Modelling river oil spills: a review

Journal of Hydraulic Research

ISSN: 0022-1686 (Print) 1814-2079 (Online) Journal homepage: http://www.tandfonline.com/loi/tjhr20

Modelling river oil spills: a review Poojitha D. Yapa & Hung Tao Shen To cite this article: Poojitha D. Yapa & Hung Tao Shen (1994) Modelling river oil spills: a review, Journal of Hydraulic Research, 32:5, 765-782, DOI: 10.1080/00221689409498713 To link to this article: http://dx.doi.org/10.1080/00221689409498713

Published online: 13 Jan 2010.

Submit your article to this journal

Article views: 163

View related articles

Citing articles: 12 View citing articles

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tjhr20 Download by: [University of Nebraska, Lincoln]

Date: 03 August 2017, At: 11:14

Modelling river oil spills: a review Modélisation de déversements de pétrole en rivière: étatde l'art

Downloaded by [University of Nebraska, Lincoln] at 11:14 03 August 2017

POOJITHA D. YAPA Dept. of Civil and Environmental Engineering, Clarkson University, Potsdam, NY, 13699-5710, U.S.A.

HUNG TAO SHEN Dept. of Civil and Environmental Engineering, Clarkson University, Potsdam, NY, 13699-5710, U.S.A.

ABSTRACT The risk of oil spills in rivers has increased due to oil storage facilities along rivers, inland navigation, major oil transport pipelines that cross rivers. Inland oil spills are more frequent than ocean oil spills but are usually of smaller volumes. Oil spills in inland waterways can have enormous environmental and economical impacts because of their closeness to populated areas and economic centres. Previous review studies on oil spill modelling have concentrated on ocean oil spill modelling. In this paper all existing major river oil spill models are reviewed. The specific needs of the river oil spill models that are different from the ocean oil spill models are identified. The physico chemical oil spill processes which form the model are discussed. A comparison of the different models are presented. Simulations are presented to demostrate state-of-the-art in river oil spill models. RÉSUMÉ Le risque de déversement de pétrole en rivière s'est accru en raison des installations de stockage le long des rivieres, de la navigation intérieure et des grands oleoducs qui franchissent les rivières. Les déversements de pétrole en eaux intérieures sont plus frequents que dans le domaine maritime, mais sont généralement de plus petit volume. Les déversements de pétrole sur les voies navigables intérieures peuvent avoir d'énormes consequences sur le plan de l'environnement et de l'économie en raison de leur proximité des zones habitées et des centres d'activité économique. Les syntheses antérieures sur la modélisation des déversements de pétrole s'étaient focalisées sur l'aspect maritime. Le president article par contre s'est intéresse a tous les modèles importants de déversements de pétrole. Les besoins spécifiques aux modèles de rejets en rivière sont mis en evidence, en insistant sur les differences avec les modèles maritimes. Les processus physico-chimiques de rejet de pétrole, qui constituent la modélisation, sont discutés. Une comparaison entre les différents modèles est présentée. Des simulations sont également présentées pour illustrer l'état de l'art des modèles de déversement de pétrole en rivière.

Introduction Early oil spill modelling efforts had been mostly concentrated on the ocean environment. Ocean spills have generally involved much larger amounts of oil than river spills, and they, in many cases, showed visibly dramatic effects. Oil storage facilities along rivers and inland navigation increases the risk of oil spills in inland waterways. In some cases, there are major oil transport pipelines that cross rivers that have the potential to burst. Inland oil spills are more frequent than ocean oil spills but are usually of smaller volumes (Owens et al., 1992). Oil spills in inland waterways can have enormous environmental and economical impacts because of their higher potential to contaminate groundwater and fresh water supplies to cities, to affect man-made structures and recreational areas, and cause long term damage to fishery and wildlife. Owens et al. (1992) estimates that in the continental United States the average total spills amount to 50,000 barrels a year. Revision received May 30, 1994. Open for discussion till April 30, 1995.

JOURNAL OF HYDRAULIC RESEARCH, VOL. 32, 1994, NO. 5

765


The worst inland oil spill in the history of the United States occurred when an oil storage tank at West Elizabeth, PA. collapsed on January 1989 and spilled over 17,000 barrels of diesel oil into the Monongahela River. The oil slick reached the Ohio River near Pittsburgh within a day and drifted further downstream, affecting water supply intakes along the river (Miklaucic and Sasecn 1989). Recognizing the high potential and the adverse impacts of oil spills has led both government agencies as well as private industries to develop programs to prepare for and respond to oil spills (e.g. Hung 1991). An important element in these programs is the use of computer models to predict the movement of spilled oil and its possible impact. In the event of a real oil spill, a model can not only be used on a real time basis to assist the containment and recovery of the oil but also help to guide the field data collection for detailed environmental impact analysis. Computer models are also used to study scenarios of possible spills and during spill training exercises to assist the development of contingency plans and assess likely environmental impacts. In this paper, the state-of-the art of oil spill models developed specifically for rivers are reviewed. A comparison of features and algorithms available in each model is presented. The general physicochemical concepts of inland water oil spill models are discussed. Model performance is illustrated through a hindcast of the largest oil spill in the United States (Ashland oil spill). Relative importance of different parameters are also illustrated through simulations. River spill models - physico-chemical aspects The fate and transport of an oil slick in a river are governed by complex interrelated physicochemical processes that depend on the oil properties, hydraulic conditions of the river, and environmental conditions. In Fig. 1, these transport and weathering processes are summarized schematically. Figure 2 is a block diagram of a river oil spill model. In general, the movement of spilled oil in the river is governed by: a) the advection due to current and wind; b) spreading of the surface slick including both the turbulent diffusion and the mechanical spreading due to the balance among gravitational, inertia, viscous and surface tension forces; c) emulsification and turbulent mixing over the depth of the river; d) changes in mass and physical-chemical properties of the oil due to the weathering process such as evaporation and dissolution; and e) the interaction of oil with shoreline. Some of the oil droplets in the suspension may become attached to suspended particulatc matter and slowly settle to the bottom. In addition, photo-chemical reactions and microbial biodegradation can also change the character of the oil and reduce the amount of oil over a long duration.

