An Interactive Computer Model for Oil Spill Training

An 1nteract1ve computer model which can be used for mar1ne oil pred1ction research and as a training tool has been developed. It uses an existing model from the University of Rhode Island which permits tracking of surface as well as entrained subsurface oil. To this are added models of sp111 cleanup and containment as well as calculations of costs involved for each of the response techniques. The performance of a response is judged in terms of the environmental and aesthetic impact-of oil on an area. The model is set up and run for two actual spills in Narragansett Bay as well as several example spills 1n the Rhode Island area. Outside evaluators have reviewed the model and judged it useful for tra1ning and prediction.

Relationship Between X and Boom Length Loss Rate (Abrahams 1977) Boom  Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   Table   2 Sample of Simulation (Cochran et al. 1975) Sample of Calculations (Holmes 1977) Shoreline Classification System (Gundlach and Hayes 1978) Common Characteristics of Equipment Animals (Malins 1977) Oil Spill Costs (1984 Dollars) ix

INTRODUCTION
Oil spills on water have been a major problem since the 1960 1 s when demand for oil began to increase. In the early years between 1956 and 1970, 80 percent of the 38 spills· in the world on water greater than 2,000 barrels (24,000 gallons) were within 10 miles of shore (Sittig 1974 1982). The total cost of the responses to these spills was over $300 million, including $2.5 million for the Argo Merchant alone (Schiff 1980). The environmental and economic impact of these spills, has lead to extensive research, designed to stop or reduce the affect that the oil has on the environment. The first step taken has been to determine the behavior of oil in a marine environment and to use this information for planning and training.

Oil Spill Processes
The chemical and physical processes which affect spilled oil are complex and interrelated and both are dependent upon oil composition and l environmental paramaters. Among the competing processes, shown in Figure   1-1, is the oil's interaction with the shoreline. Most of these are poorly understood. It is difficult, if not impossible, to take water and oil samples during an actual spill, especially if high sea states exist, so that the bulk of oil spill research has occurred in simulated laboratory environments.
Researchers have identified those factors which seem to be the most important. These include spreading, advection (both surface and subsurface), evaporation, dissolution, dispersion, emulsification, sedimentation, biodegradation, photo-oxidation and shoreline stranding.
These various processes work at different rates and thus are important during different times of a spill (see Figure 1-2). They also affect one another. For example, if evaporation is high there will be less oil available for the remaining processes. The major processes are discussed below.
Spreading is one of the most important processes in the first 6-10 hours of the spill. Both gravitational and surface tension forces increase the spreading while friction and inertia forces tend to retard it. Oil properties, temperature and the oil's thickness on the surface influence the forces. Short-time and small scale fluctuations also affect the rate of spreading (Stolzenbach 1977) . ' I

I ,(
A schematic overview of the various combined and competing weathering processes that act on spilled oil in the marine environment (from Burwood and Speers, 36). Reprinted with permission from Estuarine and Coastal Marine Science, Vol. 2, ro 1974 by Academic PresS:-Inc.  Advection is the movement of oil by wind, currents and waves. surface oil movement is mostly a function of wind drift, especially for offshore areas. In some nearshore areas, tidal currents and waves become more important. Limited research has been done on the movement of oil by waves and the resulting calculations are not easily performed for complicated wave fields. Subsurface oil is moved by tidal currents in estuaries and influenced by Ekman drift in offshore, deeper waters.
Advection can have varying effects, depending upon spill location and weather. For example, the wind moved the Argo Merchant oil offshore (Argo Merchant 1978}, but transported the oil to the coastline during the Amaco Cadiz spill (Hess 1978}. Evaporation is dependent upon oil composition and on the environment. a crude oil.
Lighter oils, such as gasoline, will evaporate faster than Wind, high temperatures and sea states will further increase the evaporation rate. Up to 40 percent of some crudes can evaporate in one day (Jordan and Payne 1980}. Oil dissolves into seawater at rates depending on the oil's composition and the seawater's temperature and salinity. The amount dissolved is usually only a few percent of the total volume so that dissolution is not considered to have an impact as large as most of the other processes (Davidson and Lawrence 1982}. Since dissolved oils are not easily detected, more research is needed to determine how much oil is actually dissolved. 6 Droplets of oil moving into the water column is called dispersion.
Dispersion is larger for heavier oils and higher sea states, although little data is currently available to confirm this. Some of the droplets resurface, but most seem to be neutrally buoyant ahd remain in suspension. The amount of oil dispersed decreases as the oil weathers, · but the particles which have been previously created continue to disperse and/or breakdown.
The water-in-oil emulsion often formed during a spill has a viscous, "chocolate-mousse" consistency, which is created by the combination of weathered oil and water. The longer the spill is exposed to the environment, the greater the percentage of oil going into emulsion.
Heavier oils and colder temperatures tend to accelerate formation of emulsions. Clean-up of emulsions is a major problem due to the increased volume. Typical oil-in-water emulsions· contain up to 80 percent water hence the volume of a spill may be multiplied by a factor of five in the emulsion. The bulk of the oil which stranded on the shore during the Amaco Cadiz spill was in the form of an emulsion.
Sedimentation is the process where particles of sand are mixed into the water and become attached to the oil. Since oil is very close to being neutrally buoyant, only a small amount of sediment will cause the oil to sink. This process occurs nearshore and is dependent upon depth, type of bottom, oil properties and the amount of turbulence caused by currents or waves. Once on the bottom, movement of this oily sand is dependent upon bottom currents.
Biodegradation is the transformation of oil by microorganisms. Only certain type of organisms are included in this process and anything that effects the population such as amount of light, nutrients and temperatures, will influence the rate at which organisms consume the oil. The impact of biodegradation is important only in the long term due to the relatively slow rate at which it operates. No field work has been done to study this phenomena, only controlled studies in laboratories.
Weathering of the oil by sunlight in the presence of oxygen is called photo-oxidation . It is dependent on the amount of light, oil composition and oil thickness. It has a very low rate and is usually ignored, except in special cases.
The behavior of oil near the shore is complicated and involves many oceanographic processes. The currents in the nearshore region are both complex and dynamic depending upon the region's physical oceanography and the manner in which waves diffract and break (see figure 1-4). Beach slope, local bathymetry and winds also influence water movement. In addition, there is a great deal of turbulence present due to breaking waves which can affect how any oil present is transported or deposited.
Stranding of oil on the shoreline is also greatly influenced by the tidal range. Oil left ashore during the transition from high to low tide, (Figure 1-5) may be refloated again during the next high tide. This was a recurring problem during the Amoco Cadiz spill (Hess 1978

