Oil Spill Modeling for the Enbridge Line 3 Replacement Program with Response for Planning, Preparedness, and Environmental Purposes

The proposed Enbridge Line 3 Replacement Program (L3RP) would replace the existing aging pipeline from Hardisty, AB, Canada to Superior, WI, USA (Figure 1). Separate environmental assessments were prepared for each regulating body investigating completely unmitigated (i.e. no emergency response, containment or collection) full bore ruptures into a range of sensitive waterways. In Canada, an Ecological and Human Health Risk Assessment (EHHRA) (Stantec et al., 2015) and technical report (Horn et al., 2015) were prepared for the National Energy Board (NEB). In the United States, an Assessment of Accidental Releases (Stantec et al., 2017) investigating unmitigated releases and a Supplemental Release Report (Horn et al., 2017) investigating the effects of oil spill response options and smaller volume releases were part of an Environmental Impact Statement (EIS) (MN DOC, 2017) that was prepared for the Minnesota Public Utilities Commission (PUC) and Minnesota Department of Commerce, Energy Environmental Review and Analysis (DOC-EERA).

Each regulatory authority required an assessment of the potential impacts associated with accidental releases of crude oil along the proposed replacement pipeline routes. These assessments were to include estimates of potential extents, both spatial (i.e. distance) and temporal (i.e. timing), and the degree of associated impacts from large-scale releases along the pipeline routes. Watercourse crossings were the focus for both the Canadian and US assessments and hypothetical releases were to be modeled at multiple representative locations along the pipeline to bound the range of potential effects and allow for broad environmental comparisons among and across the routes and between areas with similar features. The Canadian EHHRA was focused on the single proposed route through Alberta, Saskatchewan, and Manitoba, while the US EIS assessment considered the preferred route and multiple route alternatives mainly throughout Minnesota (Figure 1).

In total, eleven sites were investigated to bound the geographic and environmental variability that could occur over the entire pipeline corridor, should a release occur at any point along the pipeline at any time of the year. The sites represent a diverse set of water and land features including large rivers down to small streams and ditches, with a range of flows/velocities, varying amounts of turbulence (e.g. waterfalls), differences in channel morphology (e.g., sinuous to straight), and several lakes. They also represent a wide range of bank/vegetation types and land uses, including several different forest types, protected areas, cultivated land, wild rice lakes, recreational areas, and populated areas. Three hypothetical releases were modeled at each representative location to account for the variable geographic and environmental conditions under wintertime low, summer/fall average, and springtime high river flow conditions.

The most common cause of pipeline failure is third-party damage (e.g. line strike), which is more likely to occur in areas that have a reduced depth of cover (i.e. soil depth of earth that is used to bury a pipeline within the trench) (Stantec et al. 2017). The focus here is on hypothetical releases simulated at Location 10, the Mississippi River at Palisade (Figure 1). At this location, the planned replacement would include a horizontal directional drill (HDD) crossing, whereby approximately 0.5 km on either side of the watercourse crossing, the pipeline would begin to “dive” from a nominal depth of cover of approximately three feet to a depth that was tens of feet below the bottom of the river at the crossing. In addition, while the EHHRA and EIS investigated completely unmitigated (i.e. no emergency response, containment or collection) full bore ruptures (Stantec et al., 2015; 2017, Horn et al., 2015; MN DOC 2017), the Mississippi River at Palisade included an additional assessment which investigated more realistic (i.e. smaller) release volumes (i.e. average historical accidental release volumes instead of full bore ruptures), releases onto land (that would enter the waterway; Figure 2) as well as an effective emergency response (i.e. response mitigated) (Horn et al. 2017). The goal was to assess the differences in the transport and fate of released oil during more realistic smaller-volume response-mitigated hypothetical releases and, where applicable, compare them to the unmitigated releases modeled in the EIS.

