplets and gas bubbles are initially formed as the reservoir fluids exit the well; a buoyant plume phase where the individual droplets and bubbles are driven upward by their buoyancy along with entrained water; and a far-field phase where the buoyant plume reaches a terminal layer and oil (gas) rises as individual droplets (bubbles). The terminal layer is a result of the entrainment of dense deep water into the plume of oil and gas and further reduction of the plume buoyancy caused by dissolution of gas or formation of gas hydrates. In the buoyant plume there is potential for coalescence of droplets/bubbles as well as potential breakup of the larger droplets/bubbles. In the far-field phase. The smallest oil droplets will rise so slowly and disperse so widely that they may not be able to generate a continuous surface slick.

It is clear from the basic conceptual understanding described above that the gas bubbles and oil droplets act as the driving force that transports hydrocarbons towards the surface. Since the rise velocity of a buoyant droplet/bubble scales roughly as the square of its diameter, the magnitude of this buoyancy-driven transport is very sensitive to diameter. The goal of SSDI is to reduce the size of oil droplets released during a blowout in order to reduce their rise velocity and delay or mitigate oil surfacing. SSDI does this by reducing the interfacial tension (IFT) of the oil, and allowing the high turbulence near the discharge point to further reduce droplet sizes.

Reducing the droplet size also allows more rapid dissolution of the water soluble components in the oil (Brown et al, 2011) because the oil has a larger surface area to volume ratio. Increasing the surface area enhances microbial degradation (Hazen et al, 2010, Baelum et al., 2012, Brakstad et al. 2015) by providing more locations for microbial attachment. Greater dissolution and degradation increases the density of droplets (and therefore reduces the rise velocity) because the lighter compounds in the oil tend to be more soluble and biodegradable (Passow et al. 2012).

By keeping the oil in the water column longer and reducing the amount that surfaces results in surfacing of very thin oil films. These thin films (typically classified as “sheen”, “rainbow” and “metallic” according to the Bonn Agreement Oil Appearance Code (BAOAC, 2011) are too thin to entrain water and form water-in-oil emulsions and are more likely to naturally re-disperse by waves within hours. In contrast, large numbers of multiple millimeter size oil droplets surfacing over a smaller area are expected to form thick oil slicks that can rapidly emulsify and persist for weeks, increasing the likelihood of oil reaching shore. The difference in persistence between thick emulsifying (> 0.2 mm) and thin non-emulsifying (< 0.1 mm) surface oil films has been observed in several experimental subsurface oil releases and real incidents (Johansen et al. 2003, Daling et al., 2015, Rye et al., 1996, and Rye et al., 1997).

Droplets of 70 μm are believed to be sufficiently small to remain permanently entrained in the water column when surface slicks disperse (Lunel et al., 1995). Larger droplets could remain permanently entrained in the water column during a subsea release because more time for dissolution and biodegradation could lead to droplets that reach neutral buoyancy, as described above. Reducing the amount of surface oil and reducing its persistence has the important benefits of improving the safety of well-control responders; reducing potential harm to marine mammals, birds, and turtles; and reducing possible negative impacts on nearby coastal areas.

Initially, the API D3 committee planned a research program that included laboratory studies followed by a field experiment along the lines of the DeepSpill experiment performed by SINTEF at 840 m depth in Norway in 2000 (Johansen, et al. 2003). However, after considering the challenges of obtaining regulatory approval for field experiments, API decided to focus on laboratory, mesocosm, and modeling studies. The laboratory and mesocosm studies were broken into six phases, which were contracted to and led by SINTEF in Trondheim, Norway.

Phases-I and II were performed in the large Tower Basin (3 m in diameter and 6 m high) at SINTEF in Trondheim, Norway. Phase IV was performed in close cooperation with the University of Hawaii; Phases III & VI were performed at Southwest Research Institute’s (SwRI) hyperbaric facilities in San Antonio, TX, US; and Phase V was performed at the US Bureau of Safety and Environmental Enforcement’s (BSEE) OHMSETT Facility in Leonardo, New Jersey, in close cooperation with SL Ross Environmental Research, Canada.

