The Oil Spill Recovery Institute: Past, Present, and Future Directions (2003)

The central element of the present OSRI research and development effort is the Prince William Sound Nowcast/Forecast system (PWSNC/FC). As shown in Table 5-1, the Predictive Ecology program provided about $533,000 to Nowcast/Forecast projects from FY98 to early FY02, out of about $1.6 million total program funds. Thus, 33 percent of predictive ecology funding went to support direct Nowcast/Forecast work. As shown in Table 6-1, the Applied Technology program provided about $763,000 to Nowcast/Forecast projects in the same period, out of about $1.8 million total program funds, or 43 percent. This may be an underestimate of the overall financial emphasis on Nowcast/Forecast: for example, it does not capture staff time (especially that of the technology coordinator, who is actively involved in running the model; funds spent since inception of this position in FY99 to early FY02 totaled $253,989, or an additional 14 percent of the applied technology budget). In addition, it does not reflect how other projects might be providing information that ultimately supports the model. Clearly, however, OSRI is devoting a significant proportion of its total efforts to modeling and its supporting components.

The PWS-NC/FC is a follow-on of a model system developed under the Sound Ecosystem Assessment (SEA) research program, which was funded by the Exxon Valdez Oil Spill Trustee Council (EVOSTC). Under the SEA program, a physical-biological model of PWS was developed with the goal of better understanding the physical-biological coupling and changes in the PWS ecosystem. This scientific understanding is needed to evaluate the potential impacts of oil spills on the PWS ecosystem, as well

as other subarctic marine ecosystems. Thus, such a research and development effort is a fitting part of the mission of and was supported by OSRI.

As the SEA program was coming to a close, OSRI embarked on a new phase of model development. Whereas the focus of the SEA physical-biological model was to develop an understanding of processes and physical-biological coupling using hindcasts of specific periods, which were compared to field-collected data for verification, the present focus of the OSRI modeling program is in real-time nowcasting and forecasting on a daily basis. The BAA for this OSRI project (issued in early 1998) was brief, but requested the development of a “nowcast/forecast capability of currents and conditions that are relevant to the risk and costs of oil spills.” The contract was awarded to the team that had developed the SEA model, which designed its proposal to include a real-time data acquisition and forecasting goal.

Central to the SEA physical-biological model system was a hydrodynamic model that predicts currents, temperature, and salinity distribu-tions in four dimensions (3-dimensional in space and dynamic in time). The hydrodynamic model (Mooers and Wang, 1998) is an implementation of the Princeton Ocean Model (POM), widely used worldwide for such applications. The OSRI PWS-NC/FC POM application is undergoing continuing development and is being run in a near real-time nowcasting and forecasting mode at the University of Miami Rosenstiel School of Marine and Atmospheric Science (RSMAS, C. Mooers, PI).

OSRI recently added the existing Oil Spill Contingency and Response (OSCAR) model to the PWS-NC/FC. OSCAR was developed by SINTEF in Norway (Reed et al., 1995, 2000). OSRI purchased a license to OSCAR and houses a working copy of the model in the PWSSC, with the plan that OSRI staff and other investigators will run the OSCAR model in Cordova using data provided by the RSMAS POM through a web linkage.

OSRI has awarded a grant to the University of Alaska Anchorage (UAF, P. Olsson, PI) to develop a mesoscale atmospheric model that will produce a spatially and time-varying wind field to input to the hydrodynamic model at RSMAS. The existing Regional Atmospheric Modeling System (RAMS) model will be applied to PWS. The goal of the atmospheric modeling is to capture the orographic effects of the mountainous terrain of the islands and coastline around PWS, which is agreed to be the primary cause of the high variability in wind speed and direction across PWS at any given time. Describing this wind field variability in space and time is critical to accurate simulations of the hydrodynamics, as recognized by OSRI.