Fig. 1. Oil slick transformation in rivers. 766

JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 32. 1994, NO. 5

START RIVER DATA Geometry Shoreline conditions

T

SPILL DATA Spill volume,location.duration Characteristics of spill material

''

HYDRAULICS Read flow boundary conditions Compute velocity distribution


'

YES FLOW TH ANGED ?

WEATHER DATA Wind direction and magnitude Temperature

1

APPLY SPILL PROCESSES Advection Horizontal diffusion Mechanical spreading Dissolution Evaporation Shoreline deposition Vertical mixing Emulsification Biological Effects

1

OUTPUT | END OF SIMULATION? | YES | STOP

Fig. 2. A sample model structure. When modelling oil spills in rivers the special needs of the narrower water body (compared to ocean) and wide variation of flow conditions need to be recognized. In particular the following aspects need to be recognized and included in the model. - Rivers can have large current velocities with drastic variations across the river width. Rivers also encounter flow divisions due to in stream islands or artificial obstructions. Therefore, it is necessary to model the flow velocities accurately. - Rivers are confined water bodies with sharp curves and islands. The oil spill reaches a shore shortly after the spill. The movement of the oil slick down the river is profoundly affected by the oil interaction with shores. The oil spill model therefore, should have a good computational scheme to detect the parcels reaching the shore including the complex shoreline shapes and simulate the oil interaction with the shore in a physically real manner. - In rivers where discharge is controlled, contingency planners seek features such as what happens to the oil spill if the flow is increased or decreased. To be able to model oil spills under such conditions require that oil spill model be run in unsteady flow conditions. - In rivers with drastic water level fluctuations the shoreline location and the shape changes rapidly. The oil spill model should be capable of detecting these changes as the water level changes to find extent of the slick area. JOURNAL OF HYDRAULIC RESEARCH, VOL. 32. 1994, NO. 5

767

- Depending on the size and geographical terrain of the river area, the use of wind data from a nearby weather station may need a modification to correct for the actual effect felt at the water surface. - Ekman layer effects (Madsen, 1972) that are sometimes considered in ocean oil spill models have much less significance in river oil spills models and therefore can be ignored. Review of river oil spill models


Most of the existing oil spill models were developed for marine environments.Only a few were developed for rivers. These models are by Tsahalis (1979), Fingas and Sydor (1980), Shen and Yapa (1988), Reed et al. (1991), Shen et al. (1991), Reed and French (1992), and Yapa et al. (1994). Processes included in these models are summarized in Table 1. Table 1. Existing river oil spill models RIVERSPILL

WPMB

NRDAM

Flow computation

Empirical 1-D (secondary currents computed at bends)

Time dependent 2-D depth averaged

Mean Flow (supplied)

Transport computations

(surface layer) 1-D, provisions to compute some 2D effects

2-D surface only

3-D surface layer k suspension, vert uniform velocity

Advection and Turbulent Diffusion Mechanical Spreading Variable shoreline Deposition Evaporation Dissolution Vertical Mixing GIS data Simulation in the presence of ice Biological effects Grid sizes used Application

Yes

Yes

Yes, but circular

always

No

ROSS

ROSS2

ROSS3

Time dependent quasi 2-D (1-D network coupled with stream tube ) 2-D surface layer only

Time dependent quasi 2-D (1D network coupled with stream tube ) 2-D surface layer &; suspension over the depth

Similar to ROSS2 but on non rectangular bathymetry based same as ROSS2 but the size of the grid is non-rectangular, varies with water level

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

layer

data

No

No

Yes

Yes

Yes

Yes

No

No

Yes

Yes

Yes

Yes

No

No

Yes

Yes

Yes

Yes

No

No

Yes

No

No

No

No

No

Yes

Yes

Yes

Yes

No

No

Yes

No

No

No

N/A

1000m

few hundred to few thousand meters St. Marys, St. Clair, Detroit, Niagara, St. Lawrence

500ft / 1000ft

200ft / 500ft / 1000ft Ohio, Monangahela, Allegheny, St. Lawrence

Variable

lower River

Mississippi

Montreal area

harbor

St. Marys, Clair, Detroit

St.

St. Clair River-Lake System

Tsahalis (1979) The first known River Oil Spill model is RIVERSPILL developed by Tsahalis (1979). The model was applied to a 1205 km section of the lower Mississippi river. The model is capable of operating in either deterministic or stochastic mode. In stochastic mode RIVERSPILL used historical wind and current data to estimate the probability that an oil spill will reach a specific region. RIVERSPILL is a one-dimensional model with some provisions to compute two-dimensional effects. The current velocities in the river were computed based on site specific empirical relationships. The primary velocity was considered to be constant across the main channel. In river bends, a secondary current is also computed.

768

JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 32, 1994, NO. 5


In addition to advection, the mechanical spreading of oil was computed based on Fay's (1969, 1971) equations. The oil slick shape was always considered to be circular. When oil hit the shoreline, the amount deposited on the shore was computed based on the area of the circular slick that overlaps with land. Two accidental oil spills were hindcasted by Tsahalis (1979). Model simulations seems to agree with the observations reasonably well. RIVERSPILL was developed in Fortran for the main-frame computers. WPMB model (1981) Fingas and Sydor (1981) developed a two-dimensional model for surface oil slicks applied to a short reach of the St. Lawrence River near the Montreal harbour. Their model used a two dimensional hydrodynamic model developed by Leendertse (1972) to compute the velocity distribution in the river. The entire oil slick volume is represented by a large number of individual parcels. The drift velocity of these parcels are determined by the wind factor approach. The area of each oil parcel is calculated, independent of neighboring parcels, by Fay's spreading laws for circular slicks. Since the spreading of each parcel in the slick is affected by the distribution of the entire ensemble, this approach for simulating mechanical spreading is incorrect. The shoreline deposition was computed by a picket gate method. In this method when oil reaches a shoreline a given amount of oil is removed, depending on the shoreline designation. The type of hydrodynamic model used consumes large amounts of computer time when simulating oil spills under transient (unsteady) hydraulic conditions in long rivers. NRDAM (1991, 1992) Reed et al. (1991) developed a Natural Resource Damage Assessment Model and applied to some of the rivers in the Great Lakes area. NRDAM includes the processes of evaporation, spreading, shoreline deposition and resuspension, vertical mixing, dissolution, oil degradation, mass removal by cleanup, and sediment-oil interaction. NRDAM velocities are obtained from a previously computed and stored data base, not dynamically computed based on flow boundary conditions and governing equations corresponding to the present conditions. It is not clear how the velocities used in the most current version are computed, but in early versions of NRDAM the velocities were computed based on stream tube method (Reed et al., 1991). NRDAM has a uniform velocity profile in the vertical direction of water column. The surface velocity is different from the depth averaged velocity by 10%. In this model the transport equations for both pollutant and biota are solved by using Lagrangian Parcel method. Oil spill transport and physico-chemical processes are simulated using three-dimensional computational schemes. NRDAM consists of three sub-models; physical fates, biological effects, and economic damages. The model can simulate multiple shoreline types, and spatially variable ecosystem habitats. The model is integrated with a geographical information system (GIS) data base. NRDAM is intended as a tool to be used for damage assessment. It can estimate the number of birds, mammals, fish killed and the impact on ecosystem. Monetary values for each species is also available in the database. The cost and benefit of technically feasible techniques for restoration of impacted resources can also be estimated.


