Click for next page ( 26


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 25
State-of-the-Art Overview: Physical Oceanographic Processes, Features, and Methods of Potential Importance to the ESP INTRODUCTION This chapter provides a state-of-the-art overview of available information on the major issues reviewed by the Physical Oceanography Panel and considered to be of potential importance in meeting the MMS requirement to predict the risk to the environment from OCS oil activity. It includes for each, as feasible, an assessment of the present state of knowledge and of the information of particular importance to the ESP and MMS and an indication of the major research needs considered most likely to enhance the current knowledge. The material is arranged by subject matter and it is in no way intended to represent priorities, priorities will differ according to the physical setting. The panel has focused on the movement of oil by the water, considering, where appropriate, the effects of ice and of sediment. Its findings have been grouped for discussion into the following major sections: Transport, Stirring, and Mixing Processes; Numerical Models; Sea Ice; and Sediment Transport. The Problem of Assessing Impacts of Oil Exploratiom A Physical Oceanographic Perspective Before discussing the state of knowledge of the physical oceanography of a region and the adequacy of this information for impact assessments of oil and gas exploration or production, it is appropriate to consider the specific physical information that is needed. The problem of predicting the movement and concentration of material released into the ocean can be formally stated as follows: Given a source of some material (e.g., oil, gas, or routine discharge) as a function of space and time, what is the probability that the material's concentration at a particular spatial point and time will be greater than some specified value? In addition, it is also necessary to know how probable it is that the flux of the material into the sediments at a particular point and time will exceed some given value, and likewise to know the same for the flux into the atmosphere. The primary physical oceanographic processes that must be considered in predicting the movement of material released into water are: 1. Advection or transport: These terms refer to flows that move patches of material around but do not significantly distort or dilute them. 2. Stirring: This is the process whereby flows with strong shear and strain fields on the scale of the patch size generate "streakiness," with tendrils of material from the patch drawn out into unpolluted water and streaks of water intruding into the patch. By itself, stirring does not alter concentrations, although it affects the probability of finding material at a particular point. 3. Mixing: This process is responsible for the decrease in concentration of material. At the most fundamental level, mixing is accomplished by molecular diffusion intermingling water with other molecules. However, molecular mixing is usually coupled with stirring to produce 25

OCR for page 25
26 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF turbulent mixing wherein stirring produces concentration gradients on scales small enough where molecular mixing can efficiently erase those gradients. As discussed below, estimates of turbulent mixing rates are scale-dependent (Eckart, 1948~. Collectively, these three processes are referred to as "exchange." Both horizontal and vertical exchange must be considered since, for flow fields with a complex spatial structure, exchange in a particular plane can be dependent on velocities in the orthogonal direction. In addition, the density of the material and biological and chemical processes can play roles in the probability problem stated above. If different from that of the ambient seawater, the material's buoyancy can result in transport and mixing at rates that differ from those of water parcels (e.g., sinking, accumulation in surface convergence zones, and differential wind drifts). Biological and chemical processes can produce effective sources and sinks of particular materials and introduce additional exchange mechanisms (e.g., adsorption to sinking particles). Transport Processes in the Water Column The fate of biological, chemical, and sedimentary constituents in the coastal zone results from a convolution between transport processes and the mechanical and chemical properties of the various constituents. Coastal circulation and the attendant variability in physical parameters characterizing the coastal ocean result from complex interactions between processes with a broad range of time scales, from interannual periods to surface-gravity-wave periods of a few seconds. As a consequence of these diverse motions, describing the circulation is both challenging and expensive. Oil and pollutants are carried from one place to another by currents. But surface spills are also moved relative to the water by the wind. Waves break up and mechanically modify surface spills and drive the modified material below the surface. The material drifts with subsurface currents sometimes to reappear later at the surface under calmer conditions. Products from surface spills and effluents from drilling operations or from subsurface leaks (from pipelines or blowouts) may ultimately end up in bottom sediments, possibly accumulating to unacceptably high concentrations in localized regions. They may even be transported from place to place within the sediments over long periods. All of these processes are of potential importance in estimating the fate of spilled or leaked material. The first and second processes, advection by currents and wave effects, are of major importance in the immediate translation and dispersal of a spill. Sediment Transport Processes The physical processes responsible for the deposition, mixing, resuspension, and transport of bottom sediments are most closely tied to the long-term effects of petroleum exploration, development, and production. A portion of the chemicals of environmental concern emanating from drilling activities, discharge of coproduced waters, and oil spills eventually passes to the bottom by adsorption to fine, suspended particulates or by incorporation into detrital materials, which settle out in regions or during periods of deposition (see, e.g., NRC, 1985; U.S. DOI, laded). The subsequent fate of the particulates and the associated chemicals is then largely determined by patterns of physical mixing, resuspension, and transport. Vertical mixing and resuspension of surface sediments tend to disperse initially high concentrations of contaminants and to increase chemical interactions between particulate and dissolved phases (see, e.g., Bothner et al., 1987~. Horizontal transport often leads to further dispersal and lower contaminant concentrations (NRC, 1983), but it may also lead to the physical concentration of contaminated particulate material in depositional environments. Toxics in the bottom sediments, pore waters, and material suspended just above the bottom may then enter the benthic food web, depending on the bioavailability of the material to the local benthic community (Boesch et al., 1987; Howarth, 1987; Neff, 1987~.

OCR for page 25
STATE-OF-THE-ART OVERVIEW Sedimentary accumulation and subsequent release of toxics may prolong the impact of a spill or discharge long past the initial occurrence. Thus, physical processes responsible for the deposition, mixing, resuspension, and transport of bottom sediments are closely tied to the long- term effects of petroleum exploration, development, and production. Boesch et al. (1987) have defined long-term effects to include effects that persist for a long time as a result of some brief activity and effects that result from low-level, chronic exposure over a long period. Examples of the former include oiling of sediments or sedimentary accumulation of undegraded hydrocarbons in the aftermath of an oil spill and the impact of drilling muds and cuttings from exploratory drilling. Examples of the latter include chronic releases of oil during production and repeated discharges of drilling muds and cuttings during development. In all cases, impacts are likely to be worst in shallow-water, depositional environments (Boesch et al., 1987; Howarth, 1987~. The effects of chronic discharges on the deeper depositional environments of the OCS are still largely unknown, however (NRC, 1983; Boesch et al., 1987; Neff, 1987), because of the difficulty of separating long-term effects from natural environmental variability. Space and Time Scales Oceanic flows have energy at many different space and time scales. Physical oceanographers often discuss motions in different frequency bands separately, as this panel does below. Although this is convenient for organizing information and understanding the mechanisms involved, care must be taken in superimposing different frequency bands to obtain the total flow field. The Fourier decomposition of a current-meter record can be recombined to give the flow versus time; however, band-pass-filtered records of currents and pressure (for example) will not satisfy the Navier-Stokes equations when there are significant nonlinearities in the flow. The problem becomes even more severe when looking at the movement of particles in the flow the Lagrangian description of the motion because the evolution equation for particle position involves a nonlinear function (the flow velocity) of the position. Simple Eulerian flow fields varying in time and space with a single frequency and wave number give particle motions with a complex spectrum, containing both harmonics and a zero-frequency component. The latter corresponds to a net drift rate for a particle- a Lagrangian mean flow which is different from the average velocity measured at a point (the Eulerian mean). The difference is called the Stokes velocity (see, e.g., Longuet-Higgins, 1969~. Flows only slightly more complex can lead to chaotic particle trajectories and efficient turbulent mixing (see, e.g., Zimmerman, 1986~. When the Eulerian flows have a broad frequency spectrum, the Lagrangian motions become even more complex and can have a spectrum quite different from the Eulerian one. The probability of a particle entering a particular volume of space can depend upon the flows in all parts of the Eulerian spectrum; of particular concern are those bands in frequency and wave-number space that are not resolved by a given model. The dependence of stirring and mixing on the complex relationship between the Lagrangian and Eulerian spectra implies that turbulent mixing is scale-dependent: the inferred rate of mixing depends strongly on the range of scales that are resolved. In addition, turbulent diffusion processes do not always transport material at a rate proportional to the larger-scale gradient, nor is the flux vector necessarily parallel to the mean gradient. Although it is almost universal practice to model subgrid-scale exchange processes as a kind of diffusion, that practice may be inappropriate, especially in a region with strong and variable topography and density fronts. Forcing Mechanisms Predictive capability is usually premised on the identification and understanding of the mechanisms that couple response to forcing. The preceding section has illustrated that the coastal ocean is subjected to forcing over a broad range of periods, ranging from interannual variations in the coupled ocean-atmosphere system (for example, the E1 Nino-Southern Oscillation (ENSO) 27