The RPS OILMAPLand and SIMAP models were selected for this assessment to determine the overland and in-water movement and behavior of crude oil following each hypothetical release. OILMAPLand is a two-dimensional modeling system that was used to simulate the movement of released oil in the environment (Figure 3; left). It simulates the flow of oil over land as it travels over the land surface and into a surface water body. The model itself has three components, including an overland release model, a surface water transport model, and an evaporative model that describes the weathering of oil in the environment under specified conditions. SIMAP is a three-dimensional modeling system that was used to simulate the physical fates of crude oil in the water (Figure 3; right). It estimates the distribution (as mass and concentrations) of whole oil and components of oil on the water surface, on shorelines, in the water column, in sediments, and evaporated to the atmosphere. This comprehensive SIMAP modeling system allows for a more in-depth understanding of the behavior of oil in the environment, when compared to OILMAPLand. Oil fate processes included in SIMAP are oil spreading (gravitational and by shearing) of floating surface oil, evaporation, transport, randomized dispersion, emulsification, natural entrainment, dissolution, volatilization of dissolved hydrocarbons from the surface water, adherence of oil droplets to suspended sediments, adsorption of soluble and sparingly-soluble aromatics to suspended sediments, sedimentation (i.e. sinking oil), and degradation. As the oil weathering processes are simulated, the product remaining within the environment becomes increasingly dense and viscous as there is a reduction in volatile and soluble content (e.g. lighter ends). Of note, the SIMAP model uses total suspended sediment concentrations within the water column, organic carbon content of the material, and settling rates of particles through the water column to predict the total amount of oil that is able to be transported to sediments within the simulated timeframe. Both the OILMAPLand and SIMAP models are used extensively by industry and governments (French-McCay, 2004; Horn and French-McCay, 2015; Horn and Fontenault, 2018; Horn et al., 2018). Detailed descriptions of the algorithms and assumptions in the SIMAP model and validations to field and laboratory data may be found in published papers (French McCay, 2002; 2003; 2004; 2009; French McCay et al. 2015; 2018). Hindcasts and modeling studies span marine, estuarine, and freshwater environments, including river cases and orimulsion studies (French et al. 1997; French McCay 2003; 2004; French McCay et al., 2005; Horn and French McCay, 2015; French McCay et al. 2015; 2018a,b,c).

Site-specific and season-specific geographic and environmental parameters were gathered from federal, state, and local data sources to capture the conditions within the receiving environment over the course of any given year for each of the eleven modeled hypothetical release locations. Several key parameters include elevation, land cover data, shoreline type, habitats, bathymetry, temperature, wind and currents (spatially and temporally variable speed and direction), and total suspended solids concentrations within the water column. In addition, the chemical and physical properties of the crude oil and the release volume are characterized. This includes composition by chemical constituent, as well as density, viscosity, volatility, solubility, and other key factors that will affect the way in which the oil behaves and weathers, once released into the environment. Modeled release volumes representative of full bore ruptures were modeled as underwater releases from the center of the river channel. For the Mississippi River at Palisade additional release volumes were simulated including the full bore rupture (11,840 bbl), Initial Valve Placement (IVP) full bore rupture (8,693 bbl) for the hypothetical release locations on land, and the average historical accidental release volume (HARV; 1,911 bbl) from all Enbridge pipelines that ship crude oil and are on the Mainline system for the hypothetical release locations on land (Table 1). IVP and HARV release volumes were investigated at two on land locations (Figure 2).

The inclusion of multiple hypothetical release locations and release volumes simulated at the Mississippi River at Palisade were different than the broader assessments (EIS and EHHRA), which included full bore, unmitigated releases of two crude oil types (Bakken Crude and Cold Lake Blend), under three environmental conditions (wintertime low, summer/fall average, and springtime high river flow conditions) at each of the eleven locations (66 individual scenarios). The broader assessments considered ice cover, which would affect the movement and behavior of a release. However, here we will focus on the Mississippi River at Palisade due to the inclusion of response considerations.

Response actions modeled in this study included surface and shoreline containment booming as well as surface skimming technologies. Surface and shoreline containment booms were modeled at each of the four Control Points approximately 20, 40, 50, and 60 km downstream, which were to be activated following the release at hours 4, 5, 6 and 7. Booms were positioned in a ‘V’ shape where the offshore arm extended into the river channel and funneled oil flowing downstream into the point of the ‘V’, where a collection skimmer was mounted to maximize surface oil collection, and the nearshore arm protected the shoreline from potential oiling. Within each Control Point, a single unit of surface skimming equipment was included; either an ELASTEC TSD 136 Smooth Drum or ELASTEC UNO48G Grooved Drum. Response plans were provided by Enbridge (2017), which identified the timing and placement of each containment boom and surface skimmer.