Oil
Suggested cross section through an oil pool held against the beach face by wind and wave stress. Some of the other processes in Figure 1-1 may be important in special cases, but are generally not addressed in the literature and poorly understood.

Responses
The reasons for a response and the methods used can vary greatly and are determined by the size, time and location of the spill and the oil's characteristics. The major reasons for taking action are to protect human life and to minimize ecological impact. Some alternate motives are to minimize the socio-economic and aesthetic impacts of the spill. A trade-off between these aspects must usually be carried out since funds and manpower are generally limited. Trade-offs can also be influenced by outside considerations such as heavy weather, eliminating any possibilities of response, or political pressure.
There are many steps which constitute a response, and the magnitude of the response varies from spill to spill. An on-scene coordinator must assess the behavior of the oil and evaluate all environmental parameters. Action must then be taken to contain the oil and protect any vulnerable areas. Finally, the oil must be cleaned up and any areas damaged by the oil or response methods must be recovered and rehabilitated.

10
Organization of responses to oil spills begins at the national level. Regulations were initiated in 1968 with the National Multiagency Oil and Hazardous Materials Contingency Plan (Sittig 1974) and updated in the 1970's by the Federal Water Pollution Act (Federal Register 1975).
This legislation delegates the U.S. Coast Guard as the agency which monitors potential spill sites, inspects oil facilities, enforces the regulations, prescribes fines and supplies the on-scene commander (OSC) for marine spills not in inland waters. The Coast Guard also oversees and instructs regional and local officials in a response. The legislation authorizes equipment purchases and designates the responsibilities of other parties such as the Environmental Protection Agency and the Department of Defense.
At the local level, the Coast Guard has supplied a format to be followed for contingency plans which include plans of organization and areas of responsibility (U.S. Coast Guard 1978). Local authorities have expanded the plans to include details of response (Garry 1981), as well as site-specific considerations (Bell 1981, andHum 1977).
In the private sector, companies which are involved in some aspect of the oil business have developed plans and purchased equipment in order to protect themselves from liabilities which may occur if oil is spilled at their facility. A company has two options if the purchase of equipment is not practical. The first alternative is to join a cooperative in which each of the companies have invested in equipment and 11 training to decrease costs to individual companies (Franklin 1977, Hubbard andAllen 1979 (Hess 1978). The organization of a response team can be complicated (see Figure 1-

MARINE.
Tl.tllMIH"1.  contractors have used hoses with 7,000 psi water pressure to clean a marsh, destroying the roots of the remaining vegetation in the process (Owens and Foget 1982). The problems can clearly be overcome by proper training of managers and other personnel.