OCR for page 25
28 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF process) to the atmospheric forcing responsible for the generation of surface-gravity waves. Some forcing mechanisms are better understood than others: the forcing imposed by the barotropic tide on the continental margins is probably the best-understood forcing mechanism, and the influence of adjacent deep ocean currents and eddies may be the least-understood forcing mechanism. Each mechanism or process responsible for forcing the coastal ocean is modulated as a function of space and time. Predicting coastal circulation and its statistics thus entails a knowledge of at least the amplitude and variation of the processes that drive the coastal ocean. For example, currents are often observed to converge in the vicinity of Cape Mendocino, California; there, the convergence results in an offshore transport of coastal waters. This process is of obvious importance in determining the path of water masses initially on the shelf. Whether this convergence results from offshelf oceanic processes or simply reflects spatial variations in the wind that forces the coastal ocean is not known. Oil Spill anti Circulation Models The above points regarding mixing and transport have important implications for the models used in oil-spill-risk analysis. Generally, the models resolve only a limited set of scales, often just the seasonal mean circulation. In the absence of most of the temporally and spatially varying parts of the spectrum, the predicted Lagrangian motion may miss many aspects contributing to drift, especially on the shorter time scales. The OSRA model used by MMS deals only with inert surface-layer material, although MMS has sponsored some work involving simultaneous calculation of the fates of the oil a prediction of some of the chemical and physical changes in the hydrocarbons. This report focuses primarily on the prediction of exchange of passive materials; it is likely, however, that other processes are also important. The OSRA model deals with a point patch (a material particle only) and does not resolve mixing processes or, given the lack of small-scale detail, much of the stirring process either. Different realizations of the random aspects of the movement of of} spills come only from wind drift variability, not from the oceanic currents. Vertical redistribution of the material by turbulent mixing is not included, although this may result in dilution, reduced evaporation, different transport (due to vertical shear in the horizontal currents), and enhanced horizontal mixing (e.g., vertical shear dispersion). These points indicate that, in assessing the adequacy of a practical model for a task such as oil-spill-risk analysis, it is necessary to evaluate the potential transport, stirring, and mixing caused by many different processes. It is important to recognize that all models are inherently limited in their predictive capability. Lorenz (1969) demonstrated that a model calculating from initial conditions derived from data would diverge from the actual system within a finite time. Two factors were responsible: errors in measurement of the flow (and other physical quantities) and uncertainties in the values at points where no measurements were taken. While the predictive capability of a model depends on the dynamics, the physical processes incorporated in his model have similarities to those acting in the atmosphere and in the ocean. Although using new data to readjust the model ("data assimilation") can greatly improve the predictions, it cannot eliminate the errors (as is obvious from weather forecasting experience). Errors in model dynamics and in the forcing parameters applied will also limit the model's predictive capabilities. Diminished predictability also occurs when an attempt is made to extend information into a region where inadequate or no data exist. The ability to predict the trajectory of an actual spill is certainly important for spill containment and management; thus, the extent of our ability to make such predictions is certainly relevant to leasing decisions. But there is also another related question: how well can we predict the statistical variability of dispersal? Failure of a model to predict individual trajectories does not necessarily mean that the statistics produced are wrong; for example, radioactive decay cannot be predicted at all, yet models that describe the statistical probability of such events work extremely well. How well fluid dynamical models will reproduce the statistics of trajectories in the ocean is not known. Frisch and Orszag (1990) caution, ". . . it is well known that detailed properties of turbulent flows at far-off times cannot be predicted. However, even the statistical

OCR for page 25
STATE-OF-THE-ART OVERVIEW 29 properties of these flows may be 'uncomputable.' . . . [This] would imply, in the context of meteorology for example, that while the weather clearly is not predictable at long times, neither, in fact, is the climate." (Note that "far-off" is measured in the time scales of the dominant motions as described above and may be only on the order of days.) Again, the capability to predict statistical probabilities of spill trajectories will depend on the nature of the dynamics of a system, the degree to which the model resolves different scales, and the reliability of the input that describes the forcings and boundary conditions. It is simply not known how well even an optimal model can do. It is important to stress that data are needed both as independent estimators of trajectory statistics and as input and verification for modeling. Scope of the Overview The content of this chapter is restricted in two ways. First, attention is focused on physical oceanographic processes that are of direct importance to the motion and fate of oil in oceanic waters. Primary emphasis is given to processes that control the advection of oil in surface and near-surface waters. Second, only processes that are active over the OCS are considered, because these are the waters that are-under federal control; Processes that are .... specll~lc to nearshore (l.e., state-controlled) waters, such as In nays and estuaries, are not - included. Although nearshore processes are not represented in the OSRA model, oil is assumed to hit the shoreline if it reaches particular sections of a grid (imposed on an area map) that encompasses the shoreline. These sections cover areas extending well into OCS waters (see Fig. 5~. These restrictions reflect the primary bias of ESP physical oceanography. They do not imply that nearshore and benthic processes are unimportant in a full consideration of the ecological impacts of oil spills but simply place such processes beyond the purview of this review. As a consequence of surface concentration of oil and relatively rapid weathering, the principal physical oceanographic problems that must be addressed are understanding and predicting the motion of oil in surface waters over periods of up to about 30 days. During this 30-day period, response to wind forcing is very important to the net motion and variability of spill trajectories. Small-scale spatial and temporal processes in the near-surface environment (e.g., fronts, convergence zones, Langmuir cells, shingles, and interleaving) can affect the course of spill movement and alter spill dynamics substantially. The variability of underlying currents within this time frame is also important. Flows with temporal scales substantially greater than a month contribute to individual mean trajectory paths but do not contribute substantially to the variability of an individual trajectory over the time frame of interest. Tidal motion is important only insofar as it contributes to mean motion (through tidal rectification), affects smaller-scale dynamics (e.g., through the generation of internal waves), and is a contributing factor to horizontal dispersion (through stirring). For subsurface transport of spilled oil, the time frame of interest is extended to 30 to 90 days following the spill's release. This extension is made to account for the dilution of subsurface oil associated with relatively large spill events to low concentrations (<1 ppb). The first 30 days after the oil is released remaids of greatest interest, however, because this is the period when oil is most toxic and has the largest impact on the environment. .1 HANSPORT, STIRRING, AND MIXING PROCESSES Wind Stress Drag Coefficient and Space and Time Resolution Determination of surface-wind stress depends on (1) determination of the wind speed and direction at the appropriate space and time scales and (2) knowledge of the factors necessary to transform the wind speed into wind stress. This latter is usually accomplished by determining a drag coefficient, which depends on several oceanographic and meteorological parameters. For example, Walsh et al. (1986) showed that stress and near-surface drift determinations are sensitive not only to the geostrophic wind at the top of the Ekman layer, but also to the air and surface

OCR for page 25
30 ~ - ~' W-- l W:- ~m, :\ \3 \ me\ - - - ,/ Al ~ i : Id -~< ado: a: ~ ~ ~ ~ : 0 ~ a. ~ ~ ~ 0 ~ ~au~oaa ~ al J O ~_, - - ^~ ~ O ^ ~ ~ ~ O hi C. - - me:- a - - ~ - - _ - - 00- ~ - ~ ~ ~ ~ ~ _ ~ ~ Y ~ ~ O ~ 1 ~ ~ ~ ~ 0(I ~ ~ .4 ~ ~ O -~ 0 ~ ~ ~ ~ ~ ~ ~ ~- ~ ~ ~ ~ ~ a. ~ a. ~ ~ ~ ~ ~ ~ ~ 0 ~ ~ ~ ~ ~ ~ o In - " of To - - C ~also 0~ - "~0~. 0 0 C. ~ ~ ~ < ~ ~ C ~- \ Z O ~O O ~ N _ it' ~ , - ~ o o^- _ o8- lo . . _ _ Am, Yew .~_ O O ~Q . em-, l _ ~ o U) o .s ~3 .> ._ . . - oo ox Cal o U] C~ . . = V) = ; c: 8 ~ C;S Ct ~ _~ ._ ~ o Z ~ ~ o .= ~ O o ~ ~ . - Ct - C\~ ~ a84 ~=