The conceptual design of the PWS-NC/FC is that the atmospheric model is to operate in real-time (a “nowcast/forecast”) and pass wind fields to the hydrodynamic model. The hydrodynamic model will in turn

prepare a Nowcast/Forecast simulation and pass both the wind and current predictions to the oil spill model. Both the atmospheric and hydrodynamic models will be automated to run every 6 hours, 24 hours a day. The oil spill model may be run at any time at OSRI to provide near real-time predictions of oil movements and fates. The running of the oil spill model is not automated, and the intention (for the near term) is that it be run by OSRI staff as needed.

The PWS-NC/FC is conceptualized to include biological and ecosystem model components, but these components are not yet developed. Whereas the SEA physical-biological model of PWS contained zooplankton and larval herring components, the principal investigators of those efforts are no longer involved in the OSRI effort. Thus, the biological modeling component is conceived as a future development effort. The present focus attempts to overlay the oil spill model’s output on maps of fish species distributions developed using acoustical techniques.

Given the extent of its financial commitment, OSRI clearly sees its modeling activities as critical elements of both the Applied Technology and Predictive Ecology programs. As discussed below, the committee has reservations about the purpose and perceived primary use of the modeling program as presently configured.

As currently envisioned, the models are planned for use in a predictive mode at all times as well as during a spill, and it is on this basis that they are classified as components of the Applied Technology program. For example, planned uses include providing real-time spill trajectories and 48-hour forecasts to guide oil-spill dispersant use decisions. That is, if the models demonstrate that the projected trajectory will take oil away from modeled concentrations of subsurface organisms, then it may be appropriate to use dispersants. Likewise, if the modeled subsurface dispersed oil concentrations are below some threshold value, then that information could also be used in making dispersant use decisions.

The problems with this thinking are two-fold. The first deals with the accuracy and validity of the model predictions and the fact that the models have not been developed to the point that they can be reliably used in this mode. At the present time, the existing modeling predictions are unreliable because they are being based on either a single wind record from a buoy located in the center of the Prince William Sound or 30-km-gridded atmospheric model forecasts of winds at the water surface. The surface current and oil trajectory is highly dependent on the accuracy of the wind field, and the model system lacks the necessary detailed wind information to support it. The value of the oil model output as a spill

response tool is questionable also because it lacks a good validation of the hydrodynamic and wind models with measurements from different parts of PWS. OSRI and the principal investigators recognize these and other limitations, and plans are underway to develop more realistic wind forcing and increase the accuracy of model simulations. However, the ability to make accurate forecasts in real time is many years off, and the funding necessary to develop such a capability will be considerable.

One of the most important modules in the physical modeling program is proper validation. Based on the information provided, the field data collection programs do not appear to provide adequate real-world data to verify model predictions. If OSRI’s long-term goal is to have their model used if future oil spills occur, they will have to show that the model accurately predicts the conditions that would occur within PWS.

An additional problem with the OSRI Nowcast/Forecast system deals with the acceptance of the model and OSRI’s role in providing spill forecasts within the response community. Potential users in the response community include: the U.S. Coast Guard (USCG), the National Oceanic and Atmospheric Administration (NOAA), Alaska Department of Environmental Conservation, the responsible party, and other stakeholders involved in dispersant use decisions and other spill response issues (in situ burning, boom placement, skimmer deployment, etc). Legal authority for oil spill response offshore in U.S. waters has been given to the USCG, which relies on recommendations from NOAA for oil trajectory predictions when an oil spill occurs. NOAA utilizes its own modeling group and in-house models during all spill-response activities within U.S. coastal waters. Thus, this aspect of the OSRI Nowcast/Forecast system is redundant. However, there is an opportunity for OSRI to provide input to the response community by developing a better understanding of relevant processes, particularly of atmospheric and hydrodynamic transport and of the fates and effects processes involved.