769


ROSS (1988) Shen and Yapa (1988) developed a two-dimensional River Oil Spill Simulation (ROSS) model to simulate surface oil slick transport in rivers with or without ice covers. The model used a Lagrangian discrete-parcel method to simulate the transformation of oil slicks. Transformation processes considered include mechanical spreading, turbulent diffusion, evaporation, dissolution, and shoreline deposition and resuspension. ROSS was applied to three rivers Detroit, St. Clair, and St. Marys in the Great Lakes region. An interactive version coupled with graphical interfaces is also available for microcomputers (Yapa et al., 1989). ROSS2 (1991) ROSS has been extended to simulate oil slick transformation in both surface slick and suspended oil. This model ROSS2, therefore, includes the vertical mixing of oil into the water column. The model has been used in hindcasting of actual oil spills in St. Lawrence River and Ohio-Monangahela River (Ashland Oil Spill). Complete details of the ROSS2 model and the simulations can be found in the report by Shen et al. (1993). ROSS2 has been applied to Ohio (upper 203 km), Monangahela (200 km), Allegheny (115 km), and St. Lawrence (upper 160 km) rivers. The model has been completely integrated with the flow model, data entry interfaces, and visualization interfaces. ROSS3 (1993) ROSS3 is the latest version of the ROSS series of models. It has several important upgrades over ROSS2. First it used a natural stream-tube co-ordinate system to compute the flow distribution. Then it uses the same quadrilateral co-ordinate system to compute oil spill transport. The coordinate system used allowed the velocity computations to be limited only to areas where there is oil. This fact combined with the use of stream tube model allows good velocity computations as shown by their comparisons with observed data (Yapa et al., 1994) at a fraction of the time required for two-dimensional curvilinear co-ordinate system based flow models. As demonstrated by Yapa et al. (1994), the new co-ordinate system results in significant improvement in the oil transport simulation. The improvement of the results become critical at several situations: a) where the river is split by an island; b) when the river branches off to two or more branches; c) when the river shoreline changes are too sharp to be defined by conventional rectangular grid systems; d) when the rise and fall of water levels cause the shape of the shoreline to change. ROSS3 is a completely integrated model that allows user the traditional user friendly interfaces as well as the ability to set up the flow model and the oil spill model for new regions through the menu based interface. Hydrodynamics of the river Modelling the transport of oil spills require a complete hydrodynamical information of the modelling domain. Strictly speaking the hydrodynamics is affected by the presence of the oil slick. However, the state-of-the-art on this particular area is such that the influence of oil slick on the hydrodynamics of the underlying water body is not known. General argument is that the effect is not significant because the oil slick thickness is small compared to the depth of water and the area covered by the oil slick is small compared to the surface area of the hydrodynamic body. Consequently, in oil spill models the hydrodynamics is treated independent of the oil slick and then the 770



information from the hydrodynamic model is superimposed during the oil spill modelling. This method provides reasonable results in most cases and is generally accepted as the method of computation. The hydrodynamics of the river can be computed by using finite difference or finite element method. These models can be one-dimensional, two-dimensional or three-dimensional depending on the needs of the oil spill model. One-dimensional oil spill models are out of date and will not be discussed here. The best oil spill models for rivers available today use quasi three-dimensional. Their hydrodynamical information generally comes from two-dimensional or stream tube computations. Therefore, in this section we will discuss only the two dimensional and stream tube type hydrodynamic computations. Two dimensional computations - rectangular grids In a two-dimensional hydrodynamic model the depth-averaged shallow water equations are used. Numerous numerical models exist in the literature for determining the two-dimensional flow distribution in shallow waters. Examples are Leendertse 1970; Hamilton, et al. 1982. In these models the equations are written based on a rectangular grid system. The models provide reasonable simulations. Stream tube method can also be used to compute the two dimensional velocity distribution. Although mathematically less elegant than solving the two dimensional shallow water equations, stream tube method has proven to give good results if enough cross sections are used to describe the river geometry. One potential problem is that the velocity computations are originally done at the cross sections and, therefore, an interpolation to a rectangular grid system is needed before using in standard implementations of the oil spill models. The complete details of the implementation of this method is given in Shen et al. (1993). Two-dimensional computations - bathymetry based coordinates In all of the models that used rectangular grids the velocities are either computed on a rectangular grid system or assigned to one after computing the stream tube velocities. This results in discretization error especially in rivers because the flow velocities are dominant in the direction of flow. These discretization errors can be significant when dealing with oil interaction along the shorelines. This inaccuracy may be reduced by making the grid size finer, but a two-dimensional representation of a river with a very fine rectangular grid requires inordinate amounts of computer memory. A rectangular grid system is not the best way to represent irregular shapes like a river shoreline. All two-dimensional river models that represented a water body through a rectangular grid system had the same problems. The model could not account for river shape changes with water level fluctuations, and ad hoc treatment is required for handling slick transport around islands. Because of the complex shapes of rivers and the dominant velocity component in the longitudinal direction, the results from the river hydrodynamic models are generally improved when the computations use boundary fitted coordinates (e.g. Sheng and Choi, 1989; Johnson, et al. 1989; Spaulding and Liang, 1989). All of the models above are very time consuming for long and complex shaped water bodies. Therefore, in ROSS3 (Yapa et al., 1994) used the stream-tube method to compute the flow distribution in rivers and retained the same bathymetry based co-ordinate system for oil transport computations. The implementation is based on a cross section and cell based concept that also uses advanced data structures for programming. The velocity computations are compared with


771

observed data and matches well (Yapa et al. 1993). The method is well suited for water bodies with natural complex curvatures, many obstructions (natural islands or otherwise), and water level fluctuations that change the shoreline shape and location.