OCR for page 25
STATE-OF-THE-ART OVERVIEW 31 ocean temperatures, the mean horizontal temperature in the planetary boundary layer, and the surface roughness. Their work is a good example showing the significant sensitivity of numerical model results to the data used as input. The difficulties of converting wind speed and direction to wind stress by determining a drag coefficient usually can be overcome. However, specification of the wind field based on observations from coastal stations, ships, sea-level atmospheric pressure, low-level cloud motions, satellite microwave scatterometry, and instrumented buoys usually is inadequate and is the limiting factor to determining the surface stress distribution. The nature of the difficulty in relating the wind at an offshore location to that observed at a coastal station is described by Weisberg and Pietrafesa (1983~: The surface wind field over the South Atlantic Bight . . . varies on seasonal, synoptic, and diurnal time scales . . . over the entire region while the sea breeze induced diurnal oscillations are coherent only over the coastal area.... both the synoptic and sea breeze oscillations were found to be seasonally modulated.... The coherence between stations was also found to be seasonally modulated, with winter time synoptic scale fluctuations being coherent over the entire [South Atlantic Bight] . . . while only marginal coherence occurs in the summer. A distinct seasonality therefore exists in both the ability to predict offshore winds from coastal station data and in the matrix of linear operators . . . used for that prediction. Since the structures of the synoptic disturbances change as they progress offshore, the matrix of linear operators depends upon the vector wind at the coast and not just a single component of that vector.... During the fall season, the time series are significantly coherent for time scales longer than 1.5 days; the phases are very nearly zero; and the predicted series are underestimated by as much as 30~0 percent in amplitude with somewhat better results for u (east/west velocity) than for v (north/south velocity). During the winter season, the time series are most coherent at time scales longer than 2 days, the phases are very nearly zero, and the amplitudes are either underestimated or overestimated by as much as 30 percent. (Copyright 1983 by the American Geophysical Union.) Another example of wind field complexity is presented by the detailed observations of the Coastal Ocean Dynamics Experiment (CODE) (Beardsley et al., 1987), which showed significant temporal and spatial structure to the wind field off northern California during the upwelling season. Their measurements showed . . . after the atmospheric spring transition the airflow in the marine layer is dominated by the North Pacific high, and the surface wind field over the shelf is characterized by periods of strong (7-15 m/s), upwelling favorable alongshelf winds lasting for up to 30 days, interrupted by shorter periods of much weaker winds directed either equatorward or poleward. These periods of weak or reversed winds typically last several days and are called wind relaxations, even though they are primarily associated with coastally trapped perturbations of the marine layer along the central and northern California coast and not with a large-scale weakening of the North Pacific high. The atmospheric boundary-layer measurements made in CODE suggest a simple conceptual model which can explain much of the physiology or structure of the marine layer and associated surface wind field during periods of persistent upwelling-favorable winds. During these periods, which represent the quasi steady state regime during the upwelling season, the inversion base of the marine layer drops eastward towards the coast until it intersects the coastal mountain range at a height of several hundred meters, and the associated thermal wind produces an along-shelf wind jet which has a maximum speed just below the inversion base. Turbulent mixing tends to homogenize any stratification in the marine layer beneath the jet and couple the jet to the ocean surface, producing strong upwelling-favorable winds over the shelf. Day/night heating/cooling over the narrow coastal strip beneath the marine layer generates a weak cross-coast secondary circulation which causes the core of the along-shelf jet to drop in elevation and shift onshore. This diurnal change in the marine layer structure explains both the daytime acceleration of the surface winds observed over and near the coast and its offshore decay and the associated offshore increase in the subdiurnal along-shelf wind. Thus, the quasi-steady component of the wind stress has a significant curl (up to 1 m/e/km in wind speed observed) over the inner shelf during periods of active upwelling. This mean summer atmospheric bounda~y-layer regime is occasionally interrupted by synoptic and/or mesoscale events or anomalous conditions. Analysis of the CODE observations suggests five types of events, two primarily synoptic-scale conditions which lead to stronger-than-normal upwelling-favorable winds over the shelf and three primarily mesoscale events which lead to wind relaxation. About half of the wind relaxation events observed in 1981 and 1982 are believed to be associated with either coastal-trapped gravity currents or internal Kelvin waves that

OCR for page 25
32 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF propagate northward in the marine boundary layer along the central and northern California coastal mountain range. (Copyright 1987 by the American Geophysical Union.) Surface Wind Drift and Ekman Dynamics Price et al. (1987) reported on an important verification of the classical Ekman theory of wind-driven transport in the ocean surface layer, based on a careful analysis of upper ocean data from the Long Term Upper Ocean Study (LOTUS) (Briscoe and Weller, 1984~: By assuming that the momentum balance of a steady wind-driven current was between the turbulent stress caused by the wind and the Coriolis force caused by the earth's rotation. Ekman derived the archetypal solution for the vertical structure of a w~nd-driven current ... ,..__ a._ .~ ..... results from [the solution]. The first is that the current profile from Ekman's theory has a spiral structure, called an Ekman spiral, in which current amplitude decays by one e-folding [a factor of 1/e] over a depth D as the current vector rotates to the right [in the northern hemisphere] through 1 radian. Typical values, from observations, are D = 30 m and the eddy coefficient, A = .05 m2/s. However, the range of inferred A covers more than an order of magnitude so that neither A nor D can be regarded as well known. The detailed specific structure of the spiral depends on A being constant in depth and time, which now seems unlikely to hold in the upper ocean.... [Recent] theories yield somewhat different spiral structures, but there is no consensus on, for example, the sense of the depth dependence of A. The structure of the mean wind-driven current thus remains an open theoretical question. A second and fundamental result from the theory is that the vertically integrated current, or volume transport per unit width, is given by the Ekman transport relation.... But just as D is not known beforehand with confidence, neither is [the depth where the wind-driven current vanishes]. However, the magnitude and direction of the transport follow directly from the presumed momentum balance between wind stress and the Coriolis force and are independent of A or any other aspect of vertical mixing.... There have been repeated, but inconclusive, attempts to verify the Ekman transport relation directly by using in situ measurements of wind and currents. Although wind-driven transport more or less to the right of the wind [in the northern hemisphere], its magnitude has seldom been found to be consistent with Ekman transport computed from estimated wind stress to closer than a factor of about 2. This has not been interpreted to mean that the [Ekman transport is not given by the equation] in principle; there are significant technical difficulties in making accurate in situ current and wind measurements, some of which have only recently been appreciated and overcome. There are also analysis and interpretation problems in stying to separate the wind-driven current from the measured current.... By separating the wind-driven current from the measured current and by constructing a coherent average over a long record, [they] find that the Ekman transport relation is consistent to within experimental error. The mean current has a spiral structure qualitatively similar to an Ekman spiral. In this case, however, the scale depth depends on the stratification, and in general the dynamics of the spiral appear to be much richer [more complex] than implied by the original Ekman theory.... The principal results of [their] analysis are that (i) the Ekman transport relation was found to give an estimate of wind-driven transport consistent with the transport estimated from in situ current measurements, and (ii) the mean current was found to have a spiral-like structure that is strongly surface trapped on account of the solar heating and the resulting stable stratification. A simple numerical model that takes into account the important effect of stratification was successful in simulating the diurnal variability of current and the mean current spiral. (Copyright 1987 by the AAAS.) , ~ ThPr`~ are tern n~t~xr^th~r Present-day systems like the LOTUS surface buoy and the VACM instruments make it possible to obtain the kind of data required to build and test models of the Ekman drift with all the natural complexity taken into account. Definition of the Minced Layer Muller and Garwood (1988) defined the mixed layer as follows: The "mixed layer" is [the] part of the upper ocean where temperature and salinity are quasi-homogeneous with depth, according to some appropriate criterion. This layer has to be

OCR for page 25
STATE-OF-THE-ART OVERVIEW distinguished from the "turbulent boundary layer," which is the part of the upper ocean that contains turbulence generated by air-sea interaction processes. Traditionally, it has been presumed that the mixed layer is the vertical extent of an earlier turbulent boundary layer and that therefore the depth of the turbulent boundary layer at any given time and geographical position is less than or equal to the mixed layer depth. This maxim has recently proven to be incorrect (H. Peters, University of Washington, Seattle) . . . it is now very clear that the turbulence, which is at least initiated by the "changes of energy, buoyancy, or momentum across the air-sea interface and is hence properly considered a part of the ocean surface turbulent boundary layer, may penetrate the pycnocline well below what would be deemed the mixed layer by any of the algorithms for determining mixed layer depth. The more traditional measures of mixed layer depth, based solely on temperature profiles, often do not apply in special regions. One such region is the western equatorial Pacific (R. Lukas, University of Hawaii, Honolulu). There are cases in which the temperature profile by itself (neglecting salinity) is clearly hydrostatically unstable. The high-resolution salinity observations reveal that salinity controls both density structure and the often shallow mixed layer depth in this region. Hence the monsoonal rains may play a significant role by stratifying the upper ocean with fresher water that overlays a remotely subducted (or previously created) warmer and saltier mixed layer. (Copyright 1988 by the American Geophysical Union.) Mixed Layer Turbulence According to Muller and Garwood (1988~: Because observations of the vertical fluxes of momentum, mass, and heat are still lacking in the upper ocean, observations of the dissipation of turbulent kinetic energy are still the single best evidence of the intensity of turbulent mixing In the mixed layer. Such observations, if they are of sufficient vertical and temporal resolution, provide information on the depth of mixing. They also yield a measure of entrainment if the net sources of turbulence are known or can be estimated. The net dissipation for the whole mixed layer cannot yet be computed with great precision because it is still not possible to observe dissipation accurately in the top several meters of the ocean, the region that probably has the largest rate of shear production of turbulence. Nevertheless, the order of magnitude of the net dissipation may be computed. Similarity theory, which applies to the atmospheric surface layer, can be used to extrapolate dissipation values from several meters depth to the surface through the unobserved near-surface zone. When this technique is applied to a deep mixed layer that is strongly free convective, it is found that the wind shear production plus the estimated buoyant production of turbulence is inadequate to explain observed rates of mixed layer deepening (M. Gregg, University of Washington, Seattle) . . . Although there are other possible explanations for this discrepancy, occasional observations of "bursts" of very high dissipation rates suggest an additional (previously unexpected) source of turbulent kinetic energy. The phenomenon may be related to the sudden injection of energy from breaking surface waves. If this is the case, the similarity relationship between dissipation and the friction velocity (which is used successfully in the atmospheric surface layer) may be inadequate for the oceanic turbulent boundary layer. Although breaking waves are technically a conversion of mean wave energy to turbulent kinetic energy (a shear production mechanism), the phenomenon may act more like the buoyant transport of turbulence in a free convection regime in that it is not dissipated locally but is transported vertically to the base of the mixed layer and there converted to potential energy by the action of entrainment. There are a number of critical questions. Clearly, profiles need to be extended to the surface. The upward profiler may provide a solution (T. Dillon, Oregon State University, Corvallis). We don't yet have adequate horizontal sampling. Inadequate consideration of horizontal variability in the case of the atmospheric surface layer also caused an apparent breakdown of the expected similarity scaling between dissipation and the surface friction velocity, and the similarity theory was ultimately verified only with adequate sampling (C. Fairall, Pennsylvania State University, University Park). Acoustically detected bubbles may provide a tracer to determine the vertical extent of turbulent transport that is caused by breaking of surface waves (see Thorpe, 1985; also W. Large, National Center for Atmospheric Research, Boulder).... Near the ... surface, bubbles injected downward from the surface following the breaking of surface gravity waves are the main scatterers. The intensity of the backscattered signal hence provides a measurement of the extent of the bubble penetration. Such observations of the bubble envelope, correlated with profiles of dissipation, may shed light on the role of breaking surface waves in mixing and entrainment. 33