The modeled response option included conservative assumptions such as effectiveness or efficiency (<100%), based upon tank tests and field experiments investigation rates of collection. In addition, the modeling considered environmental conditions and associated thresholds for collection, should wind speed, current speed, or wave height become too great or the thickness of surface oil became too thin. Stationary automated skimmers were assumed in this modeling and once set up (during daylight hours only), were allowed to have continuous collection throughout the day and night. Efficiency of skimming was assumed to be identical between day- and nighttime hours.

OILMAPLand results for the original planned location where nominal depth of cover for the pipeline was 3 feet (Location A) resulted in no oil predicted to reach the watercourse for the HARV release volumes (Figure 4; Table 2). Essentially, not having sufficient volume to reach the Mississippi River, all of the oil was predicted to be retained on land, fill depressions, and evaporate. Therefore, Location B was chosen 20 m (65 feet) to the west-northwest. Due to small differences in topographic features that would likely result in different overland trajectories of released oil. This resulted in oil being predicted to enter the Mississippi River. Approximately 89% of the oil initiated from an IVP release reached the Mississippi River from HRL-B and between 46 and 53% of oil initiated from an average volume pipeline release reached the Mississippi River depending on the oil type and season (Figure 4; Table 2).

Once oil was predicted to reach the Mississippi River using OILMAPLand, the SIMAP model was initiated to predict the three-dimensional movement and behavior of oil in the water using the volume that was predicted to reach the river (Table 2). The effects of response activities at each CP on the trajectory and fate of the oil for each release scenario was predicted to assess the differences in the transport and fate of oil during response-mitigated hypothetical releases. This study focused on the potential surface, shoreline, and sediment oiling that may result from hypothetical releases, as well as the potential for containment and collection. As an example, shoreline oiling and surface oil thickness (by mass) is presented to demonstrate the larger amount of oil upstream of each CP, when compared to the downstream locations (Figure 5). In addition, two mass balances figures are provided to demonstrate the different trajectory and fate of oil based upon different release volumes under the same environmental conditions (Figure 6). Over the five-day simulation, no oil was predicted to sink, due predominantly to the low quantity of total suspended solids within the water column during the simulated timeframes.

The boom at each CP is predicted to retain a large portion of oil and the skimmer removes it from the river. Downstream of each CP, there is a smaller and more discontinuous mass of oil than areas immediately upstream of each CP. Comparing CP1 to each successive downstream CP, a smaller mass of oil was predicted to remain on the water surface. Under average river flow conditions, floating surface oil was predicted to first reach CP1 roughly 21.6 hours following the release. The large initial mass of floating oil contained by the boom resulted in the skimmer operating at full capacity, but the encounter rate exceeded the collection rate. With additional oil continuing to make contact with the containment boom, the excess oil was predicted to eventually make its way under and around the boom. By the time the floating surface oil was predicted to reach CP3 (after roughly 2 days 9 hours) and CP4 (after roughly 3 days 5 hours), there was substantially less mass of surface oil available to be contained and collected. While nearly 20% of the released oil from full bore ruptures of Bakken were predicted to be contained and collected under average river flow conditions, there was not a sufficient amount of oil that was needed to reach CP1 in the HARV scenarios and oil was predicted to strand on shorelines and evaporate.

The high viscosity and lower evaporation rates for Cold Lake Blend, when compared to Bakken crude oil, resulted in thicker predicted surface oil thicknesses under identical environmental conditions. This resulted in what may be perceived as a more successful predicted response for Cold Lake Blend under the modeled conditions. However, it is important to note that the evaporative flux from Bakken was predicted to be greater and would still result in less oil being present in the aquatic environment following a release, compared to Cold Lake Blend. For the Bakken IVP cases, this corresponded with approximately 20% collected, with 25% ashore, and approximately 55% evaporated. For the Cold Lake Blend IVP case, this corresponded with approximately 50% collected, 30% ashore, and 20% evaporated.