In the crisis atmosphere following an oil spill, most participants recognize that a plurality of models would lead to different trajectory predictions, and these would add to the uncertainty and arguments about appropriate response measures. There is also an issue of liability if the OSRI model predicts that the oil will go in one direction with little harm or impact when the oil in fact goes in a different direction and causes unforeseen damages. The same problems of liability also apply to the NOAA model, but as an agency of the U.S. government, they are less likely to be a target of a lawsuit if their model predictions are inaccurate. Alyeska Pipeline Services Company (Alyeska) also has a similar model system (Alyeska Tactical Oil spill Model, ATOM), which it uses for contingency planning, drills, and spill impact evaluations, but it would defer to NOAA’s modeling predictions in the event of a real spill response.

During a spill response, the initial deployment of equipment will be under the control of the Federal On-Scene Coordinator (FOSC) and the spill response coops. The strong probability is that they will depend on modeling input from NOAA’s Seattle office, although NOAA may use information from other entities such as OSRI, as it is NOAA’s job to integrate all available information. It is important to note that in real spill situations, there are always limitations on response equipment and personnel. Prioritization of response efforts is of primary importance. Unless strongly convinced otherwise, the FOSC is going to defer to NOAA’s advice for setting the response priorities. OSRI would need to coordinate extensively with NOAA for its efforts to be useful in a spill response situation.

These same types of considerations and limitations relate to the use of the commercial spill trajectory program (OSCAR) purchased by OSRI. This can be a useful planning tool, and is also useful in developing and implementing oil spill response drills. In a real-world spill response, however, it would be questionable whether the USCG would defer to OSRI’s predictions over those provided by NOAA through its science support coordinator, at least given the present status of the atmospheric and hydrodynamic models. In addition, NOAA’s predictions are described as a probability distribution, reflecting the uncertainty in the wind and current forecasts input to the oil trajectory model. OSCAR’s output is a single (deterministic) trajectory. Uncertainty in this forecast is not easily described without performing a sensitivity analysis of multiple runs manually implemented. As configured, it is doubtful that such uncertainty would be analyzed by OSRI and passed on to responders during a spill response in real time, as would be needed for OSRI to contribute constructively to a response.

At the present time, the NC/FC system is receiving real-time data from the data buoy in PWS, transmitting it to the University of Miami, where the nowcast/forecast hydrodynamic model is run, and then sending the data back to Cordova so that they can be coupled with the OSCAR model for making predictive trajectory runs. This is being done automatically every six hours. The value of having this information transferred back and forth on a daily basis is questionable, and maintaining and archiving data use funds that could be devoted to research. There has been a suggestion that this has been a temporary demonstration phase to show that they can transmit the data to Miami for processing and return on short notice. To continue this in perpetuity is of questionable value.

In summary, running the Nowcast/Forecast system is resource intensive from both a data and personnel viewpoint, and the OSRI legislative mandate does not justify this dedication of resources. Here are some rel-

evant considerations to illustrate the issues related to real-time operational use:

• Who is responsible for the day-to-day operation of the ocean model now and in the future?

• How is this coordinated and adjudicated with other models that the government uses?

• What is the accepted, approved, and legally determined role of the results of this model?

• Who assumes liability for decisions based on the results of this model?

• Who will configure OSCAR on a real-time basis (type of oil, environmental parameters, etc.)?

• Who will run RAMS on a real-time basis?

• What real-time data networks provide input for the ocean, atmosphere, and fate models?

• Who will pay overtime for the operational infrastructure and maintenance?

Rather than focus the modeling in a real-time response mode, a much more appropriate focus for OSRI’s modeling effort is in support of scientific purposes, such as contingency planning, training and community outreach, ecological risk assessment, understanding the Prince William Sound ecosystem, identifying data gaps, and designing experiments to obtain additional data on ecosystem function. Predictive forecasts are not needed for any of these purposes; rather, the model would be used in a hindcast mode, where wind and current conditions and oil spill behavior can be validated, thereby adding credibility to other model predictions (such as water column concentrations).