OCR for page 25
34 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF When the "dissipation method" is Tended to the ocean, Reynolds stress profiles can be estimated by assuming a balance of dissipation and local shear production. If this method is applied to measurements at the equator, however, it is found that there is a discrepancy between the observed rates of dissipation (and inferred Reynolds stresses) and the assumed sources of momentum (T. Dillon, Oregon State University, Corvallis). More investigation is needed to deternune if in fact there are significant discrepancies in the momentum budget or if the dissipation technique may not be applied to the possibly unique equatorial mixed layer because (for example) the radiation of internal gravity waves becomes a significant part of the turbulent kinetic energy budget. (Copyright 1988 by the American Geophysical Union.) Surface Waves and Stokes Drift Muller and Garwood (1988) summarized: Surface waves are an integral part of mixed layer dynamics. It is generally believed that most of the atmospheric momentum and mechanical flux is first absorbed by the surface wave field. However, it cannot be retained there, and it is quickly dissipated into the underlying ocean by whitecaps and other wave-breaking processes. Considerable effort has been spent to construct models of the evolution of the surface wave field under the influence of wind forcing, nonlinear interaction, and dissipation. One of the most advanced models is that developed by the WAM (Wave Modeling) group (G. Komen, Royal Netherlands Meteorological Institute, De Bilt). The best estimates from this model of momentum and mechanical energy fluxes from the wave field to the ocean are large. For a wind speed of 20 m/s, an energy flux of a few watts per square meter is calculated, which greatly succeeds typical turbulent fluxes estimated below the surface wave zone. Also, the momentum and energy fluxes from the atmosphere to the surface waves and from the waves to the ocean are of the same magnitude, and only a small fraction is used for wave growth. Under certain fetch conditions, the momentum flux into the ocean turns out to be even larger than the downward momentum flux at 10 m height above the ocean. Something seems to be wrong. The discrepancy may be resolved bar changing the model parameters or the spectral parametenzation at high frequencies, but it might also indicate that the atmosphere surface layer is not a "constant flux layer" because of deceleration effects over growing waves. (Copyright 1988 by the American Geophysical Union.) In addition to their role in near-surface mixing, surface waves may contribute to a Lagrangian mean surface drift velocity, known as Stokes drift. Stokes drift results from nonlinearity of the surface wave field and increases with wave height; essentially, particles or passive tracers travel farther forward with the crest of the wave than they travel backward with the trough. Kenyon (1969) has estimated the Stokes-drift velocity as a function of wind velocity by using the directional wind wave spectra for fully developed seas of Pierson and Moskowitz (1964~. He found the ratio of surface Stokes drift velocity to wind speed measured at 19.5 m above sea surface to range from 1.6 to 3.6%, which is large enough to make a significant contribution to the overall surface wind drift. Response to Severe Storms Allen et al. (1987) stated that strong storms cause large flows and increased transports and mixing in coastal areas. For example, although the typical mean flow in the Middle Atlantic Bight is 0.05 m/s, episodic storm currents associated with subinertial motions succeed 0.4 m/s and last for several days. The strength and pattern of the storm-induced flow is not well known and is probably a function of coastal geometry, the size and shape of the storm systems, and the rate at which the storms move and intensify. The effect of the large currents, mixing, and the transport associated with the storms on the shelf budgets and on the transport of material are important unsolved coastal problems. Coupled meteorology and physical oceanography programs [are needed] to understand the detailed cyclogenesis and subsequent meteorological forcing. (Copyright 1987 by the American Geophysical Union.) A recent example of a coupled meteorological and oceanographic experiment was project GALE (Genesis of Atlantic Lows Experiment), which studied these processes over the

OCR for page 25
STATE-OF-THE-ART OVERVIEW southeastern U.S. continental shelf in early 1986. Blanton et al. (1987), reporting on the oceanographic studies carried out during GALE, wrote: The GALE study area was located in an area where major cyclones develop during winter (Colucci, 1976~. The occurrence of these extratropical cyclones is manifested by wind forcing over the continental shelf in the 2-10 day synoptic period. Cold air outbreaks that follow the passing of cyclones advect cold, dry continental air across the relatively warm shelf and Gulf Stream waters. Cold air outbreaks produce offshore winds that can last several days and strongly influence the observed mean winter wind stress, directed toward the southeast (Weber and Blanton, 1980~. Synoptic wind events have spatial scales similar to the along-shelf scale between Cape Canaveral and Cape Hatteras. This results in coherent wind forcing over the total shelf domain. Wind speeds are typically more than two times greater over the shelf than over the adjacent coast (Lee and Atkinson, 1983; Blanton et al., 1985~. (Copyright 1987 by the American Geophysical Union.) Organized Motions Muller and Garwood (1988) wrote: The development of organized cellular motion in the mixed layer can be seen by surface scattering Doppler sonars (J. Smith, Scripps Institution of Oceanography, La Jolla, Calif.~. The Doppler shift of the sonar return signal provides a measurement of the velocity field. These organized motions or secondary flows have a vertical scale comparable to the depth of the mixed layer and are frequently identified as Langmuir cells. Langmuir cells are a classical phenomenon (Langmuir, 1938), yet there is still dispute about how they are generated. One widely accepted cause is related to an interaction between surface gravity waves and Reynolds stresses. Other possible causes or contributing factors include the surface buoyancy flux, planetary rotation and rotation stress, and dynamic instabilities that are not directly caused but are modulated by the surface wave field. There may be more than one mechanism leading to phenomena subjectively identified as "Langmuir cells." The quantification of the energetics of these Langmuir cells is of particular importance for understanding mixed layer dynamics. Is their total kinetic energy content to be considered a part of the turbulent kinetic energy budget? Although these circulations are apparently not ubiquitous, are they an organization of the "normal" integral scale motions of the turbulence generated by shear production? Is their energy available for mixing in the thermocline? Are these motions dissipative, that is, quickly dissipated/altered when the source of energy is removed? Are energy and momentum from these cells transferred to internal waves in the entrainment zone, and do these waves contribute to mixing well down into the pycnocline? Does the present-day parameterization of the turbulent kinetic energy budget adequately include the effects of these organized motions? (Copyright 1988 by the American Geophysical Union.) Diurnal Cycle and Shallowing Mixed Layers Muller and Garwood (1988) summarized: Recent field experiments and theoretical investigations have concentrated on entrainment or deepening aspects of the ocean surface mixed layer. Now there is a growing concern with the shallowing of the mixed layer, in particularly with the diurnal shallowing. Data taken during the Tropic Heat study show that there is a significant diurnal cycle of mixing on the equator that had not been observed oreviouclv (Peters) Dic.cination chnnae.c he two ~ ~ ~ rig -~~~J I ~~~~~~~ ~ rat ~~~ ~ ~ orders of m~l~nihlA.e The high v~lil~.c Of Aiccin~tir~n n`~nPtrat`~ ~11 into the th~rm~f~l;~^ a - A ma., an. ~_. ~ ~4 48~. . . . ~ &~ - ,&~1t it "~ w1 All ~ll~L1 "Lid Wow HILT Lily Lll~1 lilU~llilO Ally 111~y be associated with the breaking of downward propagating internal waves that are generated by nighttime convective motions in the mixed layer. Convective cloud lines are a possible cause for the diurnal cycling of mixing at the equator (C. Gautier, Scripps Institution of Oceanography, La Jolla, Calif.~. This mechanism has a strong diurnal variability with Pronounced nighttime coaling hur.ctc concurrent with wind stress bursts. lye equatorial mixed layer is worth special attention because it may prove to be a "laboratory" for certain mixed layer processes. Since the vertical component of planetary rotation vanishes, rotational aspects might be less complex at the equator than at mid-latitudes. The large vertical shear of the equatorial undercurrent is a unique source of turbulent kinetic energy that can be 35