Training
There are many different training programs which focus on different aspects of combatting oil spills as well as different levels of personnel. Schools and workshops have been developed which may last 2-5 days. An intensive five-day course for management personnel is offered at Texas A. and M. University (Payne 1981). The agenda of this course is shown in Figure 1-7. Great Britain has a workshop for local managers.
such as town engineers or fire chiefs (Cormack 1977). Traveling workshops in Canada, which train 20-30 people in three days, are adapted to cover the environment in the location in which the workshop is offered (Zimlick-Owens 1979). Shorter one-day seminars cover a more limited field. Duerden (1979) discusses a program which enables local fireman to begin a limited response without waiting for other personnel to arrive.
Role-playing has been developed by the Coast Guard (Kangeter 1977) and for private industry (Marcus. 1977) as a training technique. Both of these allow a manager to be put in a situation where he/she must make decisions regarding a spill, as well as to fend off political or public relation problems.
14 Other aids include manuals for an on-scene commander (Foget 1979 andByroade 1981), video tapes and 16mm film with manuals (Kay 1977), as well as instruction books for the general public (Omohundro 1980a and l980b  evaluated by assessing the validity of the processes modeled. There have been three major reviews of modeled processes since 1977. Stolzenbach et al. (1977) reviewed techniques for modeling surface oil processes, concentrating on advection. In 1982, Davidson and Lawrence were searching for a trajectory model to be used for offshore work. They reviewed 15 models for advection, spreading, evaporation, dissolution, and emulsification as well as surface diffusion and vertical diffusion.
Surface diffusion, used to model small scale effects which are not included with wind and current advection, is defined by the reviewers as another form of advection, and vertical diffusion is another name for dispersion. The most extensive review is that of Huang and Monastero (1982) who reviewed 35 models (see Table 2-2). The reader desiring more detail concerning modeled processes is refered to these reports. Models on the list in Table 2-2 are referred to numerous times in the following paragraphs. The processes which are contained in these models are summarized in Tables 2-3 and 2-4.
A brief description of the methods used for modeling oil spill processes is presented below. Not all techniques are discussed, only those which are generally accepted being included. Little field data has been collected concerning these techniques, with few significant advancements made in most methods used since 1978. Fay developed a model which balances the forces of gravity, inertia and friction to determine the rate at which oil spreads (Stolzenbach et al. 1977). TRis method, which gives a good order of magnitude to the  1980 1980 1980 1980 1980 1980 1980 1980 1979 1979 1979 1979 1979 1979 1978 1978 1978 1978 1977 1977 1977 1977 1976 1976 1975 1974 1974 1973 1972 1970  "' "' No P-C or 1-C prOCUHI at.alat ...
Ito P-C or 1-C proceaaaa 11-lac ... . l a f · -· -h .... 11.~1 •• Pay' a apr ... tna "affect" ta aS..lac ... Mo P-C or 1-C procaaMa 11-llat ... size of a spill as a function of time, is used in most models. It has however not been proven to work in high sea states. Other researchers are developing random diffusion models but such methods are not yet commonly accepted.
The modeling of advection is divided into surface and subsurface oil transport. Most models move surface oil at between one and five percent of the wind speed plus the current. They do not agree, however, on a drift angle resulting from the Coriolis force. The values of the angle varies between zero and 30 degrees, with the majority using no drift angle. Most models use wind from a · single point over a large area. This is a poor parameterization in a wind field with significant shear present. Water currents contribute to surface and subsurface oil movement. The best results occur if actual data are used but the availability of these data is limited. Computer simulated current or inferred current values are generally used.
Evaporation has been measured in laboratory settings and the two most popular models are one by MacKay and one from the University of Delaware Monastero 1982, Wang, et al. 1976). The Delaware model divides the oil into components and evaporates each component separately. MacKay's model evaporates a percentage of the oil based on its thickness. Both of these methods give questionable results in high sea states.

23
Dissolution occurs at a slow rate and is ignored in most models.
The technique used in the University of Toronto (UOT) model (see Table   2-2) is based on observational data and could be adopted with extensive J experimentation. The USC/API model determines the rate of dissolution as a function of six parameters but these are difficult to measure and no experimental data are available to support this method. Less sophisticated models tend to group dissolution and dispersion together and use a constant rate which is a function of time, temperature and/or sea state. The only real data has been collected by Audunson (1982).
Both the URI and the SLICKFORCAST models use these.
Emulsification is a difficult process to simulate because little is known about the factors which affect it. A simple method is used in the SEAOOCK model. This technique arbitrarily reduces the oil present by one percent when the wind speed is greater than 20 mph and the spill is in shallow water. Complex models, such as the Toronto and the USC/API models, contain comprehensive emulsification models but these are empirically based and require a significant amount of input data which is not easily obtained.
At this time, no model contains feasible techniques for photo-oxidation, biodegradation and sedimentation. Most models also do not provide for shoreline interaction. In general, the oil trajectory is simply terminated at the shoreline. The modeling of processes still needs to be developed but most of the composite models perform adequately in simulating the specific tasks which they were designed for. 24 Many of the composite models could be used as a basis for the work discussed below. The OSSM model and the Drift model by Hunter can be run interactively but lack subsurface processes. The SEADOCK and SLICKFORCAST models also lack some processes. The Toronto and USC/API models are extensive but have a mixture of theoretical and empirical processes which are too complex. The Massachusetts Institute of Technology has published a model (Oil Spill Clean-up 1981) but its emphasis is on economic impact and regulation and does not contain a sophisticated oil behavior model. The University of Rhode Island model has most of the processes needed and is simple, flexible and easily accessible. It has been selected as the base for the work presented here.