An example application is the planned dispersant impact analysis included in OSRI’s current project list. However, while this is an appropriate research project for OSRI funding, it is not possible to evaluate the implications of changing (potentially increased) oil concentrations in the water column with application of dispersants without examining exposure and toxicity in more detail than is currently possible with OSCAR. The issue is the trade-off of floating and shoreline oiling of wildlife and habitats versus impacts on water column organisms. Chemical dispersants will increase concentrations in the water column. Whether this is significantly toxic is the question to resolve the dispersant use issue. OSRI plans to perform the dispersant scenarios in-house, without a competitive

proposal process (G.L. Thomas, Oil Spill Recovery Institute. Personal communications, April 11, 2002.).

Because of the assumptions inherent in the construction of any model, an oil spill model is not an exact replica of the real world. The role of model building by OSRI should be viewed as advancing the state of models for oil spill response and for the prediction of ecological consequences. OSRI can make a contribution to oil spill response modeling by improving the knowledge base used to develop model algorithms. A second important use of OSRI models is to predict the ecological consequences of an oil spill, or of different possible oil spill scenarios. Model development should therefore include ecological effects.

Below is a summary of capabilities of the PWS NC/FC model components, the scientific validity and accuracy of the model simulations, and the degree to which the models represent or are pushing forward the state of the art.

The hydrodynamic model in the SEA program (Mooers and Wang, 1998; Wang, 2001) was designed to be run as a hindcast for specific periods of time in order to evaluate physical-biological coupling and ecosystem dynamics. The model was forced with observational wind data from several stations, tidal data from three tidal stations, and freshwater runoff calculated from annual rainfall uniformly spread along the shoreline. The boundary was defined at Hinchinbrook Entrance (HE) and Montague Strait (MS) and forced with measured data at those locations. This implementation provided considerable understanding of the circulation dynamics within PWS.

The POM implementation under the OSRI program has been expanded to include the Alaskan Shelf from Yakatat Bay to Kennedy Entrance in Shelikof Strait. Earlier modeling and measurements at HE and MS showed inflow/outflow to be complex. GLOBEC investigations have identified eddies (20-200 km) that come up against the continental margin and onto the shelf, which have a role in exchanging waters in and out of PWS. There is large variability from year to year, and it is not always clear what is upstream or downstream on the shelf. On-shelf exchange between the coastal shelf current and PWS is complex and seasonally dependent.

As the nontidal circulation is driven by shelf processes, moving the boundary out to sea allows the longer-term circulation to be addressed.

This approach makes sense, as the only alternative is to force the model with high-resolution (vertically and temporally) measurement data at the entrances, and only one acoustic doppler current profiler is available at HE. However, intensive measurements could be made at the two entrances for specific events, which could be used to verify the dynamics within PWS. While efforts at validation have been made, there are still considerable differences and unknowns in the dynamics within PWS that should be addressed (such as circulation driven by freshwater inflow and local wind dynamics).

Only one wind station in central PWS is online in the NC/FC system at present (although other wind stations have existed and will be coming online), and any hydrodynamic model needs better forcing, particularly for winds. Winds vary tremendously across PWS because of orographic steering. There are important mesoscale features in the winds (gap winds, barrier jets) that will influence circulation and oil transport. The plan is to drive the hydrodynamic and oil models with gridded winds produced by a mesoscale atmospheric model.

Freshwater inflow also has an influence on circulation (Mooers and Wang, 1998), however including this forcing appears not to be in plans for immediate future for lack of sufficient data. A sensitivity analysis could be performed to determine how sensitive the model is to this forcing.