OCR for page 25
42 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF volumes of shelf water which can be traced by satellite images (Evans et al., 1985), moored instruments (Churchill et al., 1986), drifters (Bisagni, 1983), or through radioactive tracers (Orr et al., 1985~. Despite the availability of these exchange mechanisms, very little offshore water is found to penetrate onto the shelf (Chapman et al., 1986~. The dynamics of the shelf break front have been somewhat puzzling because it is often not a density front, hence it may not play an active role in the momentum balances. Further, its tendency to "anchor" at the shelf break suggests that somehow this is a special location. A simple theory by Chapman (1987) appears to explain the front's existence . . . in a way that addresses the above constraints. Simply stated, shelf water is treated as a passive tracer adverted by the barotropic flow field. At the shelf break, a front is maintained by the balance of geometrically induced vertical spreading with offshore advection and lateral mixing. Deflections of the front, once it is formed appear to be due to wind effects (Ou, 1984a,b), the passage of eddies, and of hydrodynamic instabilities (Ramp et al., 1983~. (Copyright 1987 by the American Geophysical Union.) A series of fronts is maintained on the Bering Shelf by the varying balance between buoyancy input and turbulent mixing caused by the tidal currents as the depth of the water changes (Coachman, 1986) (see Chapter 3~. Surface-floating material, such as oil, will collect in fronts and follow their movements. Cross-Shelf Transport Allen et al. (1987) stated: Although the large-scale along-shelf flow in many regions has been described to lowest order, the structure and strength of the cross-shelf flow is poorly known. Cross-shelf flows are difficult to measure because they are weak and have short spatial scales. At the outer edge of the shelf, episodes of very strong offshore flow occur, but they are hard to measure because of their short along-shore scales and their episodic nature. Nonetheless, the cross-shelf flow is critical to exchange of water, heat, salt and nutrients land oil]. In addition, the cross-shelf flow transport of particles and various dissolved chemicals is of direct practical importance. The cross-shelf component is also important dynamically, often providing a clearer diagnostic of the flow [dynamics] than does the along-shelf component. (Copyright 1987 by the American Geophysical Union.) Buoyancy-Driven Flows According to Allen et al. (1987~: The continental margin represents the region where saline oceanic waters contact and mix with the fresher waters associated with runoff from land. Since there is usually a density contrast between the two types of water, associated structures in the currents are expected. These effects are most dramatic in high-latitude regions with large runoff, such as Norway or southern Alaska, where salinity contrasts cause large density contrasts and large currents. Despite their importance, buoyancy driven currents are not well understood. (Copyright 1987 by the American Geophysical Union.) Studying buoyancy-driven flows and their variability involves several aspects of oceanography including hydrology, meteorology, glaciology (in some regions), and forcing. Tides Tides are ubiquitous features of the marine environment caused by the gravitational attractions of the sun-moon-earth system. They are generated primarily in the deep ocean basins and then propagate over the continental shelves and into coastal waters as long gravitational waves damped by bottom friction. Tidal propagation may range between being almost normal to the shelf, as for the semidiurnal tide on the U.S. Atiantic coast, to being primarily an alongshore Kelvin wave, as for the semidiurnal tide on the Pacific coast. For the diurnal tide, a continental-shelf wave is often present as well. Daifuku (1981) shows that in the Mid-Atlantic

OCR for page 25
STATE-OF-THE-ART OVERVIEW 43 Bight, the Kelvin wave accounts for most of the diurnal surface tide, whereas roughly 80% of the diurnal current variance is due to a continental-shelf wave. Tidal currents typically run from 0.01 to 0.1 m/s, with values reaching up as high as 1 m/s in the vicinity of certain banks, shoals, and passes. While the alongshore tidal variance is an important signal in many, but by no means all, conditions, its cross-shore variance usually dominates the variance due to other processes. Tidal currents are sufficiently energetic to vertically mix the water column inshore of the 50-m isobath in the Bering Sea shelf (Schumacher et al., 1979) and on Georges Bank, in the Great South Channel, and on Nantucket Shoals (Garrett et al., 1978~. In other cases, the effective mixing is restricted to a well-mixed bottom layer. With rather simple wave models, it is possible to match the observed sea-surface elevation and bottom-pressure records on the shelf and slope. The model and observed currents, however, can be drastically different, with the observations varying significantly over a shorter distance scale than would be expected from the modeled wave lengths. It is believed that small-scale bathymetry and an irregular coastal boundary may be largely responsible for this effect (see, e.g., Rosenfeld and Bearcisley, 1987~. Furthermore, when continental-shelf waves are present in the diurnal signal, their currents are effectively independent of the tidal sea-level changes. The conclusion to be reached is that verification of a model against sea-level and bottom-pressure records does not verify the model for currents. Furthermore, in order to account for their shorter scale of horizontal variation, currents must be verified on a denser network than is required for sea level. Due to nonlinearities in the governing equations, tidal motions can generate mean flow. This flow may be an important component of the overall surface drift, as it apparently is on Georges Bank (Loafer, 1980; Hopkins and Garfield, 1981; Butman et al., 1983; Greenberg, 1983, for example). Internal Waves The generation of internal waves by the interaction of surface tides with topography has been well documented observationally and theoretically. These internal waves take one of two forms depending on the linearity of the generation process. In the linear regime, the internal waves have tidal periods (usually semidiurnal), are generated at the continental shelf, and propagate shoreward, starting as a tidal beam but changing to a lower-mode wave as the higher modes lose their energy through dissipation. Simple models (Rattray, 1960; Baines, 1973; Prinsenberg and Rattray, 1975) illustrate the basic physics involved, while observational data presented by Reid (1956), Lee (1961), Torgrimson and Hickey (1979), DeWitt et al. (1986), and numerous others demonstrate that the waves can be significantly modified by the natural background variability. The high shears associated with these internal tides near their generation region can potentially increase the rate of mixing occurring at the shelf break. The nonlinear regime is typified by trains of internal waves occurring at regular intervals of tidal period. They are essentially generated by the interaction of a tidally varying flow with topography to generate transient internal waves, modified by the tidal current advection, at particular phases of the tidal current (Hiblya, 1986~. As the tidal current changes, internal waves propagate shoreward and evolve into a train of solitary waves as shown by the model of Lee and Beardsley (1974~. There are numerous observations of these waves propagating shoreward over continental shelves, earlier by Ewing (1950), through the observation of surface slicks, and then later by Halpern (1971), by thermistor measurements. More recently, satellite observations have demonstrated the presence of similar internal wave trains propagating shoreward over many shelves, as summarized by Apel et al. (1975) and Sawyer (1983~. The surface convergences occur at intervals of a wavelength and are associated with the onshore propagating wave packets. They can collect and transport shoreward floating material such as oil, as reported by Shanks (1987~. Shanks also suggested that oil caught in these convergence zones could kill or injure larvae that are often concentrated there. Furthermore, he stated the possibility that the downwelling currents at the convergence zones could pull less buoyant fractions of an oil spill underwater, making them less accessible for cleanup.