State-of-the-art of physico-chemical processes and their implementation in river oil spill models The governing equations for the motion and spread of oil on the water surface and the water column can be written by modifying the conventional advection-diffusion equations. The modifications come in the form of addition of physico-chemical processes that oil undergo. These governing equations are described in detail for a two-dimensional two-layer (surface and water column) system in Shen et al. (1993) and Yapa et al. (1994). The governing equations for the three-dimensional case is briefly described by Reed et al. (1991). The governing equations can be solved by an Eulerian method. Eulerian methods were used by a few early ocean oil spill models. They are very rarely used now because of numerical diffusion problems and excessive computer times required for solving the system of equations. Lagrangian parcel method has become the most popular method of implementation of oil spill transport and fate algorithms. Lagrangian discrete parcels method Lagrangian discrete parcel method is inherently stable although the time step should be compatible with the grid size (Roache, 1972). The scheme does not require the solution of a system of equations. Since the computations follow the parcels, the scheme is more efficient than Eulerian schemes. Typically in models using Lagrangian Discrete Parcels Method oil is represented by a large ensemble of small parcels. The movement of each parcel is affected by the physico-chemical processes. Once the parcels are released in the water body, their discrete path and mass are followed and recorded as functions of time relative to reference grid system fixed in space. Then the density distribution of the ensemble can be interpreted as the concentration of the oil. Since the movement of each parcel is dependent upon the distribution of the entire ensemble, all parcels must be traced to a time level before proceeding to the next. The method has been described in detail in Shen et al. (1993) and Yapa et al. (1993). In a Lagrangian parcels application, the total mass of the system is always conserved. Advection In rivers advection is the main mechanism that governs the transport of the surface oil slick and the suspended oil. Once spilled into the river, oil is generally transported downstream. The advection of surface oil is caused by the combined effects of surface current and wind drag. The advection of subsurface oil is the movement of suspended oil droplets entrained in the flow due to the subsurface current. In almost all models the surface advection is simulated using the wind factor approach. The most commonly used value for contribution from the wind velocity is 0.03. The surface current velocity is approximated to be 1.1 times the average velocity of the water column. Justification for the factor 1.1 is available in several earlier publications (e.g. Shen and Yapa, 1988). 772



The wind velocity component is added to the current velocity component vectorially to determine the total surface drift velocity. Once the drift velocity is known the movement of oil parcels are computed by numerical integration. The time step must be compatible with grid size to make sure that the simulation is accurate. The criteria necessary to satisfy the relationship between time interval and grid size are given by Roache (1972) and Cheng et al. (1984). Horizontal diffusion due to the turbulent fluctuation of the drift velocity is simulated based on the random walk analysis (Fisher et al. 1979). The computational details of the advection simulation and the diffusion coefficient is given by Shen et al. (1993) and Yapa et al. (1993). Flume experiments by Sayre and Chang (1969) indicated that the diffusion coefficient for surface dispersants can also be computed based on flow depth and shear velocity. The advection computation of surface oil as well as the oil in the water column can be done with good accuracy with the exception of one problem. The wind data is usually obtained from a nearby meteorological station. The correlation between the wind station data and the wind that exists near the river water surface depends on a number of factors including the geography of the area and the river. There is very little research on how to adjust wind data for the discrepancy, and one would have to rely on Weeks and Dingman (1972) study. For simulating the advection of oil in the presence of an ice cover, the best study available is that by Cox and Shultz (1981). This simulation is based on three categories: smooth, small roughness elements, and large roughness elements. It is expected to give reasonable simulation of the oil slick underneath the ice cover. However, this is an area that needs further research to improve the understanding of how the oil moves in the presence of ice. Mechanical spreading Mechanical spreading is the horizontal spreading of the surface oil slick due tothe balancing forces of inertia, gravity, viscosity, and surface tension. Mechanical spreading is known to terminate when the oil slick becomes thin and ruptures into patches. The spreading of oil increases the area of the slick and enhances other weathering processes such as evaporation, dissolution, and emulsification. Therefore, it is an important process affecting the fate of the spilled oil during the early stages of the spill. The narrow water bodies cause the oil to hit the shores after a short time. Once the oil reaches the shore at least in that vicinity oil will not be subject to mechanical spreading. Since spreading can cause rapid expansion of the slick during the initial stages of the spill it should be included in the model in general. The mechanical spreading process only affects the spreading of surface slick. The spreading on a free water surface is different from the oil spreading when the slick is under an ice cover. S p r e a d i n g in o p e n w a t e r Fay's spreading theory (1971) is based on a rather comprehensive description of the spreading mechanism and has been verified by laboratory experiments (Fay 1971; Hoult and Suchon 1970) and other analytical solutions (Fannelop and Waldman 1971). Fay's spreading theory is derived for single component, constant volume slicks with idealized configurations in quiescent water. This theory considered the spreading of oil as a result of two driving forces, gravity and surface tension, counterbalanced by inertia and viscous forces. The spreading of an oil slick is considered to pass through three phases. In the beginning phase, only gravity and inertia forces are important. In the intermediate phase the gravity and viscous forces dominate. The final phase is governed by the balance between surface tension and viscous forces.