The recent hydrodynamic model validation (Bang et al., submitted) is comprehensive and is to be commended. Given the complexities in the underlying physical processes, periodic validation is necessary to calibrate the model and understand its strengths and limitations. It is not surprising that forcing functions have a dominant effect on the model predictions. One of the major weaknesses in the circulation model has been the modeling of through-flow from the Alaska Shelf. In particular, the assumption of equal and opposite volumetric transport at HE and MS is now known to be erroneous and will require closer attention in future modeling. Subsequent work has shown that there may be significant outflow through HE, and this needs to be incorporated in future modeling (Bang and Mooers, in press). Furthermore, (as noted above) the impact of coastal orography on the wind flow needs to be investigated and is likely to have a significant effect on any spill trajectory calculations. Finally, discrepancies in the predictions, mainly at the mesoscale and smaller scale compared with the basin scale, seem to reinforce the need for modeling the scale dependencies in the forcing functions using appropriate spatial models. The model will need further verification after the forcing is improved.

The atmospheric model component has only recently been funded, and is not yet implemented in the NC/FC system. It will be an application of an existing mesoscale model, RAMS, which has been previously applied in other locations by the PI. The funded proposal and PI were favorably rated by both reviewers as technically sound, with the reservation that a postdoc to work on the project was not identified at the time of funding. This is an appropriate approach, as the wind forcing needs to be spatially variable at the scale being addressed by the mesoscale model, because of the high mountainous terrain surrounding PWS.

The entire domain will cover much of Alaska and the adjacent North Pacific, using a nested grid system of 64-km resolution in the outer domain, to 16-km resolution within it, and to 4-km in PWS. The nested design will invoke boundary issues, which will need to be resolved. In addition, the 4-km resolution within PWS may not be fine enough for the narrower passages between mountainous terrain and in passes. However, overall it is expected that the atmospheric forcing predicted by the RAMS will improve the hydrodynamics and oil spill simulations.

OSCAR (Reed, 2002) is a fairly comprehensive model for spill trajectory and fates calculations, accounting for most major physical and chemical processes and using algorithms that are state of the art and used in other similar oil spill fates models (e.g., ASCE, 1996; French et al., 1999). Fates processes included are advection, spreading, evaporation, emulsification, natural dispersion, dissolution, adsorption of soluble components and sedimentation, volatilization from the water column, degradation, stranding on shorelines, and such responses as booming, removal, and chemical dispersant application.

One important process, which is not included in OSCAR but should be, is the interaction of oil with suspended particulate matter and sedimentation of such material. This process has been shown to be important in a number of spills, and certainly could be important in much of Cook Inlet if modeling were expanded into that region, and in PWS at certain times of the year when stream, river, and glacier input are at their highest. Research is available for the development of an algorithm (Payne et al., 1987, 1989).

For the dispersant model, OSCAR uses an empirical dispersion rate based on limited field trials. Thus, there are many hypothetical functions in the algorithm, and much uncertainty in the results. Clearly more empirical data are needed to develop appropriate algorithms. Meanwhile,

sensitivity analysis on the model would elucidate the uncertainty in present model results.

The focus of OSCAR is to model spill response scenarios and the fate of the oil resulting from such (Reed et al., 1995). The interface is designed to allow the user to play “war games” with response scenarios. Each simulation is set up individually, with databases of model input and a number of assumptions assigned by users. Thus, users need to be well-trained such that they can make appropriate selections of these uncertain inputs and interpret the results. Eventually, once the models are adequate, they should allow comparison to what is known from the Exxon Valdez spill. It will take some time (days to weeks) to run multiple scenarios with the present interface design, and these kinds of sensitivity analyses to varying model inputs will not be possible to run in a real-time response situation.

OSCAR does not account for the uncertainty in the velocity field and the resulting impact on the spill trajectory and travel times. A stochastic treatment of the velocity field or some kind of error propagation algorithm can potentially resolve this. The trajectory predictions will only be as good as the imposed velocity field and other environmental inputs. Given the sparse forcing data for the hydrodynamic model and wind field and spatial variations and uncertainties, it is impossible to describe such velocity distribution in deterministic detail. Clearly, a stochastic treatment is more appropriate as is routinely done in subsurface flow and transport in petroleum reservoirs and aquifers. The stochastic approach has also been applied to oil spill impact modeling (French et al., 1999; French McCay et al., 2002). Galt (1995) and Lehr et al. (2000) make the point that uncertainty must be taken into account in model simulations for oil spill response. The uncertainty quantification aspect is missing in the Nowcast/Forecast modeling exercise. This can lead to a false sense of security or overconfidence in modeling.