OCR for page 25
44 PHYSIC NOW OF THE U.S. OUTER CONTINENTAL SHELF Lagrangian Motions Lagrangian motions, as determined by the use of drifters, have been characterized in terms of their diffusive properties (see, e.g., Davis, l985b) and their means, which may not always correspond to the Eulerian mean (see, e.g., Chelton et al., 1987~. Davis found that drifter displacement statistics in CODE indicated that the probability density of particle displacements was reasonably well modeled by eddy diffusion with an anisotropic and inhomogeneous eddy diffusivity. At an offshore2di~tance of the order of 10 km, he found the cross-sh~lf component of the diffusivity KXX~ 10 m /s, and the alongshelf component K ~ 3 x 103 m /s, K increased offshore, while K decreased. Although he found that eddy diffusion may adequately characterize the mean scala~ransport, there seemed to be no simple relation between lateral eddy fluxes of momentum and mean shear. Use of the above Lagrangian determinations of KXX and K as estimates of the Eulerian horizontal eddy viscosity leads to variable errors of at least an order of magnitude. Particle-pair statistics describe stirring processes such as the dispersal of a scalar contaminant cloud. Davis (l9SSb) found that these processes cannot be modeled as diffusion in CODE, even if appeal is made to a scale-dependent diffusivity. Examination of particle-separation probability densities suggests that the relative velocity between widely separated particles is approximately normally distributed. The relative velocity between closely spaced particles, however, is intermittent, perhaps because closely spaced particles can be trapped within the same small-scale convergence. Chelton et al. (1987) found that off central California the drifters gave results consistent with current-meter measurements and surface dynamic topography in July 1984. In contrast, the drifter trajectories for the two winter surveys were difficult to rationalize in terms of the flow patterns inferred from other data. For example, most of the February 1984 drifters moved in a generally southward or southwesterly direction. Yet the geostrophic flow was consistently poleward in the drifter survey region. Similarly, the January 1985 geostrophic flow was quite strongly poleward in the drifter region, but the drifter trajectories are more indicative of variable flow. Windage of the drifters was not believed to be a problem. Chelton et al. (1987) considered a more likely explanation for the discrepancies between drifter and hydrographic data to be poor representation by the hydrographic data of near-surface currents. However, the difference between Lagrangian and Eulerian mean flows could be real. NUMERICAL MODELS Circulation Modeling Circulation modeling was one of the topics considered at a recent workshop on U.S. plans for research on the physical oceanography of the continental margins, held in Boulder, Colorado, from March 30-April 1, 1987. The workshop report (Allen et al., 1987) summarized the topic as follows: Oceanographic models range from qualitative conceptual models through analytical and laboratory models to extremely complex numerical models. Each type [has] a role to play. Although emphasis should be placed on numerical models, analytical models will always be important because they provide quantitative expression of individual physical processes. They are thus useful for the interpretation of both field observations and results from more complex numerical models. The various types of numerical models each have important applications. A simple idealized model is sometimes the best way to study a single process. The comprehensive model, which in principle could contain all macroscopic ocean processes, provides interpretations of observations and extensions of experimental results. Extension is especially important because no field program can hope to study all parts of the coastal ocean. [Numerical models also allow extension in time as well as in space; this is essential to get sufficient statistics.] A well-verified and well-understood numerical model could, with proper inputs, be used for quantitative prediction in regions where observations are limited. Further, the results of numerical models are needed so that observational programs cart be planned to distinguish clearly between competing hypotheses and also to provide a context for other models, for example, for biological oceanography or sediment transport.

OCR for page 25
STATE-OF-THE-ART OVERVIEW As an element of a . . . program [i.e., ESP research], a modeling effort should take as its objectives both the improvement of our modeling capability and the use of that capability for the study of specific processes or regions. Some issues concerning the improvement of modeling capability can be clearly identified: (1) Most Misting models need improvement in their parameterization of processes that are smaller than the grid scales, both vertical and horizontal. Mixing and energy dissipation are especially significant on continental margins because of the shallowness of the coastal ocean and [the large gradients in properties]. Refined understanding of dissipation in surface and bottom layers, as well as of interior mixing processes, should be reflected in improved formulations of these processes in numerical models. A related issue is understanding the extent to which processes with different time scales (e.g., surface waves, tides, and wind-driven motions) can be separated in a nonlinear ocean. (2) The construction of appropriate lateral open boundary conditions has proved troublesome in practice; improvement is needed. Correct representation of the offshore boundary conditions for coastal models is not well established and may be complicated by phenomena such as upwelling filaments and warm ring impingement. Further, the fact that the shelf has the characteristic of a waveguide complicates the imposition of [boundary conditions across the shelf at the upstream and downstream ends of the domain of interest]. (3) Driving forces at the surface and at the coast need to be better incorporated. Wind stress, freshwater runoff, stresses due to wave breaking, surface heat exchange, and surface evaporation and precipitation all need to be included. [Better understanding of the accuracy of fr~rc~n~ function description needed to obtain desired model outout accuracy is also needed.! 45 _= ~ r (4) Data-assimilative models need to be devised to serve as both diagnostic and predictive tools. The derivation of the full benefit from a set of observations depends on the use of such models. (Copyright 1987 by the American Geophysical Union.) Testing of Numerical Models There is a need for synthesizing the results of field programs and modeling efforts to achieve the maximum utility from both (a particularly important goal for MMS's OSRA modeling efforts). Validation is needed to elucidate how well models reproduce the necessary processes and phenomena. Allen et al. (1987) continued: Only by combining the results of individual field and modeling efforts can an increase in understanding (and thus utility) be achieved. As numerical models become more comprehensive, they must be subjected to continuous testing. Field results will be interpreted through the dynamical concepts embodied in the models. Because of the importance of models to the program, they must be carefully evaluated first by comparing them quantitatively and objectively with observations and second by interpreting their results in terms of simpler, process-oriented analytical or laboratory models. There is little use for models that are not well tested and well understood in terms of their dynamical behavior. Synthesis of field experiments also needs to take place on two levels: first, a quantitative description of all of the interesting phenomena that can be resolved and second, an understanding in dynamical terms. The critical questions are: what processes dominate at what places and times, and for what reasons? Oceanographic models can, and should, be used to help achieve this synthesis. Models can be used to interpolate and fill in gaps in sparsely sampled data sets, and models can be used to further a dynamical understanding. (Copyright 1987 by the American Geophysical Union.) With regard to the objective evaluation of numerical model performance, Willmott et al. (1985) further commented: With the development and use of simulation models becoming a major focus in the geophysical community, the need to evaluate a model's performance comprehensively and objectively or to compare competing models has become an important but underinvestigated aspect of modeling research. Not only is the model evaluation literature sparse, but the discussion is often specific to a small class of problems (e.g., air pollution or solar radiation models) and frequently the recommendations are contradictory. (Copyright 1985 by the American Geophysical Union.) Willmott et al. (1985) presented several techniques for quantitatively comparing model predictions with the results of observations:

OCR for page 25
46 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF . . . [A] small set of complementary difference measures can represent an objective and meaningful description of a model's ability to reproduce reliable observations precisely or accurately, regardless of whether the events of interest are scalars, directions, or vectors. The core of this set of difference measures is made up of the root-mean-square error, the systematic root-mean-square error, the unsystematic root-mean-square error, and the index of agreement, although the mean absolute error and a modified index of agreement supply related but useful information. [Bootstrapping also] provides a general and reliable way to evaluate the difference indices or, for that matter, any statistic of interest. When these difference measures are used in conjunction with the appropriate univariate statistics and data-display graphics, the operational evaluation of the performance of one or more models can be comprehensively accomplished. (Copyright 1985 by the American Geophysical Union.) They concluded with the statement that their methods . . . may be extended to several other interesting problems, such as the comparison of model-predicted and observed flow fields. Model-predicted and observed wind velocity maps, for instance, could be quantitatively compared. If the model-predicted and observed variables are time series, on the other hand, time-dependent errors within the model could be detected by the calculation and interpretation of the difference measures at lags other than zero. To gain even further insight into the nature and sources of the error variable or field, it may also be useful to partition the difference variable into its spectral (cf. Weisberg and Pietrafesa, 1983) or eigenvector (cf. Preisendorfer and Barnett, 1983) components. Several other extensions also could be conceived, but even when the [suggested] evaluation is conducted in its most basic form, the ability of one or more models to reproduce nature accurately can be dependably assessed. (Copynght 1985 by the Amencan Geophysical Union.) Modeling the Spreading and Dispersion of Oil A comprehensive review of the state of the art in oil-spill-fate modeling was recently completed by Spaulding (1988~. Earlier model reviews included Huang (1983), Huang and Monastero (1982), Davidson and Lawrence (1982), and Stolzenbach et al. (1977~. General reviews of the fate of hydrocarbons in the marine environment have been presented by Jordan and Payne (1980), Mackay (1985), NRC (1985), Payne and McNabb (1985), and Payne and Phillips (1985~. The purpose of this section is to highlight the current state of the practice in the modeling of spreading and dispersion of oil. These two processes have been selected for review because they are closely tied to near-surface physical oceanographic processes (see Fig. 6~. An NRC report on oil-spill dispersants also reviews oil-spill-fate modeling and the chemistry and physics of dispersed of! as well as the use of dispersants (NRC, l989b). Spreading Spreading is one of the most important processes in oil-spil1 dynamics, because it determines the areal extent of spilled oil and affects the various weathering processes influenced by surface area, including evaporation, dissolution, dispersion, and photo-oxidation. Spreading has historically been considered to be controlled by the driving forces of gravity and surfacetension and the retarding forces of inertia and viscosity. Various researchers have investigated this process based on this conceptual model, and several methods are available for use in its modeling. Hayes (1971) three-regime spreading theory is the most widely used approach (Huang, 1983~. Most other methods are variations of Fay's spreading theory, incorporating diffusion and dispersion, random Fickian diffusion, and the thick-thin slick approach (Mackay et al., 1980a). These modifications are an attempt to account for observations that show that 80-90% of the total area of a slick consists of thin sheen and about 10% thick slick. Most of the oil, however, is observed to be in the thick slick (Huang, 1983~. They also attempt to address the fact that turbulence at the sea surface can dominate spill spreading in the final spreading regime rather than the surface tension-viscous force balance employed by Fay (1971~. None of these