10URNAL OF HYDRAULIC RESEARCH, VOL. 32, 1994, NO. 5

773


The oil spreading theory by Fay was developed for highly idealized slick configurations floating on calm waters. Although, the same method has been adopted for moving waters (of both ocean and rivers) through superimposition, it has never been tested against any such data under controlled conditions. Obviously the shearing effects of the moving currents will impose forces on the slick that are significantly different from the ones used by Fay. Other more complicated formulations like Foda and Cox (1980) have little practical value. More research is needed to understand the oil spreading in the presence of currents. S p r e a d i n g u n d e r ice c o v e r For spreading under ice the three phases would be buoyancy-inertia, buoyancy-viscous, and viscous-surface tension. Yapa and Chowdhury (1989) have shown that the buoyancy-inertia phase lasts for only a very short time, and that viscous-surface tension phase is not present in spreading under ice covers.They developed equations for radial spreading under ice covers in either constant volume or constant discharge mode. Weerasuriya and Yapa (1993) developed equations for onedimensional spreading under ice covers. For simulation of oil spreading under solid ice the work presented by Yapa and Chowdhury (1990), Weerasuriya and Yapa (1993) is robust and compares well with laboratory data for the limited cases they analyzed. However, field situations can have very complex ice conditions that limit the use of above theories. This oil spreading under ice work was developed for calm water conditions. For oil in broken ice conditions the relevant information is available in the work by Yapa and Belaskas (1993) and Venkatesh et al. (1990). The spreading of oil in the presence of moving broken ice is not understood at all. Of the river oil spill models available ROSS and ROSS2 use the oil spreading under ice theory developed by Hoult et al. (1975). Hoult's work is now considered out of date. ROSS3 uses the more recently developed theory for oil under ice by Yapa and Chowdhury (1990) and Weerasuriya and Yapa (1993). NRDAM can simulate oil spreading in the presence of ice, but the detailed information on the algorithms used is not available in the published literature at present time. Evaporation Evaporation is the loss of oil mass to the atmosphere and occurs immediately after the spill. Spreading increases the surface slick area causing the evaporation rate to increase. The amount and rate of evaporation depend on the composition of oil, wind conditions, temperature, and surface slick area. Highly refined oil can lose 75 percent or more of its volume through evaporation within a matter of days. Oil is a mixture of complex hydrocarbons. By assuming the oil to consist of a mixture of several hydrocarbons, evaporation can be modelled using a method known as the multi component method. In this method the characteristics of each individual hydrocarbon and the composition of the mixture is used to compute the evaporation rate. The problem is however, that the oil composition vary widely and accurate differentiation between different grades of oil is difficult. Nevertheless, this method is considered to give reasonable estimates for evaporation. An alternative and a simplified method has been to assume that oil characteristics can be described by one set of parameters that gives the average evaporation rate for the oil. This latter method is known as the single component method. Both methods were developed by Mackay and his co-workers. Which method is best suited for an application depends on the detailed needs of the output and the computational effort. In laboratory conditions the former method gives better results. However, since the actual amounts evaporated in field conditions can not be measured and are different from laboratory conditions, it

774


is not clear whether the additional accuracy gained in computation present any significant advantage to the real estimates, when viewed against the additional computing effort required. Both methods are described in detail in Mackay, et al. (1980). NRDAM uses the multi component method, where as ROSS, ROSS2, ROSS3 uses single component method to compute evaporation


Dissolution Depending on the composition of oil it may or may not have appreciable solubility. Usually it is the lighter hydrocarbons that dissolve in water which are the same ones that evaporate faster. Although the amount dissolved is small compared to the evaporated amount the models generally compute this amount because of the high toxicity of the dissolved oil to biological life. Cohen et al. (1980) provides a method to compute the dissolution rate. Additional data to compute dissolution can be found in the work of Lu and Polak (1973). Computation of the dissolution can be done using a multi component or single component approach just as in evaporation. Shoreline deposition In rivers, oil reaches the shore in a shorter time after a spill when compared with ocean spills. Oil once reach the river shoreline, may be deposited along the shoreline, and later re-entrained into the river current. The amounts deposited and re-entrained depends mainly on the shoreline type (sand, marsh, gravel etc.) and the hydraulic condition of the river. Since this process significantly affects the downstream transport of oil in a river, it should be modelled. There has been very little research done to understand the process of shoreline deposition and reentrainment of oil. The best available methods are based on exponential decay method and oil holding capacity by Gundlach (1987). In exponential decay method different shorelines are assigned a half life value. Then the amount of oil retained by the shoreline after a given time can be computed using the standard half life equations. In North America many river shorelines have been assigned an Environmental Sensitivity Index (ESI) and color atlasses are available showing this information. Correlations between ESI values and half life exist and can be used for simulations. In addition to the exponential decay each shoreline also has a maximum holding capacity which determines the maximum amount of oil that a given area can retain. Oil in excess of that amount is returned to the water. Vertical mixing and emulsification In turbulent waters, some of the oil is dispersed into the water column as suspended droplets. These dispersed oil particles may form emulsions, however, emulsion formation is less likely in river spills because most of these spills are refined products as compared to ocean spills which consist mostly of crude oils. Refined products are much less likely to form stable emulsions. Wave and tidal action in coastal and ocean environments are major contributors to the turbulence that causes break up of oil in marine spills. In rivers, while tidal and wave action is generally less than in ocean, rapids (high turbulence sections) and flow over artificial barriers (dams, spillways) or natural barriers (water falls) cause enormous turbulence that breaks up the oil and mixes almost completely into the water.


