The Refugio Beach oil spill originated from a pipeline break on 19 May 2015 near Refugio State Beach, which is located approximately 20 miles west of Santa Barbara, California. An estimated 500 barrels (21,000 gallons) of crude oil flowed from the shore side of Highway 101 into the Pacific Ocean. Trajectory and fate modeling of the oil were provided to the Incident Command to support the response. Several factors were particularly challenging for oil spill modeling in this incident. The spill entered the ocean through the surf zone, a distinct dynamical region with variability on small spatial and temporal scales, which coastal circulation models generally do not resolve. The regional winds were also highly variable and in some locations forecast models did not reflect the on-scene observations. A final complication was the presence of numerous natural oil seeps in the region. This posed challenges both for model initialization and validation of modeling results.
In the days following the spill, above-background levels of tarballs were observed coming ashore on various beaches remote from the spill. Chemical analyses indicate that some of the tarballs likely originated from the spill. Hindcast modeling of the spill was conducted to examine transport between these locations and the spill source. Modeling simulations showed regional connectivity in approximately the correct time frame between the spill site and beaches in Ventura County and Santa Monica Bay.
On 19 May 2015, a break in Plains Pipeline Line 901 resulted in a spill of approximately 2,500 barrels (105,000 gallons) of oil. An estimated 500 barrels (21,000 gallons) of the oil flowed from the shore side of Highway 101 into the Pacific Ocean at Refugio State Beach, near Santa Barbara, California (Figure 1). The oil released from the pipeline was a blend of crude oils being transported from the Heritage, Harmony, Hondo, and Holly oil production platforms.
As part of the oil spill response effort, NOAA’s Office of Response and Restoration provided scientific support including overflight observation of the spill, information on fate and effects of the crude oil, and forecasts of surface oil movement. Operational forecasting of surface oil movement is critical to the response for planning, allocation of resources, and timely direction of response assets. This particular spill presented several unique challenges for oil trajectory modeling.
One challenge was the location of the release. Oil flowed down the beach and entered the ocean through the surf zone. The surf zone, the nearshore region of breaking waves, is dynamically distinct from the coastal ocean. Transport and dispersion within the surf zone is influenced by much smaller scale dynamic features (e.g., longshore currents, rip currents). Oil is moved alongshore by breaking-wave and wind-driven currents, and may be ejected offshore from the surf zone to the inner shelf by transient rip currents and/or offshore winds. Although detailed surf zone modeling is possible, it requires high resolution and inclusion of wave-current interactions. The surf zone is not typically resolved by the regional coastal ocean models that are used to drive transport in oil spill modeling. This is also a limitation for trajectory modeling of oil spills that originate offshore. In this case, it limits the ability to provide detailed information on variability in shoreline oiling below the model grid scale (typically >1 km). However, in the case of the Refugio spill, the origination of the spill on the beach affected the initialization of the model, requiring assumptions to be made about the initial transport and footprint of the floating oil.
Uncertainty in the oil footprint used in the model initialization can be reduced by overflight surveys, which map the observed floating oil. NOAA overflights began the morning of 21 May, two days after the initial release, which occurred in an area known for its abundant natural oil seeps. The Coal Oil Point area is home to seeps that release on the order of 100 barrels of oil per day and are among the most active in the world [Hornafius et al., 1999]. It is not generally possible for aerial observers to determine if observed floating oil is from a spill or a natural source, unless the oil can be clearly traced back to an origin point. In the Refugio spill, the oil from the local natural seeps and the leaking pipeline both originated from the same geologic formation so even their chemical makeup was similar, requiring sophisticated chemical analyses to differentiate the two.
In spite of these challenges, oil spill trajectory forecasts were produced and utilized by the response to guide operations from the first day of the spill through 26 May. These forecasts identified with reasonable accuracy the heaviest regions of shoreline oiling along the coast adjacent to the spill site and extending several miles. However, several incidents of above-background levels of tarball deposition also occurred on beaches remote from the spill site, e.g., along the Ventura coastline, in Santa Monica Bay, and in the Channel Islands. This prompted a hindcast analysis to examine regional transport pathways and time scales of transport to examine the likelihood of these incidents arising from the Refugio spill. This hindcast modeling was further refined as part of the Natural Resource Damage Assessment following the spill response.
Movement of oil was modeled using the General NOAA Operational Modeling Environment (GNOME) [Zelenke et al., 2012]. GNOME is an oil spill trajectory model in which the surface oil is divided into a large number of small particles of equal mass that move under the influence of surface ocean currents, wind drift, and horizontal mixing. GNOME also includes algorithms that simulate surface oil weathering, e.g., evaporation and dispersion.
Oil spill trajectory models rely on hydrodynamic model forecasts and/or observations of winds and currents as inputs. Real-time forecasting during spill response necessitates reliance on meteorological and ocean forecast model results, increasing the uncertainty associated with the oil trajectories. In contrast, hindcast modeling can incorporate observational data from the time of the spill or reanalysis products (models that are re-run with available observational data assimilated).