The OSCAR model seems to have been validated using a variety of synthetic problems by comparing the results with analytic solutions. Some limited validation of the trajectory and concentrations using field data has also been presented, although statistical analysis of goodness of fit is not performed.

OSCAR utilizes the POM current predictions for its simulations. Current data transfer from POM is done via an ftp (file transfer protocol) site, using an automated download onto an OSCAR-dedicated computer at OSRI. Presently, the wind data used is from a single record from a station in the center of PWS. The plan is to use the gridded predicted winds from the mesoscale atmospheric model, passed through the data transfer file from the POM.

The POM runs every six hours for a forecast of 48 hours. The current

vectors are saved at simulation intervals of one hour on a 1-km grid (with 11 vertical layers in sigma coordinates) and downloaded to OSRI in Cordova every six hours. The download takes about 30 minutes and involves 20-30 megabytes of data per file. The data files are archived by date and time in a sequential set. The model automatically uses the latest hydrodynamic data available from the POM, stepping through applicable dated files sequentially.

Thus, OSCAR can be run using the POM input for as long as the POM forecast, which is now two days into the future. To run a hindcast, one needs to have the appropriate POM files for the period of interest. OSRI is saving a library of POM outputs for the present year, which might be used. The University of Miami group is preparing to make month-long files from the archives to use for hindcasts. However, there are no plans to run past years or an Exxon Valdez oil spill simulation period for potentially validating the model. Such an exercise would build confidence in the model system, as there is much information available on this oil spill.

The exposure assessment model in OSCAR, as presently included in the NC/FC system, is highly simplified compared with the rigor involved in the transport and fates calculations. Exposure is evaluated as an index of oil thickness or concentration greater than a user-selected threshold, integrated and averaged within a defined polygon area. For surface-floating oil the index is m²-hrs and for concentrations in the water, the index is ppb-hrs (presumably averaged over the water column). While mapping of such exposure indexes can show where resources are relatively more exposed and can be used to compare runs in a relative sense, it will be difficult to interpret such information in a response decision regarding using dispersants. The questions are (1) whether applying chemical dispersants will increase impacts on water column organisms, and (2) whether there will be a balancing decrease in impacts on wildlife and shoreline resources. The m²-hrs of surface oiling needs to be translated into percentages of wildlife or numbers of birds oiled. The ppb-hrs of concentration exposure is highly subject to the volume and time over which the integration and averaging occurs. In addition, the ppb-hrs needs to be interpreted as an impact on water column organisms, which is a function of concentration of each chemical in the oil mixture, exposure time, and number of organisms exposed (i.e., a biological exposure model is needed). This kind of analysis will be needed by decision makers planning or during a response.

The original BAA for the biological effects component requested “technologies [that] demonstrate three dimensional trajectories as well as

the resulting physical and biological environmental impact of dispersed and nondispersed oil in Arctic and subarctic marine environments.” OSCAR does not yet have the biological impact component developed enough to meaningfully perform a comparison of the impacts of dispersed and nondispersed oil spills. At the moment the analysis would be limited to mapping of distributions of surface oil and subsurface concentrations in paired scenarios.

Resource mapping into an ArcView (Environmental Systems Research Institute software) geographic information system (GIS) (named RARE = Resources at Risk, which may be imported and overlaid on model results in OSCAR) using environmental sensitivity (ESI) maps and other information, is planned but not yet linked to the model. Meetings to discuss this plan were held in May 2002 in Trondheim, Norway. Also, consideration has apparently not been made of using the GRD (Graphical Resource Database) for PWS, which has been developed cooperatively by Alaskan and federal agencies and the Alyeska Pipeline Company over the last decade. The GRD is the intended coordinated resources-at-risk GIS database for PWS developed by the Regional Response Team (RRT). The GRD is used by Alyeska’s oil spill model ATOM. OSRI should be working with the RRT, as the GRD could be read by the model system and is kept up to date with the latest information. Creating a new RARE database is a redundant exercise that could result in conflicting information. If OSRI wants to advise on spill response, they need to be more coordinated with the RRT.