The total predicted amount of oil collected at each CP ranged from approximately 0–300 metric tons (1 MT = 1,000 kg), depending on the release scenario, and 0–911.2 MT for the total of all four CPs (Table 3). Note that oil is tracked within the SIMAP model in units of mass, as density changes with time due to environmental conditions (e.g. temperature) and state of weathering. Due to weathering processes and interactions with the water, it is anticipated that the total volume of material collected, including oil and water emulsion, would range from 0–650 m³ for each CP and 0–1,965 m³ in total. In general, the modeled release volumes from the HARV did not have sufficient volume to travel overland and downstream to the first CP, where oil would have been collected. However, with the IVP and full-bore volume simulations, large masses of oil were predicted to enter the river with sufficient volume to be collected at CP1-4. The timing of first collection at each CP ranged from 16.6–75.5 hours, depending on the scenario, with later times associated with CPs that were further downstream (Table 4).

Computational oil spill modeling was used to support evaluation of the downstream movement, behavior, timing, and potential ecological and human health risks resulting from hypothetical releases of crude oil from the proposed Enbridge Line 3 Replacement Program using four of the designated CP's. The investigation involved assessing multiple hypothetical crude oil releases from the pipeline onto terrestrial and into aquatic environments with the intent of understanding the expected spatial extent, timing, and magnitude of hydrocarbon contamination, as well as the relative success of response activities.

Small differences in the topography of land had the potential to strongly influence the trajectory and fate of oil released onto land. Depressional storage (i.e. filling basins) may greatly reduce the total amount of a release that could enter nearby waterways. Similarly, the exact slope of the land from the release location has the potential to greatly influence the exact location that oil may enter the Mississippi River. Ultimately, this predicted location and volume of oil entering the water will have a large influence on the ultimate trajectory and fate of a potential release onto land. Results from response modeling may be used to predict the effectiveness of booming and skimmer strategies at CPs positioned downstream to reduce the mass of floating oil on the water surface over time.

Enbridge had planned for thirteen potential CPs downstream of the Mississippi River near Palisade crossing where containment and collection equipment could be deployed in the event of a hypothetical release. These locations would be equipped with booms and skimmers following a release and have been incorporated in Enbridge's Tactical Response Plans. While these CPs had been proposed as potential deployment locations, additional sites may be occupied in the event of an actual release to respond to the specific release situation in the most effective manner at the time of an incident.

By varying release location, volume of oil, season, and response equipment, this response modeling may be used to more completely understand the range of possible trajectory and fate of hydrocarbons, should there be a release into in the environment from the proposed L3RP at the Mississippi River near Palisade crossing. While it is unlikely that any such event would occur, and that this hypothetical release modeling is in no way a prediction of the exact location of any potential future release, an accurate representation of the ultimate trajectory and fate of a release of oil with response options was provided. In the event of a release, response options would be employed to contain and collect oil with the intent of minimizing potential effects on the environment.

The spill modeling results provided here (from Horn et al., 2017), as well as those contained within the other documents associated with the Line 3 Replacement Program (Stantec et al., 2015; 2017; Horn et al., 2015) were used to aide in environmental assessments and spill response preparedness and the preparation of emergency response plans. Results and reports served as key touchstones within public and evidentiary hearings for the environmental assessments that were required by the Canadian National Energy Board and the Minnesota Public Utilities Commission and Department of Commerce. Computational spill modeling was used to bound the range of potential effects of hypothetical releases of crude oil by varying release locations, seasons, volumes, product types, and emergency response activities. Ultimately, this multi-tiered use of oil spill modeling more-effectively communicated the range of potential effects following a release and aided regulators in the decision-making process.

RPS would like to acknowledge Enbridge for funding these studies as well as the other project team members including Stantec, Dynamic Risk, and Amec Foster Wheeler. We would also like to thank the staff of the Canadian NEB and the Minnesota DOC-EERA and PUC for reviewing a large body technical information. Finally, we would like to thank all public and tribal members that provided feedback during the scoping and review processes that informed state and federal agencies of their concerns.