Modeling Responses
There are two general approaches for modeling oil spill responses.
The first method is to model a specific response, such as a skirrmer, to determine the cost of the effort and the result that it has on the mass balance of a spill. In addition to modeling general responses, computers have been used to investigate and/or plan specific components. Swanson and Spaulding (1980) have taken a mathematical model by Cross and Hoult (1971) which simulates the interaction of oil with a boom, and combined it with real data from Abrahams (1977). The result is a model of boom effectiveness although in the technique has not been verified experimentally. A second approach is to assume cleanup parameters such as cleanup rate and efficiency and to use these as input into a composite model which determines the probability that the oil will come ashore.
These approaches may be programmed on a computer or performed by hand. Cochran et al. (1975) assumes environmental and equipment characteristics and calculates the mass balance. Table 2-5 shows a sample spill of 10,000 barrels with cleanup responses utilizing a skimmer and a boom.
Skimmer and dispersant responses were studied by Holmes (1977). Table   2-6 shows a typical calculation for responses utilizing two skimmers and a dispersant spraying unit. Fraser (1979) utilizes several of the models listed in Table 2-2. Numerous runs are performed using Cochran et al. (1975), Blaikley (1977), the BOSTM model (No. 25, Table 2-2) and RIVERSPILL (No. 12) with the probability of oil coming ashore at a given location being determined. The results are then used to determine the type, location, and amount of cleanup equipment needed. Audunson (1980) assumes a cleanup efficiency based on sea state and then uses the SLICKFORCAST model to determine the probability of the oil reaching land. None of these models however contain enough detail to simulate a reasonable cleanup technique .
Another use of computers is the U.S. Coast Guard's data base of cleanup equipment. This data base, called SKIM, stores the characteristics of twenty-six types of equipment along with their location and owner. In addition, the Coast Guard in New Haven has utilized a microcomputer for contingency planning (Harrald and Conway 1981 Cornillon and Spaulding (1978). This model has been designed to be modular so that as new algorithms are developed, they can easily be integrated into it. It was initially used to determine the impact of an oil spill on the fishing industry of Georges Bank (Cornillon et al. 1979). Details of the computer program with sample applications is presented in "Assessment of More recent applications are sunrnarized by Reed and Spaulding (1982). In this chapter, the processes modeled by Cornillon and Spaulding are briefly described; more extensive descriptions of the processes and assorted algorithms exist in the literature. This is followed by a detailed description of additions made to the model as part of this research.

URI Model
For the URI model, oil on the surface of the water is modeled as individual spillets or pancakes. Each spillet is an independent entity having its own mass, volume, oil composition and radius. Spillets are acted upon by all processes and are not affected by the presence of other spillets. 30 The subsurface regime is modeled with advection and diffusion as developed by Spaulding (1976). Oil in the water column is modeled as discrete droplets, each representing a specific amount of mass having unique oil properties. A floating three-dimensional rectangular grid is set up around the particles and is used to calculate the concentration based on the number of particles in each grid cell. The model then determines a diffusive velocity which is added to the current field.
The model as developed includes the following processes: l) advection: 2) spreading: A wind drift factor and drift angle is used for moving the surface spillets. be easily changed by the user.
These values cannot They can however be modified in the computer code. Currents transport the subsurface particles and add to the surface advection. These currents can be entered in any detail desired by the user.
Fay spreading (Stolzenbach, et al., 1977)  An important process for the training of personnel in the response to spills is the interaction of the spill with the shoreline. This process depends on the nearshore oceanographic process. Thomas (1975) and Winant (1980), have discussed wind-induced circulation in a shallow water environment. Shepard and Inman (1980) and Birkeier and Dalrymple (1975) have developed empirical equations for nearshore currents. These are just a few who have investigated nearshore processes. The modeling 32 of these complex currents requires large amounts of wave, wind, bathymetic, and beach slope data. Because such detail greatly ~xceeds our level of understanding of oil-shore interaction, it is inappropriate for this research. Instead, a simple method simulating the general movement of oil along a shoreline is used. As understanding of the spill-shore interaction improves, more sophisticated nearshore processes can be included.
The shoreline interaction routine developed here tracks the center of the spillet and prevents it from crossing the shoreline. After intersecting the shoreline, spillets are constrained to move parallel to it with the parallel velocity component. The spillet is moved away from the coast when it reaches the end of a shoreline segment or the end of a time step. A given percentage of oil from spillets intersecting the coastline is deposited on shore at the end of each time step. Subsurface particles use the same basic scheme although the entire particle is deposited on the first shoreline interaction. Details of these algorithms are contained in Appendix A. A shoreline classification system is used in the model both for the shoreline interaction and response methods. It is based on the work of Gundlach and Hayes (1978) who developed the classification system shown in Table 3-1. Complicated and time consuming field studies are needed to determine the shoreline composition, wave energy, and tidal dynamics in order to tlassify a coastline. This classification may also vary for Areas of reduced wave action. 011 may peratat tor many yeat1. Clean·up la not recommended unless oll concanlrallon la very heavy.
Areas of great biologic activity and low wave energy. 011 may persist tor years. Clean·up Is not recommended unless oll accumulallon la vary heavy. These areas should receive priority protacllon by using booms or oll aorbent materlala.
Most producllve ol aquallc environments. 011 may persist tor years. Cleaning of salt marshes by burning or cutting should be undertaken only If heavily olled. Mangroves should not be altered. Protacllon of these environments by booms or sorbant material should receive first priority. 34 different oil compositions. In this model, the ten types of coastlines have been reduced to four: rocks, beaches, marshes, and man-made structures.