In addition, the distributions of organisms that might be affected by a spill are highly variable in space and time. Static mapping of resource locations in a GIS map will be of limited use to an NRDA. More understanding of the coupled physical-biological system is needed for such purposes.

As part of the SEA program, there was a linkage between physical and biological modeling components. Plankton models (Eslinger et al., 2001; Jin et al., submitted) were run with the 3-D POM implementation by Mooers and Wang (1998). An understanding of the spring bloom temporal and spatial dynamics was gained that explained the focusing of herring populations in areas and times of plankton abundance. There were also bioenergetic models of pink salmon and herring by V. Patrick, D. Mason, and others. However, there has not been any incorporation of this or any other biological modeling into the current NC/FC modeling effort. It is unfortunate that the understanding gained in the SEA program will not be carried forward into the OSRI model system.

The models supported by OSRI are state of the art, but they are not unique models. The hydrodynamic and atmospheric models are applications of existing model code commonly used in other locations. The POM for PWS seems now to exist as two implementations: (1) the present OSRI model being operated at the University of Miami, and (2) the former SEA version of the model, now at the University of Alaska Fairbanks (Wang, 2001). The biological modeling component begun in SEA was lost, and a new effort will be required to bring this type of capability back to the OSRI system.

The OSCAR model is similar to the fates portion of the ATOM model, which has been in use by Alyeska Pipeline Company since 1990. ATOM is updated periodically and is equivalent to Applied Science Associates’ SIMAP, a further development of the fates and biological effects portions of the NRDA model (French et al., 1996). In addition, NOAA HAZMAT has its own model GNOME applied in PWS; thus, there are redundant model systems in place now for PWS, at least in regards to oil fates. OSRI should rethink how its model system is to be used and whether it should in fact be a real-time nowcasting and forecasting system or a research tool. Also, while there are redundant oil fates models, there is a lack of hydrodynamic and atmospheric models, and these areas provide opportunities for OSRI to contribute.

Although the committee disagrees with the current focus of the OSRI modeling efforts on maintaining a real-time simulation capability, it sees good opportunities to shift the emphasis. In general, we recommend that OSRI focus on development of understanding and new algorithms that would improve the state of the art in oil spill modeling. Most of the algorithms used in oil spill models (generally) are based on or are similar to those developed more than 20 years ago by Mackay et al. (1980 and previous work). OSRI is in a unique position to fund algorithm development research, which is not generally funded by industry. The following are examples of topics that could be researched.

The POM is designed and set up to simulate the water circulation in subtidal waters, and does not handle the wetting and drying dynamics of intertidal flats. While such flats are limited in area in PWS, Cook Inlet and the Copper River Delta have vast flats. OSRI could make a contribution

by funding development of modeling approaches to tidal wet-dry boundaries. Hydrodynamic models typically do not address the tidal wet-dry problem, and assume a shoreline of unchanging location.

The complexities of the COZOIL model (Reed et al., 1988, 1989) have not been included in OSCAR, likely in part because it requires considerable data on the shoreline characteristics. An algorithm of intermediate complexity to COZOIL and the simplification that is presently used in OSCAR is desirable to better estimate how much oil is retained by a shore segment and the dynamics as the tide moves in and out. Oil is known to combine with sediments and become incorporated in both intertidal and nearshore subtidal sediments. Understanding these dynamics is essential to predicting long-term impacts of oil spills on shoreline and nearshore biota.

Additional areas of research where OSRI funds could be used to elucidate and provide the basis of new and better oil fate algorithms include the influence of oil on wave height and transport by Langmuir circulation. Langmuir circulation was the subject of a recent NOAA workshop, because it is important to spreading and water-in-oil emulsification, as oil is concentrated in convergence lines.