OCR for page 25
47 So o _ ~ in a: ., o UJ - _ . C, ~ Z _ o _ oC _ LU ~ a: O in ~ . g o :' 4 - ~ 1_ 3 .e i: o i: o . - .m Go o' UJ - o o a: Z Z Z Z Z o o o o o F ~ F F a: , UJ C' O ~ IL Z in ~_ in C] , , ~ o m o - - x o to o . o U' 0 .- o 'n 5

OCR for page 25
48 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF techniques, however, addresses the inherent"patchiness" of actual spills or the thickening of spills near the leading edge. One of the important recent developments in spill spreading is the work by Johansen (1982, 1983, 1985, 19873, Johansen and Audunson (19823, Elliott (1986), and Elliott et al. (1986~. In their approach, oil is modeled as a distribution of droplets that are driven into the sea by breaking-wave events. Once in the water column, the droplets are advected and dispersed by the near-surface currents, where vertical shears are important. Most of the droplets, each with its own buoyancy, eventually resurface. Oil spreading is hence controlled by the droplet-size distribution and the shear-diffusion process. This mode} correctly predicts the occurrence of thicker oil toward the leading edge of a slick and the alignment and elongation of slicks in the Erection of the wind. This technique will undoubtedly replace procedures based on Fay's theory. ~ ne percent or spate pa~cn~ness, however, remains a pro Stem that wall require Improved insight into near-surface transport processes before substantial progress can be made. TO ~ ^ ~ _~:1] If__ :_ A__ t~ ~ ~ _~ ___ .~ -ad ~ ~ Dispersion Dispersion is generally assumed to result from wind-generated breaking waves dispersing oil in the water column (Raj, 1977; Lin et al., 1978; Milgram et al., 1978~. The simplest approach uses tabulations of dispersion as a function of sea state and time after the spill. Audunson (1979) suggested an empirical formulation based on the square of the wind speed, reflecting the amount of energy available for driving oil droplets into the water column. Reed (19803 and Spaulding et al. (19823 used a variation of Audunson's approach, including an exponential decay function, to account for weathering and mousse formation. According to this formulation, 99% of dissolution and dispersion is complete within the first few days after release of oil onto the sea surface. Mackay and Leinonen (19773 and Mackay et al. (1980b) formulated a two-stage dispersion process. The equations describing this process treat dispersion from thin and thick slicks separately, and agree qualitatively with observed behavior, but they have yet to be verified. Spaulding et al. (19823 proposed an approach that calculates the mass flux rate of oil into the water column by breaking-wave-induced turbulence, but the technique is not sufficiently developed for use in spill models. Aravamudan et al. (19823 developed a simplified but highly theoretical model of dispersion based on turbulence generated at the sea surface due to breaking waves. The approach has not been widely adopted because of its complexity and lack of validation. Recently, Delvigne (1983; 1984a,b3 completed a series of measurements of the dispersion of oil in the water column (below the breaking-wave zoned and developed a theoretical model (Delvigne et al., 1987) that was verified by laboratory data of the vertical dispersion coefficients for oil and oily, suspended particulate matter. However, the study did not address the dispersion caused by breaking waves. Chemical dispersants applied on the surface of oil slicks can decrease the oil-water interracial tension (NRC, l989b). This results in an increase in the oil surface area and the breakup of the slick into tiny droplets. These droplets may then disperse in the upper water column under the influence of natural turbulence and wave action. In some cases, depending on sea state and the type of dispersant used, additional mixing energy must be applied from a boat, as with a pressurized water spray. The concentration of oil droplets in the water column is highest near the surface and declines with depth. The depth and degree of dispersion will vary from case to case, depending primarily on sea state. Field tests conducted in moderate seas have detected dispersed oil at low concentrations (1-20 ppm) down to depths of 6-9 m shortly after dispersion. Oil on the sea surface will drift in response to wind and currents. Dispersion of oil into the water column isolates it from the effects of wind, and the dispersed oil plume will drift with the near-surface currents. Depending on specific conditions and the slick-drift forecasts, it may be tactically advantageous to disperse the oil to reduce wind effects.

OCR for page 25
STATE-OF-THE-ART OVERVIEW 49 SEA ICE Ice Modeling Several well-developed ice models are available for application to the Alaskan OCS waters (Rothrock, 1970; Kowalik and Untersteiner, 1978; Hibler, 1979; Pritchard, 1980; Kowalik, 1981 1984; Thorndike and Colony, 1982; and Overland et al., 1984). Many are even integrated with hydrodynamic models for Alaskan waters. However a sense of how well these models actually perform in representing the range of ice conditions for the OCS areas is missing. Hence, a fully coupled ice/hydrodynamics model should be applied to selected areas and times for which data are available, and a detailed model-data comparison should be done. Chapter 4 includes more detailed suggestions for improving the modeling of ice movements. Ice-Oil Interaction Summaries of the fate and behavior of oil in the arctic environment are included in Walker (1975), Mackay (1984), Payne and McNabb (1985), Bobra and Fingas (1986), and Reed et al. (1986a,b). Ice conditions are highly influential in determining the movement and final disposition of spilled oil. In open water (10% ice cover or less), the primary processes affecting the oil are spreading, advection, and evaporation under calm conditions. In rough water, dispersion and emulsification also become important (Buist et al., 1983~. Even at temperatures of 0C, evaporation occurs, with estimates of up to 30,to evaporation in 48 hours (Buist et al., 1983) and up to 40% evaporation over a 2-week period (Logan et al., 1975~. Such weathering results in greater density of the remaining oil, which in turn causes increased thickness of the oil by increasing the viscosity. The low temperatures encountered in the Arctic further increase the viscosity and also increase the oil's equilibrium film thickness. Rosenegger (1975) reported a minimum equilibrium thickness of 0.0025 m for the spread of crude oil on the water surface. When oil temperatures drop below -9.5C, oil gels and spreading due to surface forces cease to occur (Rosenegger, 1975~. As oil spreads on the water surface, some of it is deposited on, in, and under ice with which it comes in contact. This oiled ice can be transported for a considerable distance before melting occurs and the of! is released. Under freezing conditions, ice forms beneath the oil slick. The oil remains as a film on the ice if conditions are calm, but surface-oiled pancake ice develops under rough conditions. The oil will most probably be covered by snow and will stabilize until the spring melt. Deterioration of the ice sheet in the spring releases most of the oil in first-year ice back into the water; the rest travels with the broken ice sheet (Logan et al., 1975~. In ice-infested waters (10-80% ice cover), the spreading and movement of oil are highly dependent on ice dynamics. The same processes described for open water are in effect. However, ice dominates advection and spreading by increasing surface drag and thereby reducing the velocity of wind-blown slicks. Furthermore, a medium-to-heavy concentration of ice tends to herd the oil, and the the oil's motion is constrained to follow the motion of ice, which moves more in response to surface currents than does the oil. The oil remaining on the water surface is restricted to leads and to areas between floes, where damming increases the oil thickness beyond that experienced in open water. Estimates of probable oil thickness range from 0.01 m (Walker. f l9-/~) to several nunoreutus ot a meter (Logan et al., 1975~. The progressive opening and closing of leads and the continuous motion of the pack ice could transport the oil for long distances, leaving a path of oiled ice. Freezing sea water would incorporate the surface oil, and the entombed oil could then be transported up to hundreds of kilometers from the site of its release in an essentially unweathered state (Logan et al., 1975~. These factors make it difficult to predict the extent of areal contamination. Oil incorporated into ice as the result of leads having frozen may eventually be deposited in newly formed ridges, because leads are areas of structural weakness and are, therefore, particularly susceptible to pressure ridging. Oil spilled on the surface of solid ice (e.g., oil from a tanker spill) remains on the surface with spreading due only to gravity. The surface roughness of the ice determines the extent of areal coverage, and an average oil thickness of 0.03 m is expected (Buist et al., 1983~. The