The next section describes the laboratory setups and is followed by a section describing the major results. By necessity, this section is largely restricted to a summary of the major findings because there were simply too many experiments and details to include in a single paper. The interested reader is referred to the detailed technical reports and scientific papers, identified in the Reference section. The reports and papers are also available on the API web site.

Five different oil types have been used in this research program. They were selected as reference oils to cover a wide range of oil properties (density, viscosity, wax/asphaltene content, pour point and evaporative loss), see list below. Most of the experiments were done with the Oseberg crude, a light paraffinic crude with the characteristics typical of a high-volume blow-out. Further details regarding the properties of these oils can be found in the technical reports Brandvik et al., 2014 (Phase-II report).

The three different dispersant products used in this study Corexit^®9500A, Dasic^® Slickgone NS and Finasol^® OSR-52 are approved and included in oil spill contingency plans in many regions in North America and Europe. All products were used as they were delivered by their suppliers.

The primary test facility was SINTEFs Tower Basin in Trondheim, Norway. This facility consists of a 6-m high, 3-m deep tower holding 42 000 litres of natural sea water. The facility was operated at ambient pressure using release nozzles from 0.5 – 50 mm in diameter and with oil release rates of 50 mL to 400 L per minute. Further details can be found in Brandvik et. al., 2013.

To perform experiments under pressurized conditions similar to deep-water conditions, a hyperbaric chamber at Southwest Research Institute, San Antonio, TX, US was used. This chamber is 5 m deep, 2.2 m wide and holds 24 400 litres of artificial sea water. Further details can be found in Phase-III & V technical reports (Brandvik et al., 2016 a-b).

The mesocosm wave tank facility of the US Bureau of Safety and Environmental Enforcement’s OHMSETT Facility is located in Leonardo, New Jersey, US. The large dimensions of this facility (200 m long, 20 m wide and 2.4 m deep), enabled significantly longer duration oil releases than was possible in the other facilities.

The Inverted Cone method has been used by a number of researchers to study the rise of droplets and bubbles in the plume and far-field phases of a blow-out, e.g. Maini and Bishnoi (1981) and Masutani and Adams (2000). Buoyant oil droplets are injected into a downward-flowing stream of water, below an inverted cone (often referred to as an Imhoff cone) and rise through the cone, encountering gradually increasing downward velocity caused by the reduced cross sectional area. At some point the drag forces on the droplet from the downward flowing water balance the droplet’s buoyancy and the droplet becomes stationary. Adjustments in the speed of the downward-flowing water can then relocate the droplet so it is suspended in the imaging section where it can be observed. High definition images of these droplets allowed changes in the size and shape caused by dissolution, shedding of small droplets or splitting to be quantified over time. The inverted cone system simulated buoyant droplets rising through the oceanic water column in a finite height facility, albeit imperfectly, since the change in water pressure with depth is not simulated.

At SINTEF, the droplets were created with a nozzle system equivalent to that of the Tower Basin facility (1.5 - 3 mm). The individual droplets captured and quantified are therefore taken from a sample of droplets created with a turbulent jet. This makes tip-streaming easier to study since captured droplets don’t split because they’ve exceeded their maximum stable droplet size (Hu and Kintner, 1955). Further details can be found in Davies et al., 2016a and Davies at el., 2016b.

Two main approaches were used in this study to quantify the size of oil droplets and gas bubbles.

In Phase-I and II oil droplets were quantified with a laser diffraction instrument (LISST-100X, by Sequoia Scientific Inc.), which reports an equivalent-sphere size distribution by way of inversion of the recorded small-angle forward scattering signature (Agrawal and Pottsmith, 2000). The instrument takes 10 measurements every second (covering 32 logarithmically-spaced size bins in the 2.5–500 μm range). An average taken over a 30 second period was used to represent each droplet size distribution. Averaging over this period reduced uncertainties from possible drifting or pulsing in oil or dispersant flow rates and inhomogeneity in the rising oil & gas plume.