Spreading (which controls evaporation and entrainment) has been modeled for more than 20 years using algorithms that are only appropriate for calm water surface conditions. The algorithm is augmented by slick drift, oil droplet entrainment, and resurfacing, but significant data gaps exist in our understanding of oil-slick-spreading behavior.

Evaporation from a thin (well-mixed) slick and the overall percentages of oil lost to evaporation can be predicted reasonably well. The rates of evaporation losses are often overpredicted where thicker oil slicks do not behave as well-mixed fluids. Usually this is because of water-in-oil emulsification (mousse formation). Under these conditions, evaporation is controlled by diffusion within the oil phase. This is an area that is currently not modeled at all in any oil-weathering model.

The formation of water-in-oil emulsions is only beginning to be understood at a fundamental level, and it cannot be predicted from first prin-

ciples (Fingas et al., 1997). Most model predictions are based on empirical results from laboratory studies with a variety of oil types. Oil spill modeling would benefit from basic research focused on better understanding of these mechanisms, such that better algorithms can be developed.

Dissolved component and suspended particulate material (SPM) interactions included in OSCAR are based on adsorption/desorption, but other research (Payne et al., 1987, 1989), has shown that whole oil droplet– SPM interactions overwhelm rates of dissolved component adsorption and sedimentation by several orders of magnitude.

Modeling of biological effects has not been included in the NC/FC system. However, model development and oil spill effects research generally have identified several areas where greater mechanistic and quantitative understanding is needed in order to develop predictive models. Again, the following are examples of research areas.

An understanding of oil toxicity is critical to the prediction of biological effects. While acute effects on fish and invertebrates can be modeled with reasonable accuracy (French McCay, 2002b), long-term effects are being studied (Bodkin et al., 2002; Bue et al., 1998; Carls et al., 1999; Esler et al., 2002; Golet et al., 2002; Heinz et al., 1999; Marty et al., 1997; Roy et al., 1998; White et al., 1999) but have not yet been incorporated into modeling. There is a basic lack of understanding and quantitative information that would be needed to develop such models.

OSRI has focused on the distributions of specific species in PWS, with an aim of being able to predict resource exposure to oil. Research that develops a mechanistic understanding of biological distributions is highly desirable, as there is a basic lack of understanding of organism distributions and movements within PWS and elsewhere, such that they cannot be predicted. The ultimate goal would be to model the movements and exposure of organisms at several trophic levels. Presently, work is ongoing for several fish species, but other groups are not being addressed. The focus should be on those organisms most vulnerable: birds, marine mammals, and fish that utilize the nearsurface and nearshore (such as herring, pink salmon), ichthyoplankton and invertebrates in the surface waters, nearshore benthos, and intertidal organisms (both plants and animals).

An unanswered question that is often raised is: Do organisms avoid (or are they attracted to) oil? This would be a potential area for study,

either using innovative laboratory studies or spills of opportunity. Such behavior could then be included in biological models and considered in NRDA.

Recovery of impacted species and success of restoration in augmenting impacted organisms are other areas that would be much in keeping with the legislative mandate and OSRI mission. The modeling of such recovery would be useful to NRDAs for future spills.

The Nowcast/Forecast model is conceptualized primarily as a real-time spill response tool, and this and related efforts have been a large OSRI financial commitment and focus. The OSRI Advisory Board, with input from the OSRI leadership and stakeholders, will need to judge whether the emphasis on this modeling activity is an accurate reflection of its priorities and its interpretation of the OSRI mission. In the committee’s view, the goal of OSRI’s modeling activities should be to support research and, while OSRI may contribute needed information to responders, it should not compete in the response business. OSRI’s modeling efforts should help advance the state of the art, including validation and analysis using hindcasting. Modeling can be used effectively to synthesize existing information and as a hypothesis-testing tool to define questions and identify needed research.