775


The mechanism of formation of oil droplets and their entrainment into the water column is not clearly understood. Delvigne (1991) and Delvigne and Sweeney (1988) correlated dispersion rate and the droplet size distribution of the dispersed oil to the energy dissipation rate of breaking surface waves. Delvigne (1993) presented arguments to demonstrate the similarity between dispersion by breaking waves and dispersion of oil by different sources (e.g. flow over a dam, flow around an obstacle). However, the understanding of the mechanism is not very good. Emulsification is a physico-chemical process that results in a substantial increase in the apparent viscosity and volume of the surface layer due to the increase in water content in the slick. This increase in the water content is due to the incorporation of small water droplets into the slick as a result of the coalescence of large oil droplets back into the surface slick. The most commonly used method for computing the increase in water content and the change in the apparent viscosity of the emulsion is based on the equations developed by (Mackay et al. 1980). Biological fates NRDAM is the only river oil spill model that can model the biological effects. It is known to calculate the number of birds, mammals, fish, shellfish, and their young killed by a spill. In NRDAM fish eggs and larvae are assumed constant and evenly distributed across each ecosystem within each month of an annual cycle. Mortality is calculated using laboratory acute toxicity test data corrected for temperature and time exposure, and assuming a log-normal relationship between percent mortality and dissolved concentrations. The model considered both the short term kill and the long term losses. Model performance Hindcast of ashland oil spill In this section, results from the hindcast of the Ashland oil spill using ROSS2 model are presented. These results show the state-of-the-art of simulation capability of a river oil spill model. Ashland oil spill is the largest inland oil spill in the United States and a reasonable set of observed data are available for comparison with the simulations. The model ROSS2 has been applied to the Ohio-Monongahela-Allegheny River system that consist of a total length of 540 Km (Shen et al., 1993). January 1988 Ashland oil spill has been simulated using the parameters shown in Table 2. Details of parameters involved, discharge and water levels at upstream and downstream boundaries, uncertainties of the data, and limitations of the model are discussed in detail by Shen et al. (1993). Identifying the leading edge position in the field was rather subjective. Field observations by Vicory and Ahles-Kedziora (1989) and Berkey, et al. (1989) gave different values of observed leading edge positions. The simulated leading edge position is defined as the position of the most downstream oil parcel. The simulated positions are taken to be the positions at 17:00 hrs of each day. The results from the simulation and its comparison with available observed data are shown in Fig. 3a and b. Fig. 3 a shows the comparison of leading edge position while Fig. 3b shows the comparison of locations where peak concentrations occurred on a daily basis (e.g. 1/2 refers to the position on January 2). These comparisons show that the modelis capable of simulating the transport of oil with good accuracy each day.

776



Fig. 3. Comparison of simulated and observed slick positions: a) leading edge, b) peak concentration.


777

Table 2. Input parameters for simulation Parameter

Ashland oil spill

Spill Site

W. Elizabeth, PA

Oil v o l u m e / t y p e

2668 m 3 of Diesel

Spill condition

30% underwater 70% on surface d u r a t i o n = 1 hr (major part)

Wind Temperature


Grid size A x x A y Time step A i

16 K m / h r from S.E. -3°C 61m x 61m 15 min

Effect of Different Processes The models reviewed here have been applied to different geographic regions and the type of input data required for the hydrodynamics part varies significantly from one model to another. Therefore, a comparison of the relative performance of the different models reviewed here are not possible. However, and illustration of the relative effects of different processes that were approximated or neglected in some of the models is made here. In this demonstration, three computer simulations have been made using the model ROSS2 for St. Lawrence River. All three simulations are for an oil spill of 189 m3 originating entirely on the water surface. The weather conditions used in the simulation were 15 °C air temperature and wind at 14.3 Km/hr from the West for the first 3 hours of the simulation and 14.3 Km/hr from the Southwest for the rest of the simulation. All three simulations include the processes of advection, horizontal diffusion, and mechanical spreading. The first simulation was done with no evaporation, no dissolution, and no mixing. The second simulation was done with the same hydraulic and wind conditions with vertical mixing included but no evaporation or dissolution. The third simulation included evaporation and dissolution but no mixing. The results from these three simulations are shown in Fig. 4. Fig. 4a to c correspond to the surface oil distribution from simulations 1, 2, and 3 respectively. Fig. 4d shows the oil distribution in the water column for simulation 3. The four slick positions shown in each figure correspond to times of 1 hr, 4.5 hrs, 7 hrs, and 9.5 hrs after the spill. The results of the first simulation is similar to what one would get from a two-dimensional surface transport model that does not include any weathering processes. The second simulation which includes evaporation and dissolution but no mixing of oil into the water column is similar to the results from a surface transport model with the weathering processes of evaporation and dissolution. It clearly shows a smaller slick size and reduced oil concentration compared with Fig. 4a. Although the amount evaporated can vary drastically depending on the type of oil, wind conditions, and air temperature the comparison between Fig. 4a and b provide a qualitative view of the relative importance of evaporation. Dissolution and Evaporation are competetive processes, but the amount dissolved is typically much smaller than the amount evaporated. The oil evaporation rate considered here is medium level and the wind conditions are light.

778



Fig. 4. Relative effects of oil spill processes: a) No evaporation, no dissolution, no mixing; b) with evaporation and dissolution, no mixing; c) no evaporation, no dissolution, with mixing; d) same as case c) in suspended layer (water column).


779

Fig. 4c and d shows the distribution of surface oil and oil in the water column respectively. This simulation includes vertical mixing and resurfacing but no evaporation or dissolution. Comparing Fig. 4a en c provides a qualitative comparison of the effect of vertical mixing and resurfacing. Vertical mixing occurs as a result of oil slick break up into droplets and entrainment into the water column. Some of them will eventually resurface. This effect is clearly demonstrated in the simulation shown in Fig. 4 a as evidenced by the extended tail of the oil slick at 4.5, 7, and 9.5 hrs. Such behaviour is consistent with observed oil slicks.


Summary Oil spill models are used on a real time basis to assist the containment and recovery of oil, study scenarios of possible spills to assist in the development of contingency plans and assess likely environmental impacts, detailed environment impact analysis after a spill, and to help to guide field data collection. Recent emergence of user friendly models with advanced algorithms are powerful tools available for many oil spill related uses as described above. However, the user friendly integrated models allow the user to use them with virtually no understanding of the algorithms and the limitations of the model. Oil spill models are interdisciplinary and involves the simulation of highly complex phenomena. In modelling these, each modeller is constrained by scientific limitations of the knowledge base and implementation. The user must learn these limitations. Failure to do so may result in misinterpreted or erroneous results. For river spills the ROSS2/ ROSS3 models or NRDAM are the best choices. ROSS and WPMB models can simulate surface oil spills. They are of limited use for many present needs. RIVERSPILL is one-dimensional and out of date. NRDAM provides tools for damage assessment and cost benefit analysis for restoration of impacted resources. It's velocity data is based on average conditions adjusted for the given condition through interpolation. Therefore, it is not ideally suited for real time simulation especially if the river conditions vary hydraulically and unsteady transients need to be considered. NRDAM has a good user interface and is combined with GIS data. ROSS3 is a completely integrated model that allows the user to interactively input data, run the hydraulic or oil spill models, and visualize the data including the progress of the computations. It has a unique way of utilizing a non-rectangular co-ordinate system based on bathymetry. It can simulate oil spills under unsteady flow conditions, and for places where shoreline shape and position moves because of the fluctuations in water levels. It is best suited for handling complex shapes and obstructions found in rivers. ROSS2 has integrated features and is easy to use. Since ROSS3 supercedes ROSS2 no further description is necessary. In this paper, available river oil spill models are reviewed along with brief discussions on oil spill processes and the state of the art of their implementation. Processes included in each model are discussed. Model performance is demonstrated through a hindcast case study of ROSS2 application to Ashland oil spill. Further comparison of the relative effects of some oil spill processes are illustrated through simulation. Acknowledgements The authors work on oil spill modeling was made possible due mainly to several projects that were sponsored by U.S. Army Corps of Engineers and the St. Lawrence Seaway Development Corpora780