Modeled Responses
The first decision that the coordinator must make in the event of a spill is whether or not to respond to it. Spills which are small, quickly dispersed or evaporated, or blown out to sea generally do not require a response. The coordinator must be aware of the situation at all times as weather or equipment availability may interfere with decisions. In this model, if response is initiated, the coordinator may contain the spill, clean up the oil, clean up the shoreline, disperse the oil or any combination of these options.
In defining the response alternatives, each of the above options is associated with its own set of equipment. The nine equipment types Equipment efficiency is a controversial topic so a review of existing data as well as assumptions which have been used in previous modeis is warranted. Evaluation of equipment in controlled environments such as the Environmental Protection Agency facility in New Jersey (Lichte 1979, Schwartz 1979 tend to be over optimistic when compared to real spills. Poor performance in the field is usually due to weather or high sea states, although it is sometimes caused by operator error or machinery breakdown. Cochran et al. (1975) and Holmes (1977) provide efficiency values for specific equipment based on sea state (see Figure   3-1). Blaikley et al. (1977) and Audunson et al. (1982) have designated overall "combat efficiencies." These values are estimates of the amount of oil cleaned up between the start of the spill and the time that it reaches shore. In reviewing reports dealing with real spills (Hess 1978, Marcoline 1980, O'Brien 1981, it was noted that these "combat efficiencies" are also too high. One of the systems rated to be most

Containment
One of the first responses normally put into action during spills is containment or protection so it will be the first section of the program to be discussed. This modeled response makes use of booms to enclose the oil and keep it from spreading or to deflect the oil away from vulnerable areas. The boom characteristics in Table 3-3 are loosely based on the U.S. Navy system which defines 3 classes of booms having 8 inch, 16 inch, and 24 inch drafts respectively. Additional characteristics come from Bellantoni (1979), Byroade (1981), Foget (1979), and SKIM. There are no actual booms with a draft of 60 inches as in class 5. This choice has been included to model attempts to block a narrow breachway or harbor entrance by dumping sand into it, effectively stopping almost any oil from entering the protected area.  The deployment of booms during a spill requires vessels of one type or another. In this model, the vessel characteristics shown in Table 3-4 were taken from SKIM, Byroade (1981) and the Argo Merchant report (1978). The smaller vessels are in general used nearshore while the larger ones are used offshore. These vessels are utilized in other response alternatives as well.
The boom itself is modeled after Swanson and Spaulding (1980) who combined research from Cross and Hoult (1971) and Abrahams (1977). In this model, the trajectory of the center of the surface spillet must pass between the end points of the boom otherwise the oil is not contained.
After the oil is inside the boom, there are two methods by which it can leave, assuming that the current direction does not change. First, if high currents are present, oil can be entrained into the water column, so particles are created based on the loss values of Abrahams (1977).
Second, the amount of oil which the boom can hold is limited by the efficiencies described before in Figure    Another method for oil pickup during spills is sorbent wringers.
These use an absorbent belt on a pulley system with a wringer at one end to squeeze out the recovered oil. The characteristics can vary greatly among manufacturers. The characteristics used in this model, in Table   3 They remain deployed until retrieved by the user.
The other method available to the user for cleanup on water is the use of skimmers whose characteristics are in Table 3-7. These are taken from SKIM, Foget ( ), Beach (1978, and Schwartz (1979). The classes are based on U.S. Navy classifications and the efficiencies are those shown in Fig. 3  model. The characteristics of the trucks are capacity (2500-6000 gallons) and cost per hour ($40 -$15) and are taken from SKIM and Byroade (1981). For offshore spills, floating storage is deployed. In the field, the two types of containers are steel barges which range from 1150 to 150,000 gallons capacity and flexible rubber bladders which can hold 50 to 6400 gallons. The characteristics in Table 3-8 cover this range and are taken from Allen (1982), SKIM and Bellantoni (1979). The cost includes a tug at $100 per hour.
When initiating a skinvner response, the desired position is entered and the effort operates on the closest surface spillet as the absorbent efforts did. Vessels and booms can also. be deployed with a response.
When a boom is used with a skinvner, it is assumed that the boom collects the oil thus increasing the skinvner efficiency but not inhibiting the movement of the oil. The user must discontinue this response when cleanup is completed.
Cleanup on shore Shoreline cleanup requires different types of equipment and techniques which are dependent upon weather, oil composition, and shoreline type. In his manual for on-scene coordinators, Byroade (1981) has detailed 23 methods which use various types and combinations of personnel and equipment. The options which Byroade has described have been reduced for this program and configured such that one cleanup  technique has been modeled for each type of shoreline and cannot be used on other types.
The normal procedure suggested by Byroade to clean up beaches is to use heavy construction equipment. Premack (1975) and Byroade (1981) supply the cost of equipment (see Table 3  When initiating a cleanup response, the user inputs a location and decides which equipment and personnel are to be deployed. The response will then clean any oiled shore within a 1000 meter radius. Calculations are performed which assumes that the oil is dispersed over a ten meter width of beach. This is considered an average value since marshes will have larger areas and man-made structures a smaller value. The amount of oil on the shore i~ reduced by the fraction of area which an effort can cover. The user must terminate the response when cleanup is no longer needed.

