scope:

1.0 EXECUTIVE SUMMARY

With a huge amount of petroleum being transported by tankers and pipelines through the marine environment, and with increasing offshore oil exploration, drilling and production activities, the potential of an oil spill resulting in damage to our marine environment and to the critical near shore resources has aroused great concern amongst government decision-makers, oil industry officials, and the general public. Assessments have to be made of the potential risk of damage resulting from exploration and development activity based upon predictive impact evaluations of the fate of hypothetical or real oil spills. When an oil spill does occur, planning and execution of cleanupmeasures to minimize impacts also require the capability to forecastthe short term and long term behavior of the spilled oil.

A substantial amount of money has been spent by government and the oil industry over the past decade in an effort to develop the capability to predict the fate of spilled oil. Numerous models have been developed and applied, most of which could only predict the horizontal movement, or advection, of the oil slick. When oil is discharged into the marine environment it will immediately be affected by various physical-chemical and biological-chemical processes in addition to the physical advective processes. Most existing models do not have the capability to accommodate the effects of such processes although they are quite important in determining the overall fate of spilled oil. Over the past few years, a few models have been developed which combine advective and weathering processes in an attempt to provide a more comprehensive predictive capability. An inherent problem in this approach lies in the fact that these weathering processes have not always been integrated into the advection models in the most beneficial way; in other words, the later, more inclusive models essentially wind-up being a melange of ideas that have been strung together rather than a basic framework within which all process models are placed.

As is frequently the case in a relatively new technological area, the burgeoning of ideas in the absence of a well-defined problem framework has led to a proliferation of techniques designed to meet some broad objectives. Within this context an assessment of the state-of-the-art in modeling the fate of oil spills should contribute significantly to the quality of models developed in the future and the resultant decisions regarding the potential risk and/or hazard of an oil spill.

The objectives of this study, chosen with this in mind, are three-fold:

1. To assess the capabilities of selected oil spill fate models tosimulate real-world conditions;

2. To provide recommendations for the synthesis of various oil spill fate process components into a comprehensive "State-of-the-art" model; and, since this synthesized model will still have shortcomings,

3 . To recommend further improvements that can be made in oil spill simulation models.

To achieve these objectives, the study was divided into three phases. In the first phase, available information on existing oil spill simulation models was gathered and placed in one of the following process categories: (1) advection, (2) spreading, (3) evaporation, (4) dissolution, (5) emulsification (water-in-oil), (6) dispersion (oil-in-water), (7) auto-oxidation, (8) biodegradation, and (9) sinking/sedimentation. A truly comprehensive oil fate model was perceived to be composed of submodels for each of these nine processes. A preliminary review of the information was made with respect to the techniques used to simulate oil fate processes included in each model. The submodels for each individual process were grouped with others found in the literature, but not yet applied (i.e. "theoretical"algorithms), to allow for a screening and selection of those with the potential of contributing to the state-of-the-art modeling technology, based on a set of six criteria as described in Section 3.0. The main criterion used for the selection was that the process model or algorithm has a unique treatment which makes it more realistic, reliable, valid, theoretically sound, or applicable than others. The selected models were then subjected to a detailed evaluation in Phase II.

The primary goal of Phase II was to identify the model(s) which "best" represent(s) the state-of-the-art technologies for a given process. The factors considered included: theoretical soundness or realism, verifiability, level of data input, applicability to a range of conditions and localities, and accuracy of output. Details of this phase of the evaluation are presented in Section 4.0 of this report.