An alternative to the LISST-100X was developed for the later phases, the so-called ‘SINTEF SilCam’, to overcome the limited size range (< 500 μm) and interference from larger particles of the LISST system (Davies et al., 2012) and its inability to discriminate between oil droplets and gas bubbles. The Silcam uses the principle of backlighting to create silhouettes of particles suspended between a light and the camera. Fifteen images are taken per second with a path length that is adjusted for droplet/bubble concentration and to ensure that the largest particles are able to pass through the sample volume. Three different magnifications can be used, depending on the experimental setup. These magnifications result in minimum quantifiable diameters (equivalent circular diameter) of: 28, 56, and 107μm. The number of droplets per image during a typical experiment is usually several hundred or more. Errors due to out-of-focus particles are removed by use of telecentric receiving optics that have a depth-of-field that is equal to or wider than the physical restriction or path length of the measurement cell (i.e. the gap between the illumination and camera housing windows). Path lengths chosen therefore vary from approximately 3 mm for the highest magnification to up-to 50 mm for the lowest magnification.

Particle dimensions are quantified and used to determine individual droplet/bubble sizes and corresponding size distribution. Equivalent circular diameters for every particle are counted, by volume concentration, into log-spaced size bins that match the LISST-100 (Type-C spherical inversion) and extend a further 21 size classes larger than the 500μm upper limit of the LISST-100, yielding a total of 53 log-spaced size classes

Each size distribution is often based on over 1-million individual droplets/bubbles, enabling reliable statistics even for the largest droplets/bubbles. Multiple camera configurations with different magnifications were used to quantify both the small, dispersed oil droplets (28 – 800 microns) and large droplets of untreated oil (up to ~12 mm) during the large scale testing (Phase-V). For the high-pressure testing at SwRI (Phase III and VI) a version of the SilCam rated for 300 bars was developed.

To study combined releases of oil and gas, software was developed to distinguish between oil droplets and gas bubbles. This makes use of a classification scheme based on the different wavelengths of light transmitted through the droplets/bubbles. As these transmitted signatures are unique to a droplet/bubbles specific refractive index and absorption characteristic (i.e. optical properties related to composition), it is possible to automatically separate oil droplets, gas bubbles and oil-coated gas bubbles of equal size and shape (Davies et al., 2014). Further details of the method used for this classification and a full description of the hardware and analysis routines for the SINTEF SilCam are available in Davies at al. 2017.

The first study in this research program used the SINTEF Tower Basin to quantify the change in droplet size as a function of release nozzle diameter, flow rate, dispersant-oil-ratio (DOR) and dispersant injection method. These experiments were done with the same oil and dispersant (Oseberg and Corexit^® 9500). A full description of the study is available in Phase-I Final report (Brandvik et al., 2014a).

Figure 1 shows the initial droplet size distribution (volume %) for various dispersant dosages and demonstrates the decrease in droplet size that can be achieved with only 1% dispersant (DOR: 1:100), compared to the 4% dosage typically used for surface applications. In this case, a dosage of 1% drops the mean droplet size (d₅₀) by roughly a factor of 3, while doubling the dispersant to 2% increases this factor to 4. Note that these factors are at laboratory scale and as described later in this paper, these factors are considerably larger at field scales. Measurements of IFT on oil samples taken from the oil plume, show that a DOR of 1–2% typically decreases IFT in the Oseberg oil by a factor of 100–200.

Multiple dispersant injection techniques were tested in this study (upstream injection, simulated insertion tool (SIT), injection above, and horizontal injection). They all gave significant reductions in oil droplet sizes. The most important variable was the distance between the dispersant injection point and the release point of the oil. The dispersant caused the most significant reduction in droplet sizes when it was injected right before or immediately after the release point of the oil, see Brandvik et al., 2016d for further details.