For the hindcast analysis presented in this paper, surface currents were obtained from coastal radar data operated with funding from the Southern California Coastal Ocean Observing System (sccoos.org). Coastal radars measure surface currents in the coastal ocean with coverage throughout the Southern California Bight at 6-km resolution (Figure 2; a higher resolution of 2 km is available for the region within the Santa Barbara Channel).These data resolve the mesoscale features of the surface circulation (e.g., counter-clockwise Santa Barbara Eddy, Southern California Countercurrent [Hickey, 1992]).
Wind forcing was obtained from the Coupled Ocean/Atmosphere Mesoscale Prediction System (COAMPS®), developed and run by the Naval Research Laboratory in Monterey, CA (http://www.nrlmry.navy.mil/coamps-web/web/home). COAMPS is a numerical model used for wind nowcasts and forecasts. The implementation of COAMPS used in this study has a spatial resolution of 4 km and covers the coastal ocean from Oregon to Mexico. COAMPS was run as a forecast with re-initialization every 12 hours with available observations interpolated together with the previous model forecast. Atmospheric fields were then projected forward in time.
In spill trajectory models, it is common to combine a number of physical processes related to wind forcing (e.g., Stokes drift, surface drift, Langmuir circulation) into a wind-drift factor [Galt, 1994]. This has been determined experimentally to be ~3–4% of the wind speed for fresh oil in light winds without breaking waves [Reed et al., 1994]. As the oil weathers and/or if wind speed increases, the oil may spend a significant portion of time away from the surface and out of the influence of many of the processes associated with the wind forcing, and the average drift factor may be much lower. In general, this parameterization is a very useful approach but requires observational feedback during spill events [Galt, 1994]. GNOME allows the user to specify a range of values for the wind drift along with a persistence time scale, simulating the time-varying windage as the wind and wave conditions are not generally spatially or temporally constant. A base-case model run used a range of windage values from 1–4% with a persistence time scale of 15 minutes, i.e., individual particles are randomly re-assigned a windage value between 1–4% every 15 minutes. Sensitivity studies were performed using reduced (1–2%) and increased (3–4%) windage values.
Turbulent diffusive processes that spread spills horizontally are simulated in GNOME by a random walk. A diffusion coefficient of 1 m2s−1 was used to calculate random step lengths in the x- and y-directions from a uniform distribution. The current version of GNOME does not allow for spatial variability in the horizontal diffusion, so this results in a uniform spreading of the particles over time.
As discussed previously, the 6-km resolution regional coastal radar derived surface currents do not include data in the surf zone where the oil first entered the ocean. Due to this limitation, rather than initiate the spill as a point source, observations during the first 24 hours of the spill were used to initialize floating oil nearshore. Although detailed NOAA overflight surveys did not begin until 21 May, aerial photographs indicated some of the floating oil began moving offshore the afternoon of 19 May, coincident with the development of an offshore wind in the late afternoon. U.S. Coast Guard personnel flying over the site observed a 2-mile alongshore extent of shoreline oiling at 5:30 PM on 19 May and observed some floating oil remaining in the surf zone and extending offshore during a mid-morning overflight on 20 May.
Based on these limited observations, particles were initialized in the GNOME model within a polygon in a nearshore area extending out to ~1/3 km off the coastline with an east-west extent similar to the 19 May overflight. For each simulation, ~100,000 particles were released over a 24-hour period beginning 19 May 5:30 PM. A linear decay rate in the number of particles released per timestep was applied.
Without a detailed surf zone modeling component or a quantitative estimate on how much oil beached along the shoreline adjacent to Refugio Beach, we did not think it reasonable to estimate how much oil may have moved offshore versus oil that remained in the surf zone and beached in the near-field. Model results are therefore reported as relative particle densities rather than oil concentrations. For the same reason, oil weathering is not included in this analysis and the particles are conservative with no loss processes.
SHORELINE OIL OBSERVATIONS
Shoreline Cleanup Assessment Technique (SCAT) teams began surveying the coastline adjacent to the spill site very early in the response. SCAT survey data are presented in Figure 3. This map includes data collected over multiple days: the coloring of shoreline segments represents the maximum level of oiling observed on any survey. For a temporal perspective, on 20 May, SCAT teams reported observing oiled shoreline extending approximately 7 miles from Arroyo Hondo Preserve (western extent) to El Capitan State Beach (eastern extent). By 23 May, SCAT maps indicated moderate shoreline oiling extending several miles further east of El Capitan and also variable degrees of shoreline oiling around Coal Oil Point and University of California, Santa Barbara.
High densities of tarball deposition were subsequently reported by the public at Manhattan Beach in southwestern Los Angeles County on May 28 and in Ventura County between Carpinteria and Oxnard around the same time. Surveys were also conducted in these areas and reported light to moderate oiling conditions. An atypical tarball event was also reported by U.S. Fish and Wildlife in the Channel Islands between 1–7 June.