JOURNAL DE RECHERCHES HYDRAULIQUES, VOL. 32. 1994, NO. 5

tion. Students Mark Petroski, DeSheng Wang and Keerthisiri Angammana contributed in many ways and requires special mentioning. They would also like to thank many others who are too numerous to mention, but have helped in this work. This paper was prepared during the first author's sabbatical leave at the Science University of Tokyo, Japan.

References ARAVAMUDAN, K., RAJ, P., NEWMAN, E., and TUCKER, W„ Break up of Oil on Rough Seas - Simplified Models and Step-by-Stcp Calculations, Report No. CG-D-69-79, U.S. Coast Guard, Washington, D.C., 1979. BERKEY, E., PRICE, R.L., CREEGER, S.M., and JOHNSTON, W.M., Assessment of environmental effects from the January 2, 1988 diesel oil spill into the Monangahela River, Proceedings of Pittsburgh Oil Spill, Pittsburgh, PA, 1989.


CHENG, R.T., CASULLI, V. and MILFORD, S.N., Eulerian-Lagrangian solution of the convection-dispersion

equation in natural coordinates, Water Resources Research, 20 (7) 1984, pp. 944-952. COHEN, Y., MACKAY, D., and SHIU, W.Y., Mass transfer rates between oil slicks and water, The Canadian Journal of Chemical Engineering, Vol. 58, October 1980. COX, J.C., and SCHULTZ, L.A., The containment of oil spilled under rough ice, Proceedings of the 1981 Oil Spill Conference, American Petroleum Institute, Washington, D.C., 1981, pp. 203-298. DELVIGNE, G.A.L., and SWEENEY, C.E., Natural dispersion of oil, Oil and Chemical Pollution, 4, 1988, pp. 281-310. DELVIGNE, G. A. L., Natural Dispersion of Oil By Different Sources of Turbulence, Proceedings, Oil Spill Conference - 1993, Sponsored by U.S. Coast Guard, EPA, and API, Tampa, FL, March, 1993 pp. 415-419. FANNELOP, T.K. and WALDMAN, G.D., Dynamics of oil slicks, American Institute of Aeronautics Journal, 10, 1972, pp. 506-510. FAY, J.A. The spread of oil slicks on a calm sea, Oil on the Sea, (ed. by D. Hoult), 1969, pp. 53-64, Plenum, New York, USA. FAY, J.A. Physical processes in the spread of oil on a water surface, Proceedings of the Joint Conference on Prevention and Control of Oil Spills, Washington, D.C., June, Amer. Petrol. Inst., 1971, pp. 463-467. FlNGAS, M., and SYDOR, M. Development of an oil spill model for the St. Lawrence River, Technical Bulletin No. 116, Inland Waters Directorate, Water Management and Planning Branch, 1980, Ottawa, Canada. FISCHER, H.B., LIST, E.J., KOH, R.C.Y., IMBERGER, J. and BROOKS, N.H., Mixing in Inland and Coastal Waters, Academic Press, New York, 1979. FODA, M., and Cox, R.G., The Spreading of thin films on a water-air surface, Journal of Fluid Mechanics, Vol. 101, parti, 1980, pp. 33-51 GUNDLACH, E.R., and HAYES, M.D., Vulnerability of coastal environments to oil spill impacts, Marine Tech. Society Journal, Vol. 12, No. 4, 1978, pp. 18-27. GUNDLACH, E.R. and BACA, B.J., Sensitivity of Shoreline Environments and Wild Life to Spilled Oil St. Clair River, Lake St. Clair, and Detroit River: A Coastal Resource Atlas, Prepared for Environmental Protection Service, Environment Canada, Toronto, Ontario, Canada, 1986. HAMILTON, G.D. SOILEAU, C.W., and STROUD A.D., Numerical modeling study of Lake Pontchautrain, Journal of the Waterways, Ports, Coastal and Ocean Division, ASCE, 108, 1982, pp. 49-64. HOULT, D.P. and SUCHON, W., The spread of oil in a channel, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, May 1970. HOULT, D.P., Oil spreading on the sea, Annual Review of Fluid Mechanics, 4, 1972, pp. 341-367. HUANG, J.C., A review of the state-of-the-art of oil spill fate/behaviour models, Proceedings of the 1983 Oil Spill Conference, American Petroleum Institute, Washington, D.C., 1983, p. 313-322. HUNG, S.C., Seaway emergency response plan, 1991 Oil Spill Conference, San Diego, Mar. 1991. HUANG, J.C and MONASTERO, F.C., Review of the State-of-the-Art of Oil Spill Simulation Models, Final Report submitted to the American Petroleum Institute, Raytheon Ocean Systems Company, 1982. KAISER, T., GODON, W., WHITEHORNE, R., GUNDLACH, E.R., BACA, B.J., Canadian-U.S. Spill Response

Cooperation Along the Great Lakes, Proceedings, Oil Spill Conference, Baltimore, Maryland, 1987. LEENDERTSE, J.J., A water-quality simulation model for well-mixed estuaries and coastal seas, Volume I, Principles of computation, The Rand Corporation, Santa Monica, California, Report RM-6230-RC. Lu, C.Y. and POLAK, J., A study of solubility of oil in water, Environmental Protection Service, Canada, Report EP5-4-EC-75-1. MACKAY, D., and MCAULLIFFE, CD., Fate of hydrocarbons discharged at sea, Oil and Chemical Pollution, 5, 1988, pp. 1-20. MACKAY, D., PATERSON, S., and NADEA U, S., Calculation of the evaporation rate of volatile liquids, Proceedings of the 1980 National Conference on Control of Hazardous Material Spills, Louisville, Kentucky, May, 1980, pp. 361-368. MACKAY, D„ PATERSON, S. and TRUDEL, K, A Mathematical Model of Oil Spill Behaviour, Report submitted to Environmental Protection Service, Fisheries and Environment Canada, 1980.