A subsea oil release forming a turbulent jet will form a compact oil cone up to approximately 6–10 nozzle diameters downstream of the release point. Injecting the dispersant into this compact oil cone seems to be most efficient. With this approach, the dispersant will mix into the oil and reduce the IFT before the oil enters the highly turbulent zone where the oil droplets and gas bubbles are formed. These results indicate that an injection wand like the one used during DWH could be highly efficient as long as the dispersant was injected immediately above the release pipe or into the release pipe.

The sensitivity of droplet size to oil flow rate and outflow pipe diameter was also extensively studied in Phase 1 and results were used to confirm and refine a mathematical model of initial droplet size based on Weber scaling (Hinze, 1955), developed in previous studies and is described more fully in Johansen et al. (2013):

where d₅₀ is the median oil droplet size, D is the orifice diameter, A and B are empirical coefficients derived by fitting the lab results, W_e is the exit Weber number (ρU²D/σ), ρ is the oil density, U is the exit velocity, σ is the interfacial tension (IFT), V_i is the Viscosity number (μ U/σ), and μ is the dynamic viscosity of oil. Fitting the Phase 1 data to Equation 1 suggested A = 24.6 and B = 0.08.

The model presented in Eq. 1 covers cases with momentum jets and single fluid releases (oil only). For combined releases with gas and oil, a void fraction correction of the release velocity (U_n) as described in Eq. 2 is used.

Where n is the gas volume fraction. In case of large volume flows, which are buoyancy dominated, an exit Froude number correction should be applied, as described in Eq. 3.

Where Fr = U_n/(g′ D)^1/2 and g′= g [ρ_w − ρ_oil (1 - n)]/ρ_w. Further details and discussions are found in Johansen et al., 2013.

Modified Weber scaling is used later in this paper to compare its predictions with the relevant results from the Phase II-VI studies and to evaluate the impact of dispersant on the DWH oil spill. Rosin-Rammler and lognormal distributions were evaluated to fit the droplet-size distributions. The lognormal distributions gave a slightly better fit to untreated oil while the Rosin-Rammler performed somewhat better for oil treated with dispersants.

Different oil types and three different dispersant products were included in the experimental work in phase-II. In addition to Corexit^® 9500, the experiments included two other dispersants, Finasol^® OSR 52 and Dasic^® Slickgone NS. In addition to the Oseberg oil, Phase II studied, Norne blend, Troll B, Grane and Kobbe condensate (see Table 1).

The effectiveness of a specific dispersant was found to be somewhat dependent on oil type. Overall, Corexit^® 9500 was found to reduce droplet size the most, followed by Finasol^® OSR 52 and Dasic^® NS (see example with Kobbe oil in Figure 2). More details regarding the effectiveness testing of different products and oil types are given in the Phase-II technical report (Brandvik et al., 2014b).

Most of the work in this research initiative was performed at a water depth of 2–5 m. To include possible effects of the elevated pressures that occur in deep-water, selected experiments were performed at pressures up to 2500 PSI (1750 m depth) in a hyperbaric chamber at SwRI’s facilities in San Antonio, TX, USA.

In Phase-III, experiments with dead oil (oil stabilized at 1 atm, without any dissolved gas) treated with dispersants (Corexit^® 9500, 1% and 2%) were performed to study possible pressure effects on droplet formation and dispersant effectiveness. No significant effect of pressure was found on the initial droplet formation (see Brandvik et al. 2016a for details).

Combined releases of live oil (to simulate oil with dissolved gas that would be released during a blowout) and additional free natural gas under high pressure were performed as a part of Phase V using the hyperbaric chambers at SwRI. For these experiments Oseberg blend (density: 850 g/L, viscosity: 2 mPas) was recombined with natural gas under different pressures (58, 116 and 175 atm) and high temperature (80 °C) to produce a fully gas-saturated oil (density: 0.690 g/L, viscosity: 0.7 mPas). The oil was released at varying pressure in combination with different ratios of natural gas.