This region has a level of background shoreline oiling due to the active natural seep field, which is an order of magnitude higher in summer than winter [Del Sontro et al., 2007, Lorenson et al., 2009]. This complicates interpretation of the semi-quantitative SCAT data. The map shown in Figure 3 certainly incorporates both spill oil and natural seep oil. In the first few days of surveys described previously, the proximity to the source and nature of the oiling indicated much of this oiling was likely due to the pipeline release. As the observations become more remote in time and space from the spill origin, sophisticated chemical analyses are required to differentiate between the two possible sources. The model results described in the next section can also aid in describing potential connectivity between the spill site and the impacted regions.
Snapshots of relative particle densities at various model times are shown in Figure 4 for the base-case run described previously (windage values ranging from 1–4%). In this simulation, model particles move offshore into the Santa Barbara Channel and enter the counter-clockwise Santa Barbara Eddy on 20–21 May. Very few particles make landfall on the Channel Islands. Instead, particles move east, making landfall along the Ventura coastline by 25 May. The remaining floating particles move southeast, with some portion advected eastward into Santa Monica Bay and making landfall by 29 May, approximately one day later than the observed event. The highest beached particle densities occur near Refugio Beach and in the eastern Santa Barbara Channel.
A sensitivity study with the wind influence reduced to 1–2% of the wind speed (Figure 5) demonstrates the importance of westerly winds prevalent in the Santa Barbara Channel in reducing landfall on the Channel Islands. In this lower windage simulation, particles make landfall initially on the north facing shorelines of Santa Cruz and Anacapa Islands within several days of the release. Shoreline impacts still occur in Ventura County but are delayed by several days. However, in this case, no particles enter Santa Monica Bay.
A sensitivity study with a larger wind influence, a 3–4% range, was also conducted (not shown). In this case, particles move eastward in Santa Barbara Channel much quicker, making landfall in Ventura County by 23 May. The impacts in Santa Monica Bay are also magnified and occur by 27 May, preceding observations slightly.
SUMMARY AND DISCUSSION
Following the Refugio Beach Oil Spill on 19 May 2015, several incidents of increased tarball accumulation occurred on beaches remote from the spill site. This study examined the movement of modeled particles simulating floating oil that originated nearshore of Refugio Beach, with the goal of elucidating regional transport pathways and time scales of transport.
The model simulations show particles making landfall in southern Santa Monica Bay and Ventura County in a time frame approximately agreeing with observations. In the base-case scenario, a range of values of 1–4% were used for the model windage parameter, resulting in downwind movement averaging about 2.5% of the wind speed. This resulted in initial landfall impacts in Ventura County by 25 May and at Manhattan Beach by 29 May. Interestingly, the majority of public reporting of tarballs were around the same time for both sites, with Santa Monica Bay (28 May) actually leading Ventura slightly. Based on the modeling studies alone, this seems confounding—however, the modeling studies only considered movement of oil originating from the spill. Oil observations from overflights indicated very active natural seep activity throughout the time period of surveys, which under the right environmental conditions, could also manifest as notable tarball events on remote beaches. We would expect tarballs that arrived on beaches in both these locations during this time frame to likely contain a mixture of spill and seep oil.
Another notable event was the increased tarball accumulation observed in the Channel Islands beginning around 1 June. By this time, the modeling simulations suggest most of the floating oil had moved to the southeast or already beached. A NOAA overflight on 29 May saw little oil remaining in nearshore waters off Refugio Beach, but traced an approximately 10 mile long band of dark oil several miles offshore of the spill location back towards the Coal Oil Point seep field (Figure 6). By this time, the position of the Santa Barbara Eddy as observed by the coastal radar data had shifted westward in the Santa Barbara Channel. When combined with prevailing wind patterns, this shift in the position of the eddy made transport from the spill region to the Channel Islands much more likely during this time period. These data suggest the Channel Islands event may have been a result of natural seep activity rather than from the Refugio spill.
Finally, due to the lack of data in the surf zone region, the model likely underestimates shoreline oiling in the “near-field,” i.e., the coastline adjacent to Refugio Beach extending several miles. Observations indicate a significant portion of oil initially beached along this shoreline (Figure 3). The volume that beached in this near-field region was likely a function of many processes that were not included in this regional scale modeling effort, for example, variability in local onshore/offshore winds (sea breeze), the phase of tides, the strength of longshore currents, and the holding capacity and orientation of the shoreline. Higher resolution re-analysis ensemble wind products could better capture the temporal and spatial variability in the wind patterns. This, combined with high resolution surf zone modeling, could potentially provide a better prediction of alongshore transport in the near-field, including exchange between the surf zone and inner shelf, and hence provide a better prediction of shoreline oiling probabilities along this section of coastline.
Thanks to Chris Barker and Glen Watabayashi (NOAA Office of Response and Restoration) for input on the modeling approach and assumptions. Libe Washburn and Brian Emery at University of California, Santa Barbara (UCSB) kindly provided information about coastal radar data in the region operated through NOAA SCCOOS.