Oispersants
One response which sees limited use in the field is the deployment    Many people have attempted to quantify impact, although most research is directed towards the economic effect on a region. The most extensive work has been on the impact of the Amoco Cadiz (Auguier 1982, Hess 1978, Meade 1982. Recently the Massachusetts Institute of Technology has developed a model which attempts to addresses all aspects of spill impact (Nyhart et al. 1981, Oil Spill Clean Up 1981. Both of these studies are too specific and contain too many variables, so a generic method is needed which can be used for any type of location or spill. 56 Schulze's (1981)  The amount of oil on the shoreline can vary and the program calculates impact based on a minimum shoreline density~ After the Amoco Cadiz, researchers found oil in the coastline soil with densities of 5 to 50 tons per kilometer at thicknesses ranging from 4 to 100 millimeters (Hess 1978). For this research, an average width .of 10 meters is assumed to be affected and a minimum threshold value of .5 tons per kilometer is used. The threshold can be changed by the user in the plotting programs at the end of the main program.
The sensitivity to a spill is defined as the combined ecological and social impact on the area. Each region is assigned a weighting factor which is somewhat arbitrary, but can be changed in the program depending upon the research being performed. At this time, the subsurface is taken to be twice as sensitive as the surface and the shoreline region is three times as sensitive. Each of the shoreline types have been assigned a value. Rocky and man-made coasts are assigned a weight of one, beaches a weight of one and one-half and marshes a weight of two. This system results in the marshes receiving six times the weight in calculating impact as the surface.   Table 3-13 reflect actual spills as well as modeled spills. Normally, the cost of a spill is greatly increased when the oil is washed ashore . The shoreline was heavely oiled during the Tamano and Amoco Cadiz spills so the costs associated ~ith the 6 0  This program allows a user to run a specific spill and then rerun it using various response techniques. Since not every possible response and equipment is modeled in this program, the impacts and costs may not compare to actual data. The relative impacts and costs of various responses can be compared to determine which methods are more effective.
The user will learn the appropriate questions and problems associated with the various methods and can implement this knowledge during actual spills.

Model Integration
There were two steps performed for model integration after the detailed routines were developed. The first was to combine all of the modeled processes, responses and evaluation methods into a workable interactive model. Then, programs which handle all aspects of input and output data were developed. During both steps. the algorithms were designed to allow easy use of the program and to allow as much flexibility as possible. This results in three sets of programs: 1) a group for manipulating and plotting input data for the main program; 2) a main section containing the routines, for modeled processes and responses 62 and; 3) programs which process and plot the output results from the main program. The main program's framework will be discussed below followed by an explanation of the input and output programs and the resulting graphics.
The ma1n simulation is set up in sections so that for each time step, the model handles, under user control, implementation of the theoretical routines. When the user initializes the model, the program offers two major options. If the subsurface portion of the oil is not · considered to be important for a run, the program will allow an abbreviated run which does not create subsurface particles and track them. All other processes are included and the mass balance still includes subsurface oil. This alternative is preferred for simple trajectory studies as it is substantially faster. An option is also offered regarding input data. At each time step, a user can change any value of the environmental input. This option provides flexibility during a research or trajectory study by allowing use of data which is not available. One example of this is to have the wind blow from a specific direction for a certain length of time. For training runs, this option is not desireable.
During a run the user is continually queried by the program regarding the information she/he might like to see and the action to be taken. For example, the location of surface oil is displayed by a map of the spill area at the user's request. Figure 3-2 shows such a map.     More details of the processes, responses and operation of the program can be found in [Wilson] (1980) (Snooks and Jacobson 1979). Gordon, (personal communication) found that the energy spectrum of the Charlestown data is very similar to the Green Airport data for times longer than one day.
Weather Service data has been used to simulate wind predictions.
The U.S. Department of Commerce publishes a monthly summary of local weather which includes the resultant wind direction and speed each day.
For use in this work, these data were rounded off to the nearest eight points of the compass and nearest increment of 5 knots so as to simulate typical information which would be passed to an on -scene commander. For example, a calculated resultant of 9 knots with a direction of 135 degrees will yield a prediction of Northwest winds at 5 to 10 knots.
Additional examples of wind predictions will be seen duri ng an application run later in this chapter. 78 Currents Current data for Narragansett Bay was based on Gordon (1982) who suggested that tidal currents are important for mixing and wind currents are significant for sub-tidal flows. At times, density and continental shelf events such as storms can greatly influence currents. Development of a sophisticated wind current model is not within the scope of this research so values from a tidal current model developed by Spaulding and Swanson (1976)  Several studies serve as a background for the selection of the currents in Block Island and Rhode Island Sounds. Collins (1977) carried out a study using 600 surface and bottom drifters. Drifters do not give accurate speeds but trends can be established and this study indicates that northerly and north-westerly winds cause the surface currents to move offshore in the winter. During the summer, south-westerly winds cause the opposite effect. The bottom currents generally move opposite the surface but are more complex due to bottom topography. A study by Shonting (1969) indicates that the surface currents are predominantly non -tidal but the bottom currents are rotary similar to tidal currents.
The latest research indicates that most of the energy is in tidal 79 Figure 4- Figure 4-2 Block Island/Rhude Island Sound Grid currents and that this energy increases with distance offshore and somewhat with depth (Snooks and Jacobson, 1981). Since a tidal current model is available (Beachamp 1979), it is used for this study. The model covers the shelf from the western end of Long Island Sound to Buzzards Bay but for this application, just a portion is used (Figure 4-2). The grid separation is one nautical mile and makes an angle of about 15 degrees with lines of constant latitude.