Figure 3 shows experimental oil d₅₀ for the combined releases of simulated live oil and natural gas with a GOR of 1:1 at four different pressures. The nearly horizontal data points showed there was no significant effect of pressure on oil droplet formation. The one major outlier for this horizontal trend was for the untreated live oil with 1:1 GOR at 5 m pressure (SINTEF Tower Basin). Also note the GOR (volume ratio) was kept constant as the pressure was increased. Thus the gas density increased with increasing pressure. This gave the discharge more momentum, which in turn decreased both the oil droplet size as pressure increased.

The complex physics involved with the live oil and associated gas were reasonably well captured by the Weber scaling (Eq. 1 and 2). The predictions are about 10% lower than the observations for the untreated live oil & gas (solid black symbols) and in some cases as much as 25% different than for treated oil (open black symbols). See Phase-V report for details (Brandvik et al. 2016b).

Phase I, II and III focused on initial droplet formation in the highly turbulent zone immediately above the release nozzle. Quantification of oil droplet sizes was done 2 m (650 – 1300 release diameters) above the release. This was necessary in order to provide sufficient dilution for accurate measurement and a more homogeneous distribution of the droplets in the plume (Brandvik et al., 2013). The assumption being that initial droplet formation due to turbulent shearing is likely complete at this distance (more than 1000 release diameters) from the nozzle (Or et al., 2011). To test this assumption, twin LISST instruments were mounted at 2 and 5 m above the release point. This set-up was used to study possible coalescence or droplet splitting after initial formation. Four different oil types and two dispersants were included in this study. No significant changes in the droplet sizes were observed. Details are given in Brandvik et al., 2014b. These measurements suggest that if secondary processes like breakup of large unstable droplets, tip-streaming and coalescence are important, they must be occurring within the first 2 m above the release or at greater than 5 m above the release. It seems unlikely that coalescence will occur above 5 m if it was not apparent below 5 m because the concentration of droplets is rapidly diminishing with distance from the release point.

Figure 4 shows an example of “tip streaming” which is the formation of micro droplets due to turbulent shearing and ultra-low interfacial tension at the edges of large, freely-rising dispersant treated oil droplets. These droplets are usually not created by a turbulent jet, but rather by gently releasing a large droplet similar to pendant drop conditions. Such low turbulent release conditions were needed to create these large unstable multiple millimeter droplets with very low IFT (2% C9500). These droplets are too large (> d_max) to be stable under turbulent jet conditions.

Tip-streaming and droplet shedding were also observed previously in a study by Gopalan and Katz (2010). They pre-mixed oil with high concentrations of dispersant (3–7%) and released the oil under low turbulence to create large unstable droplets, with a low IFT, to study tip-streaming.

In Phase IV of this research effort, tip-streaming was studied both at UoH and SINTEF by capturing individual droplets in the observation chamber of the inverted cone apparatus described earlier. Both turbulent and non-turbulent conditions were studied. Tip-streaming was found to vary with oil viscosity, droplet size and interfacial tension (dispersant dosage). For droplet sizes in the range of 500 to 800 microns created by a turbulent jet after addition of 1–2% dispersant (Corexit^® 9500), tip-streaming was usually observed for a limited period (minutes) until a stable droplet size was formed, see example in Figure 5.

The tip-streaming process is shown to play a significant role in modifying the initial size distribution from subsea releases in situations where the oil viscosity is sufficiently high and the interfacial tension is sufficiently low. In all cases where tip-streaming was observed, the original ‘parent’ droplet reached a stable size within a few minutes after the initial formation, after which no further droplet shedding was observed.