781


MACKAY, D., BUIST, I., MASCHARENHAS, R., and PATERSON, S., Oil Spill Processes and Models, Report EE-8 submitted to Environment Canada, University of Toronto, Toronto, 1980 MADSEN, O.S., A Realistic Model of the Wind-induced Ekman Boundary Layer, J. of Physical Oceanography, 7(2), pp. 248-255. MIKLAUCIC, E.A. and SASEEN, J., Oil spill, Floreffe, PA - Case history and response evaluation, Oil Spill Conference, San Antonio, Texas, 1989 MOORE, S.F., DWYER, R.L., and KATZ, A.M., A preliminary assessment of the environmental vulnerability of Machias Bay, Marine, to oil supertankers, Parson Laboratory, M.I.T., Report No. 162, 1973. OWENS, E.H., TAYLOR, E., MARTY, R„ and LITTLE, D.I., An inland oil spill response manual to minimize adverse environmental impacts, Proceedings, Oil Spill Conference - 1993, Sponsored by U.S. Coast Guard, EPA, and API, Tampa, FL, March, 1993 REED, M., FRENCH, D., and FENG, S., Natrural Resource Damage Assessment Model, Proceedings, ASCE National Conference on Hydraulic Engineering, Nashville, TN, July 1991. REED, M., and FRENCH D., A transportable system of models for natural resource damage assessment, Environmental Software, 1993, Elsevier Science Publishers Ltd., p 191-201. Research Planning Institute, St. Lawrence River - supplement to the joint Canada-United States Marine Pollution Contingency Plan for spills of oil and other noxious substances. Columbia, S.C., 1986. ROACHE, P.J., Computational Fluid Dynamics, Hermosa Publishers, Albuquerque, N.M. 1972. SAYRE, W.W., and CHANG, F.M., A Laboratory investigation of open-channel dispersion processes for dissolved, suspended, and floating dispersants, U.S.G.S. Professional Paper 433-E, 1969, p. 71. SHEN, H.T., YAPA, P.D., WANG, D.S. and YANG, X.Q., A Mathematical Model for Simulating Fate and Transport of Oil Spills in Rivers (ROSS2), CRREL Special Report 93-21, Cold Regions Research and Engineering Laboratory, U. S. Army Corps of Engineers, Hanover, NH, July 1993. SHEN, H.T. and YAPA, P.D., Oil Slick Transport in Rivers, Journal of Hydraulic Engineering, ASCE, 114(5) May 1988, pp. 529-543. SHEN, H.T., YAPA, P.D., and PETROSKI, M.E., A Simulation Model for Oil Slick Transport in Lakes, Water Resources Research, 23(10) 1987. SPAULDING, M.L., A State-of-the-Art Review of Oil Spill Trajectory and Fate Modeling, Oil and Chemical Pollution, Vol. 4, 1988, pp. 39-55. STOLZENBACH, K.D., et al., A Review and Evaluation of Basic Techniques for Predicting the Behaviour of Surface Oil Slicks, Report No. 222, Ralph M. Parsons Laboratory, Department of Civil Engineering, M.I.T., Cambridge, MA, 1977. TSAHALIS, D.T., Contingency Planning for Oil Spills: Riverspill - A river simulation Model, Proceedings, Oil Spill Conference - 1979, Sponsored by U.S. Coast Guard, EPA and API, Washington, D.C., March, 1979. VlCORY, A.H., and AHLES-KEDZIORA, L„ Interstate in-stream tracking of the January 1988 Ashland oil spill to the Monangahela and Ohio rivers, Proceedings of Pittsburgh Oil Spill, Pittsburgh, PA, 1989. WALDMAN, G.A., JOHNSON, R.A., and SMITH, P.C., The spreading and transport of oil slicks on the open ocean in the presence of wind, waves, and currents, Coast Guard Report No. CG-D-17-73, AD-763 926, July 1973. WEEKS, W.F., and DINGMAN, S.L., Thermal modification of river ice covers: Progress and problems, Proceedings, The Role of Snow and Ice in Hydrology, Banff, 1972, p. 1427-1435. YAPA, P.D., THOMAS, R.J., RUTHERFORD, R.S. and SHEN, H.T., A Microcomputer Model for Oil Spill Simulation (MICROSS), Journal of Computing in Civil Engineering, ASCE, January. YAPA, P.D., SHEN, H.T., DALY, S., and HUNG, S.H., 1991. Oil Spill Simulation in Rivers, Proceedings, Oil Spill Conference - 1991, Sponsored by U.S. Coast Guard, EPA, and API, San Diego, CA, March, p. 593600. YAPA, P. D., SHEN, H.T., WANG, D.S., and ANGAMMANA, K., 1992. An Integrated Computer Model for Simulating Oil Spills in the St. Lawrence River, Journal of Great Lakes Research, 1992, 18(2), p. 34-51. YAPA, P. D., SHEN, H.T., and ANGAMMANA, K. Modeling Oil Spills in a River-Lake System, Journal of Marine Systems, 1994(4), Elsevier publications, pp. 453-471. WEERASURIYA, S. and YAPA, P.D., Uni-directional Spreading of Oil under Solid Ice, Canadian Journal of Civil Engineering, February 1993, pp. 50-56. YAPA, P.D. and CHOWDHURY, T., Spreading of Oil under Ice Covers, Journal of Hydraulic Engineering, ASCE, December 1990, pp. 1468-1483. YAPA, P.D. and BELASKAS, D.P., Radial Spreading of Oil under and over Broken Ice: An Experimental Study, Canadian Journal of Civil Engineering, December 1993, pp. 910-922.

782