Temperature
For sea water temperature, data has been obtained from various sources including Snooks and Jacobson (1979), Gordon (1982) Table 4-2. The wind data every three hours from Green Airport was used for the entire region. The bay is fetch limited to one nautical mile in the east-west direction and twenty nautical miles in the north-south direction. The values in Table 4-3 are assigned to the entire bay although in reality coves and inlets would have smaller waves. When reviewing the sea states calculated for 1977 and 1978, it is rare that a sea state of 2 is exceeded and this is consistent with the limited reports available for the Narragansett Bay. For the Sound region, the waves are fetch limited if the wind is from the north east or west but are not if the wind is from the south. The winds from the south are assumed to be duration limited to nine hours for this application.

Bathymetry
Depths for these runs were gathered from charts by choosing points which coincided with the current grids. The Narragansett Bay depth grid is one-fifth of a nautical mile and the spacing in the Sounds is one-half of a nautical mile.

Coastline
The shoreline has been digitized and stored using longitude and latitude and each point is assigned a coastline type. The shoreline of 83

Equipment
The characteristics and locations of the equipment availab le are predominantly taken from the Coast Guard's SKIM output for this region.
Information for a local cooperative, Clean Atlantic Associates, was obtained from Allen (1982) while Premack (1975) supplied information concerning municipal equipment.
Some equipment from outside the region Generic equipment types have been added to insure that the user will not deplete the stored equipment. For example, there are five units of each class of skinrner stored in Providence, in addition to any others listed in SKIM.

Process Validation
The process validation uses data from two different spills in Narragansett Bay . Sample runs have been performed and the model results compared to reports concerning the actual spill. Data from a small spill 86 which occurred off Quonset Point in 1976 has been reported by Noll and Spaulding (1977). The On-Scene Coordinators report supplied information about the spill of the merchant vessel PENNANT in 1973.