In agreement with previous published work (Gopalan and Katz, 2010), tip-streaming can be predicted by use of a critical modified capillary number of 0.21. This has now been validated for realistic dispersant dosages for subsea dispersant injection, and for droplets formed within a turbulent release. Viscosity played a key role in determining the significance of tip-streaming, where higher viscosity oils are more prone to tip-streaming than low viscosity oils. These viscosity effects are captured within the modified capillary number. Further details can be found in Davies at al., 2016b.

The work performed in Phase-I to V focused on down-scaled laboratory experiments with release diameters in the 0.5–3 mm range with flow rates of 0.1 to 10 L/min. These results together with the data from the DeepSpill experiment (120 mm release diameter, 840 m depth, Johansen et al., 2003) form the basis for suggesting a new algorithm for initial droplet formation using modified Weber scaling (Johansen et al., 2013). To further verify the new algorithm, laboratory experiments were performed with a more realistic size range of oil droplets. These experiments were performed at the SINTEF Tower Basin, in Norway and at the large outdoor test basin at the Ohmsett facility in NJ, USA. The oil droplet sizes targeted in this study were;

1. 1 - 12 mm for untreated oil and

2. 0.05 – 1 mm for oil treated with dispersants (1% C9500).

The challenge of quantifying these large oil droplets required development of new equipment for monitoring of oil releases since the LISST instrument used in phase-1 and II was not capable of measuring droplets greater than 0.5 mm. The new SilCam has been found to correlate well with the LISST for the smallest droplets (< 0.5 mm). More details are given in the Phase-III & V report (Brandvik et al., 2016a,b) and Davies et al. 2016b.

Figure 6 shows examples of droplet distributions from up-scaled experiments in Phase VI. The measured droplet sizes (d₅₀) are given in color above the individual distributions (50 mm, 200–400 L/min) and the predicted d₅₀ are given in the embedded box. The predicted sizes for the untreated oil are biased slightly high (1 to 10%). However, precise quantification of dispersed droplets sizes is a challenge due to the high droplet density at these flow rates (300 L/min). Comparing the yellow solid curve and the dotted red curve in Figure 6 (300 L/min with and without dispersant), suggests that injecting 1% dispersant can decrease oil droplet sizes (d₅₀) by an order of magnitude.

Another important finding from the larger-scale experiments was that dispersants mixed into the oil and caused a uniform reduction in droplet size as demonstrated by the unimodal droplet-size distribution shown in Figure 8. This unimodal behavior was observed in all the droplet size distributions measured during the Ohmsett testing and is consistent with the smaller-scale experiments performed at SINTEF and SwRI. This indicates that the injection techniques used to simulate a simple injection wand are adequate, however, larger scale tests are needed to fully verify.

Experimental results from Phases I–VI are compared with predicted values from Equations 1–3 in Figure 7. The experimental results come from a large range of release diameters, oil & gas flow rates and release velocities and compare well with the predictions from the Modified Weber scaling. Also the DeepSpill2000 experiment at 840 m depth with a 120 mm release and 1000 L/min flow rate (Johansen et al., 2003) fits nicely on the line.

These finding strongly contradict the work of Paris et al. (2012) and Aman et al. (2015) who suggested that the turbulence in the release would naturally create oil droplets less than 100 μm and claimed that SSDI had no effect on the DWH spill. They used a similar version of Equation 1 but with coefficients derived from a small laboratory apparatus using a high speed mechanical stirrer to generate turbulence. They provide no discussion of how their apparatus, that subjects oil to long periods of high turbulence, replicates the turbulence generated by a blowout (where turbulence rapidly dissipates). Adams et al. (2013) suggested several other reasons that their equation will significantly underestimate the droplet size that would be generated in a blowout. Modified Weber scaling has been implemented in most operational models used to predict fate and effect of subsea oil releases (Socolofsky et al., 2015).