Results of the final phase of this study were to be recommendations for: integrating various component models, based on the results of the Phase II into a comprehensive oil spill fate model, and steps to betaken to improve oil spill fate modeling. In the course of the Phase III work, it was determined that a synthesis of a truly comprehensive model incorporating all nine oil fate processes could not be done at this time nor was it necessary for all practical purposes. There are processes such as auto-oxidation, biodegradation and sinking/sedimentation for which there are virtually no models available, and emulsification and dispersion whose mechanisms are not well understood, although unvalidated models are available for these two fairly important processes. Unvalidated models for the dissolution process are also available, but must be regarded as highly speculative because determination o f the dissolution rate from an oil slick is imprecise. The only processes which have received a great deal of attention by researchers and are capable of producing results with fairly reasonable confidence are advection, spreading, and evaporation. Models with varying degrees of sophistication and detail for these three processes are available, and can be assembled to form a limited comprehensive oil fate model for environmental impact assessment or oil spill risk analysis purposes. Although the current state-of-the-art is such that very accurate prediction of oil spill behavior is yet to be achieved, it is probably fair to state that reasonable determination of the fate of the spilled oil can be made using a combination of only three or four processes.

Depending on the purposes for which a model is intended, the geographical areas to which the model is to be applied, and the uncertainties of some oil fate processes, one should include algorithms for only those processes which significantly affect the spilled oil under those specific conditions and in which there is a high degree of reliability in the model output. In this way, it is possible to reduce uncertainty in the model outcome while still accounting for the most critical factors. This may, however, have an inherent flaw in that many of the important processes such as spreading, evaporation, emulsification, and dispersion are highly interdependent and synergistic, and through interaction largely determine the behavior and the lifetime of a spill. Although models for emulsification and dispersion contain considerable empiricism, omission of either one from an oil fate model will lead to unrealistic results. For example, emulsification will significantly retard spreading, evaporation, dissolution, biodegradation and dispersion; if it is ignored, the effective rates of other related processes will be unrealistically high, and will result in overly pessimistic estimates of environmental impacts. The process of dispersion of surface oil into the water column by breaking waves is also exceedingly important in that it largely determines the lifetime of the surface slick. Its exclusion from an oil fate model may, too, lead to overly conservative and unrealistic impact estimates. Sinking/sedimentation can also shorten the lifetime of the surface slick, but generally only occurs in the presence of a significant suspended sediment load (a condition which becomes more critical in nearshore regions).

It is recommended in this report that additional work be done on algorithms for emulsification and dispersion since they are so critical to several of the other processes. As mentioned previously there is no need, for all practical purposes, for a great amount of effort to be put into developing a fully comprehensive oil spill fate model. Such a model should, and likely will, remain a research tool for the foreseeable future primarily because of the great deal of uncertainty in the rates and mechanisms of such processes as biodegradation, auto-oxidation, and dissolution. When these uncertainties are factored into a more comprehensive model the overall reliability of the model prediction decreases while the percentage of additional oil accounted for is probably less than 10%.

Recommended steps for the selection of existing oil spill models, as well as some factors to be considered in constructing limited comprehensive oil fate models, are presented in Section 5.0 of the report. The precise nature of such models will depend upon the purposes for which they are intended and the geographic areas to which they are to be applied.

One of the most difficult and frustrating problems encountered in the course of this project was the general lack of data, collected from either spills of opportunity or experimental spills, that can be used to validate or verify models. This lack of data seriously limits the utility of models, and can be remedied by modest data collection and analysis programs. The success of such programs lies in comprehensive planning and execution of extensive data gathering. In too many cases field data collection efforts of this type are done strictly according to the wishes of the principal investigator, and with relative disregard for techniques used by other investigators. What results is a huge number of data that are virtually unrelatable from study to study. In order to achieve the desired levels of comparability it would be useful to construct a crude composite model from existing algorithms and look at the commonalities of data requirements and interrelationships between process models with an eye toward designing appropriate, efficient experimental programs whose objectives closely match model data requirements. The assumed rate constants or coefficients employed in the model can then be verified with such data, or the data can be used to estimate the proper rate constants or coefficients.

Decision making based on model predictions, is by the very nature of the uncertainty in the models, highly conservative. The model in use by the U.S. Department of the Interior falls into this category, and has significant implications for near-term future energy decision making. Such conservatism can certainly be either justified or refined on the basis of the findings of this oil spill fate model evaluation and initiation of certain recommendations found in Section 7.0.