Quonset Point Spill
The first spill used for validation occurred on the morning of gallons (about 1000 metric tons) . The report of the on-scene commander (Pennant 1973) indicates that heavy oil came ashore at Warwick (point B) and later covered the shoreline at the other three points (A, C and D) noted in Figure 4-6a. Ultimately, a total of 13 . 6 kilometers of coastline was oiled, the heaviest area hit being the Old Mill Creek area in Warwick which . is just above point B.
The first attempt at simulating this spill assumed that most of the oil was released at the grounding site shown as a cross near the bottom of Figure 4-6a. The spill first came ashore at point A but never touched the Warwick shore (8). Given that this run did not simulate the observed spill very well, the simulation was repeated with the twenty degree wind deflection removed. The initial three spillet positions were spread over l 1/2 hours and four miles up the ship channel, assuming that the tanker leaked during and after the grounding. Figure 4-  A minimum concentration of 50 parts per billion was chosen for the impact plot (Figure 4-8). It can be seen that the subsurface dominates the impact for the first 48 hours at which time the shoreline impact becomes more important. Little additional oil is stranded after the first 50 hours so the impact is constant and curves 2 and 3 are identical in shape. The subsurface curve is irregular because the number of grids which exceed the minimum concentration change rapidly resulting in changes in area exposed and the resulting impact.
There are many reasons why this simulation did not match the actual spill. First, waves coming from the south were reported to be as high as four feet during the first eight hours of the spill . This could have caused oil to come ashore along the northern coast where this model does  Fi gur e 4-8 Penn a nt Simulation Impa c t not predict it. The wind record used in the simulation may also be erroneous due to the difference of space and time with respect to Green Airport. The wind was from the east at one point at the spill site according to the Coast Guard (Pennant 1973), but the airport recorded winds from the southeast. Furthermore, the airport records the wind every three hours so that any fluctuations in speed and direction between these times are not included in the data. This run is a good example of how a response is dependent upon the quality of data received by the coordinator. If the coordinator using only wind data from Green Airport,  Table 4-4. The predictions for days 1, 2 and 4 are in general agreement with the actual winds, however, the third day is off due to a wtnd shift during the middle of the day . This is the same type of predicted data an on-scene coordinator would get from a local weather bureau and the information which is passed to the user in this program. A student using the program for training will initially not see the actual winds that the model uses.
Maps of the spill without any response are shown in Figure       cleanup teams deployed to cover the expected stranding sites and a class one skirmier to p1ck up the oil. The maps in Figure 4-13 for the first t response indicate that the shoreline crews were efficient · in removing the oil within about 40 hours but that the skirmier was retrieved too early and the oil allowed to come ashore elsewhere. For the second case ( Figure 4-14), a boom was deployed approximately half-way down the peninsula, absorbents were used and the same shoreline cleanup teams were deployed. The boom kept the spillet from moving down the coast and allowed the cleanup teams to move into position. The absorbant cleanup method took about 8 hours longer than the skimmer but the boom helped to slow the spread. Once the oil moved past the boom, the southerly section of the peninsula was oiled and the spillet turned the corner. Oil which came ashore here was out of reach of the cleanup teams and in this simulation no additional teams were assigned. The costs in Table 4     The main program has been integrated with routines to control the input and output data. The input routines access a data base for a region and transfers the appropriate data into the proper format for use by the main section of the model. Both the input and output routines utilize graphics which allow the user to preview data before use or examine the data which is generated by the main simulation program. The graphic routines are easy to use and can generate either interactive or hard-copy graphics. 124 The model has been set up for use in the Rhode Island coastal waters. Results from the ·simulation of the actual spills in Narragansett Bay indicate that the program does a reasonable job predicting oil behavior. Sample runs were performed to display the capability of the model as a training tool.
Personnel from the Coast Guard and NOAA evaluated the model and found it promising. They felt that it has more capability than any program to which they have access and that it could be easily adopted to simulate spills in other regions.
The limitations of the model indicated by the evaluators, relate to its speed and to constraints imposed by the limited space.
both functions of the IBM computer presently being used.
These are The University of Rhode Island uses a timesharing system which during busy times only allows a user five to ten minutes of CPU time per hour. The speed of the model becomes marginally acceptable with 15 to 20 minutes of CPU time per hour. This problem can be overcome by using a dedicated computer such as a MICROVAX.
In terms of space, the URI computer requires about 1.5 megabytes to store the main model and about 2 megabytes for the input and output programs. In addition, the model uses over twenty megabytes for storing all input and output data for .a 10 to 20 day simulation. This is dependent on the size of the data base created. The URI system does not 125 allow one user to request this much space, so much of the data must be stored on tape which cannot be accessed during interactive execution. A dedicated system would also solve these problems.
If more speed and space can be obtained, more complex routines can be incorporated into the model. These may include more sophisticated There are several assumptions made in developing this routine.
First, the spillet or particle is not on land initially. The program performs some cursory chekcs but the initial spillet positions must be verified by the user. Secondly, the shoreline is digitized using longitude and latitude with dummy values between the coast and islands.
This will indicate to the program that a discontinuity exists. The sequence of the routine is described below and can be followed in the flow chart in Figure  2) The spillet trajectory first encounters a boom. The subroutine which simulates the boom is then called.
3) The spillet crosses a shoreline section. The spillet location is moved to the shore, then parallel to shore and finally projected out as in Figure  At the end of the sequence for each spillet, ten percent of the oil in that spillet is distributed among the shoreline segments crossed and the new spillet position is stored.
The algorithm is similar for the subsurface particles except that subsurface particles are not restricted by booms so this is not checked and the particle is deposited on shore when the first segment is hit. Boom Modeling The boom model described below is based on that of Swanson and Spaulding (1980) in which the theoretical work of Cross and Hoult (1971) is combined with data from Abrahams (1977). The method calculates the amount of oil that a boom can hold, the amount of oil which escapes around the boom and the amount entrained into the water column. The flowchart for the routine is shown in Figure B-1. The calculations to determine the currents under the boom are performed in another subroutine and stored for use.
The routine uses two methods to determine if oil can be held by a boom. The first method is based on the critical Froude number. Figure   B-2 shows the cross section of a boom with oil in it and current moving from left to right. As the current increases, the interface between the oil and water becomes unstable and oil is entrained. The Froude number is the critical parameter and is calculated by the equation: where F r = "tT = current velocity g = gravitational constant 6 = 1 -6 where 6 = specific gravity of oil d = draft of boom If this value is greater than the square root of two, then the water will essentially_pull all of the oil below the boom. Otherwise, a calculation is performed in which the drag forces on the oil are compared with the 132  Refer to Figure B-3 for nomenclature.
The volume of oil, is obtained by integrating the thickness of the oil in y and z: Since h is independent of y, and x = B -z,equation B-2 becomes : If the volume of oil in the spillet(s) impinging on the boom is greater than the volume which the boom can hold, another spillet having the same oil properties is created on the downstream side of the boom.
For the oil which is in the boom, the spillet position is adjusted to the midpoint between the boom endpoints . The amount of oil lost into the water column is then calculated using data from Abrahams (1977). Three linear curves have been approximated from Figure      Thank you for your demonstration last February of the Narraganse tt Bay Oil Spill Computer Model. As the NOAA Scientific Support Coordinator, I am involved in both oil spill response work and contingency planning for oil spill s on the state and federal levels in the New England area From your demonstrat ion. I can see a number of applications of your model to both local response personnel training and contingency planning.
As a training tool , your model allows an individual to become famil i ar with the various factors of wind, currents, tides. and physical characterist ics of oil wh i ch act together to determine slick movement The de termination of surface and water column oil concentrations and the weighted scoring of impacts on di f ferent shoreline t yp es identifies the need to develop a protection strategy which will minimize the overall impact I thought the additional capability of deploying response equipment and the inherent logi stical problems, both in terms of time and money, associat ed with the different response options to be par ticularly useful in giving an individual insight i nto some of the operational constraints of a response.
In the area of contingency planning, I think your model could be helpfu l in addressing the question of the the most cost effective siting of response equipment based on worst case or historically typical spill scenarios. A second area where your model might be helpful is determining in what areas and under what conditions dispersants might be considered for a spi 11 response. NOAA and the Coast Guard are currently doi ng a study of t he transportation pattern of hazar dous materials i n Narragansett Bay and adjacent coastal areas. On the basis of this study, we plan t o develop si te specific contingency plans for areas considered to be particularly "at risk". I would enjoy meeting with you again to discuss whether the Narraganse tt Bay Oil Spill Model might be useful in developing these plans 150