If the modified Weber scaling (Eq. 1–3) is used to predict oil droplet sizes for a deep-water blowout scenario like DWH, the result (oil alone & SSDI), is shown in Figure 8. The calculation assumes a light oil (0.830 kg/L), release diameters of 0.40 m, oil flow of 60,000 bbl/day and an actual GOR of 1:1 in 1500 m of water. The IFT for the oil (20 mN/m) is lowered by a factor of 100 during SSDI, see Brandvik et al., 2016d for a general algorithm. A lognormal distribution has been applied to the treated oil and Rosin Rammler to untreated oil. When using the modified Weber scaling for large-scale scenarios with such low release velocities, it is important to limit the upper tail of the droplet distribution to the maximum stable droplets size, d_max (Hu and Kitner, 1955). For example, if the d₉₅ > d_max then for the predicted distribution d₉₅ is set to d_max. In the case in Figure 8, d_max is 14.4 mm for untreated oil and 2.0 mm for treated oil.

At scales that might be possible for a large blowout, Figure 8, suggests 1–2 % dispersant reduces the mean droplet size by an order of magnitude. A droplet size reduction of this magnitude will substantially slow the rise of a significant oil volume towards the surface. Rising velocity of the oil droplets are mainly determined by Stokes law, however for larger droplets (> 0.8mm), the drag forces also needs to be accounted for (Hu and Kintner, 1955). Oil droplets between 4–10 mm (realistic range for the untreated oil) require 3–4 hours to reach the surface from 1500 m depth, while 0.2 - 0.8 mm droplets (possible range for dispersed oil) will need from 12 h (0.8 mm) to 4 days (0.2 mm) to reach the surface. This kind of delay allows for increased dissolution which will further reduce droplet diameters and increase droplet density to even further slow the rise of the droplets towards the surface. These rising velocities assume stagnant water and oceanographic turbulence, cross flows and density layers could significantly extend droplet rising time. Microorganisms may also generate “flocs” of microbial biomolecules and oil degradation products (Brakstad et al. 2015). These “flocs” may have neutral or even slightly negative buoyancy (Passow et al. 2012).

A delay in surfacing of the oil means droplets can be advected away from the release site by background currents, resulting in a thinner slick distributed over a larger area. Such thin surface oil films might be too thin to emulsify and much more readily entrained into the upper mixed layer of the water column by surface waves. This is in strong contrast to the thick surface oil slick formed by rapidly rising large oil droplets (multiple millimeters) that can emulsify and form persistent slicks that could persist for weeks thus increasing the likelihood that oil will reach shorelines. The difference in persistence between thick emulsifying and thin non-emulsifying (< 0.1 mm) surface oil films has been observed in several experimental subsurface oil releases and real incidents (Johansen et al. 2003, Daling et al., 2015, Rye et al., 1996, and Rye et al 1997).

The reduced droplet size caused by SSDI means that oil will be further advected downstream by ambient horizontal currents to the point that the fraction that surfaces, will do so far from the well site. All of these factors can greatly enhance the health and safety of the responders working at the surface near the well.

The effectiveness of SSDI was found to be fairly insensitive to the dispersant injection technique as long as the injection was less than six release diameters from the release point. Once this fact was established, most of the subsequent tests were performed with the so-called Simulated Insertion Tool (SIT), a wand-like tool inserted into the release opening. This arrangement is quite similar to that used during the DWH spill.

In the case of subsea releases, a dosage of 1% showed a significant reduction in droplet sizes and should be sufficient for the lighter oils that will be characteristic of high volume blow-outs. Larger dosages will probably be needed for heavier oils, but such oils will usually have reduced flow rates in most subsea release scenarios because heavy, viscous oils cannot flow through porous reservoir rock at the same rate as light, low-viscosity oils.

This high efficiency of SSDI is due to multiple factors the most important being that the injection occurs at high turbulence levels near the blowout and the fact that the oil is fresh and not weathered and emulsified.

Significant differences in SSDI effectiveness were found for different oil types that ranged from a condensate to heavier, asphaltinic crudes. For the latter, SSDI had reduced effectiveness. These results suggest SSDI effectiveness should be assessed during contingency planning in order to optimize SSDI effectiveness.

Three commercial brands of dispersants were tested; Corexit^® 9500, Finasol^® OSR 52 and Dasic^® Slickgone NS. Under the conditions tested in these studies, all three dispersants significantly reduced droplets sizes, however, for most oils Corexit^® 9500 generated the smallest droplets, followed by Finasol^® OSR 52, and then Dasic^® Slickgone NS.

Previous work has shown that after initial formation, droplets treated with dispersant can exhibit extensive secondary break-up through a process known as “tip-streaming”. Experiments from Phase IV suggest that droplets less than 1 mm stabilize after initial tip-streaming and remain relatively large. This suggests that tip-streaming may not be that important for most oils treated with SSDI since these will form droplets smaller than 1 mm.

As for coalescence, measurements suggest it is insignificant at 2 m or greater above the release. Closer to the outlet it was difficult to make measurements, so both droplet splitting and coalescence may be important there. In any event, these processes are included in the calibration coefficients of the modified Weber scaling that was found to nicely fit all the laboratory studies.

Oil released at varying pressure showed no significant difference in oil droplet sizes as a function of pressure (6 m to 1750 m of water) both for oil alone and when treated with dispersant. Also in experiments with simulated live oil and with live oil & associated gas, initial droplet formation showed no dependency on pressure, beyond the effect caused by the increased density of the gas (GOR at the release point was maintained at 1:1 across all pressures). This strongly indicates that SSDI effectiveness is not dependent on water depth or pressure.

SDDI also strongly influenced gas bubble sizes. This should be included in future modelling of subsea releases, since the total surface area of a given volume of gas increases as bubble size decreases, and this will increase gas dissolution into the water and hence plume buoyancy.

Modified Weber scaling given in Equation 1–3 was used to predict oil droplet sizes for several of the experiments performed in this study. The predictions are good over a wide range of flow rates, release diameters, and oil types including the larger-scale results at Ohmsett and the DeepSpill field experiment performed by SINTEF in 2000 (Johansen, et al., 2003, Rye et al., 2003). The Modified Weber Scaling suggests that for an important range of blow-out scenarios SSDI could reduce the droplet size by an order of magnitude.

These findings strongly contradict the work of Paris et al. (2012) and Aman et al. (2015) who conclude that SSDI during the DWH spill was unnecessary because they believe droplets were <100 microns even without addition of dispersant. Their findings are likely because their test system kept oil under constant high shear conditions for an extended period rather than the rapidly decreasing shear of a jet of oil released into water. As suggested by Adams et al. (2013) there are many other reasons to believe that their equation will significantly underestimate the droplet size that would be generated in a blow-out.

The combined results from the different phases in this study suggest SSDI is an important response tool, which can significantly reduce the volume of surfacing oil, increase natural degradation, and reduce the persistence and safety implications of oil that does surface. The ultimate decision of whether to use SSDI in a specific event will depend on a net environmental benefit analysis (NEBA). These studies also provide measurements at laboratory scales that can be used to validate algorithms for predicting oil droplet size. Performing field trials releasing oil have for the last four decades proven to be important for verifying models for fate and behavior of marine oil spills (Faksness et al., 2016). Despite studying a wide range of release conditions in these experiments, it is still desirable to further validate the modified Weber Scaling with an experimental deepwater release with live oil, associated gas, and dispersant injection, offering more realistic release conditions.

The authors want to thank technicians and engineers at SINTEF in Trondheim, Norway, at the University of Hawaii, USA, at SL Ross in Canada, at Ohmsett in NJ, USA and at SwRI in Texas, USA. The valuable outcome of this comprehensive four-year research effort had not been possible without their dedicated support. The main source of funding for the research described in this paper has been the American Petroleum Institute.