OCR for page 25
so PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF thickness of the slick and the low temperatures cause weathering to proceed very slowly. Eventual snow cover of the oil essentially halts weathering until spring and summer. Oil spilled beneath solid ice (e.g., a well blowout) coalesces to form a slick between the ice and water. The underside contours of the ice determine the amount of oil that can be stored in the absence of currents. Irregularities on the bottom of sea ice occur on several characteristic scales. Under smooth ice, crude oil forms films of 0.005 to 0.01 m (Walker, 1975), and differential currents of 0.03 to 0.07 m/s are required to move the slick (Stringer and Weller, 1980~. Bottom irregularities are important in containing oil under ice and limit the extent of its spread. Oil will collect in lenses a few meters across and up to a few centimeters thick (Walker, 1975) rather than in a continuous layer. Oil will remain in these undulations unless disturbed by an appreciable current along the ice bottom. Oil deposited under ice with a 0.01-m amplitude roughness requires a current of 0.25 m/s to initiate movement (Stringer and Weller, 1980~. The presence of gas escaping with oil from an underwater release may increase the spread of oil under solid sheet ice, since buoyant gas bubbles will displace the oil from concavities (Logan et al., 1 975~. From autumn through early spring, oil beneath the ice will become entombed in the growing sea ice within approximately a week, and thereafter will not weather appreciably. During this period, the buoyancy force of oil causes no significant penetration of oil into the ice from below. Large brine drainage channels in first-year ice offer the most likely means for the limited penetration that may occur. In the spring, as ice temperatures approach freezing and melting proceeds, the drainage channels open in first-year ice. The brine drainage structure of first-year ice probably provides the major pathway for the rapid upward migration of oil in the spring. From 70-80% of the entombed oil rises to the surface of the ice over a period of approximately 1 week (Walker, 1975~. As the surface snow and ice melts, the oil covers the surface of melt pools, decreasing albedo and speeding melting. The oil may reach temperatures of 5 to 10C (Walker, 1975), and rapid evaporation of the lighter fractions will occur. Once the ice has melted completely, the oil will reside on the sea surface, possibly being incorporated in a weathered state into the first-year ice of the next winter. Alternatively, such weathered residual oil may sink to the seafloor and become incorporated into the sediments. In multiyear ice, similar processes are active, but the time scale is much longer. Oil most probably becomes entombed at greater depths, where the temperature increase in the spring is slower than in thinner ice. The surface layer of multlyear ice is essentially fresh and does not contain brine channels to facilitate oil movement. The upward migration of of! is therefore much slower, but oil should reach the surface within a maximum of 4 years (Walker, 1975~. Once on the surface, its behavior is much like that described for oil on first-year ice, except that the ice will not necessarily melt sufficiently for the oil to return to the water surface. Several researchers have attempted to quantify and predict some aspect of oil-ice interactions. Free et al. (1981) developed a set of empirical equations to describe the spreading of oil in a broken ice field. The properties of the spilled oil, the size and concentration of the ice field, and the velocity of currents and wind are required as input. The equations match data obtained from ice flume tests, but certain limitations were noted. Due to the method of data acquisition, the results are valid only for one-dimensional situations, and the empirical constants are possibly biased by scale and one-dimensionality effects. Furthermore, the equations are not general enough for application to all situations. Rosenegger (1975) presented results of laboratory investigations of the flow of crude oil beneath sea ice. Functional expressions were developed that describe interracial tensions between oil and brine at the brine and ice interface, and the force required to initiate motion of an oil bubble below an ice sheet. Equilibrium thicknesses of the two types of crude oil were determined. A separate study by Nelson and Allen (1981) examined the migration of oil through first-year sea ice and the effect of entrained oil on ice growth rates. Surface insulation of the ice sheet was found to induce upward oil migration by raising the ice sheet temperature and enlarging brine channels. Equations are presented that predict brine volume changes with temperature. The thermal conductivity of sea ice can be as much as 20 times greater than that of oil. Entrained oil can therefore be expected to greatly alter ice growth rates under the oil layer. Evaporation of the lighter fractions of oil under arctic conditions has been studied by several researchers. Laboratory studies reported by Tebeau et al. (1982) showed a well-defined

OCR for page 25
STATE-OF-THE-ART OVERVIEW quantitative relationship between the physical properties of oil and evaporative exposure levels particularly when emulsification had not occurred. This study also noted a functional relationship between the rapid decrease in aqueous solubility of oil and increasing evaporative exposure. Weathering was also seen to decrease the oil-water interracial tension, but no well- defined relationship could be found. Stiver et al. (1983) presented an analytical expression to describe the extent of evaporation as a function of evaporative exposure and a dimensionless air- oil partition coefficient. The spreading and the evaporative and dispersive losses of oil in broken ice fields are poorly known at best. The consensus is that medium to heavy oils exhibit a herding effect and that the evaporative and dispersive losses decrease due to sheltering from the wind by the ice field (Cox and Schultz, l981a,b; Free et al., 1981; Reimer, 1981; Ross and Dickens, 1987a). The behavior of oil under freezing and thawing conditions is just beginning to be studied with a view toward modeling (Wilson and Mackay, 1986~. Recent work by Ross and Dickens (1987b) should also markedly improve the ability to model oil in leads. 51 Numerical Models Including Oil-Ice Interactions Extension of spill models to handle oil-ice interactions has been extremely limited. Applied Science Associates, Inc. (1984) formulated a model based on the existing state of knowledge for arctic waters. The model addresses oil-ice interactions, including drifting, spreading, evaporation, emulsification, and dispersion for spills under ice and in ice-infested waters, but excludes freezing and thawing situations. Data on ice, temperature, wind, and current conditions are derived from available atlases (Brower et al., 1977; LaBelle et al., 1983) or model predictions. Wotherspoon and Swiss (1985) present an oil-ice interaction model but do not describe the theoretical or empirical formulations employed or the results from any simulations. When oil is located under ice, researchers have a reasonable set of algorithms to describe its motion (advection), based on the work of Sayed and Abdelmour (1982), Uzuner et al. (1979), and Cox and Schultz (l9Sla,b), and to describe trapping in under-ice roughness elements (Kovacs, 1977, 1979~. If oil is incorporated in a broken ice field, we know from Coon and Pritchard (1979), Reimer (1981), Allen (19R3), and Thomas (1983) that the oil drifts with the ice. Belore and Buist (1988) have recently developed a detailed model for oil spills in snow that fills an important gap in our knowledge. SEDIMENT TRANSPORT Bottom Boundary Layer and Transport of Suspended Materials Brink (1987) wrote: The surface and bottom boundary layers represent the areas where shelf waters absorb the wind and bottom stresses, respectively. These stresses are particularly important over the shelf, since they influence a relatively shallow ([about] 100 m) depth, water column. Further, since the turbulent layers themselves are typically 5-20 m thick, they often represent a substantial part of the shells volume. Thus, understanding the shelf in general requires a knowledge of boundary-layer processes within this environment. The study of the bottom boundary layer over the shelf has recently been strongly influenced by nonlinear models coupling [long] surface gravity waves (e.g., swell) with lower-frequency motions within the bottom boundary layer. Physically, a thin ([about] 0.05 m) near-bottom sublayer associated with both the waves and currents is generated, and it is an area of extremely active turbulence. The effect of this nonlinear coupling can ultimately be parameterized as an enhanced bottom roughness. Thus, the existence of [long] surface gravity waves will act to increase the bottom stress on lower-frequency flow patterns. Farther from the bottom, rotational effects become important and "Ekman spirals" such as those observed by Dickey and Van Leer (1984) are to be expected. The original simple models of wave-current coupling (Smith, 1977; Grant and Madsen, 1979) have since been extended to include the effects of stable ambient stratification, the earth's rotation, and self-stratification due to sediment suspension (Glenn and Grant, 1987~. Accompanying these theoretical advances was the demonstration that the wave-current theories actually compare quite favorably with field observations (Grant et al., 1984). Further, accounting for wave-induced enhancement of bottom stress

OCR for page 25
52 PHYSICAL OCEANOGRAPHY OF THE U.S. OUTER CONTINENTAL SHELF considerably improves the comparison between shelf-w~de model results and observations (Brink et al., 1987a). A comprehensive review of bottom boundaty-layer processes over the shelf can be found In Grant and Madsen (1986~. While recent advances in bounda~y-layer modeling have been substantial, a good deal remains to be done In terms of field verification, subsequent refinement, and incorporation into larger scale circulation models. Bottom boundaty-layer models need more testing in the field, because truly realistic environments simply cannot be created in the laboratory. Present boundary-layer models are [only one dimensional, while circulation models] need to average bottom stresses over larger areas (several km) which may have very inhomogeneous m~crotopographies. Finally, more work needs to be done on the nearshore area where the surface and bottom boundary layers overlap. To date, only simple models (e.g., Mitchum and Clarke, 1986) have been advanced. (Copyright 1987 by the American Geophysical Union.) Knowledge of the general circulation and bottom-boundary flow conditions provides a basis for estimating the transport of suspended materials. However, with few exceptions, there appears to be a dearth of specific data on suspended matter concentrations in the water column collected in conjunction with ESP studies conducted in continental margin areas. Bottom-boundary layer and sediment transport studies carried out as part of the ESP have been performed mainly by USGS scientists via a memorandum of understanding with MMS. These studies have been carried out in Alaska (see, e.g., Cacchione et al., 1982), on the Pacific coast (see, e.g., Drake et al., 1985; Cacchione et al., 1984), and on the Atlantic Coast (see, e.g., Butman and Noble, 1978; Butman et al., 1980~. Similar work in the economically active and potentially active areas of the Gulf Coast has not been carried out. Sediment Transport and the Effect of Oil One significant fate of inputs of of! drilling fluids, or tailings, is deposition to sediments or incorporation into surface sediments (NRC, 1983, 1985~. Incorporation of compounds from spilled oil or from oil compounds chronically discharged from coproduced waters into surface sediments may have effects on benthic organisms for months to years depending on the amount and type of oil and the type of benth~c ecosystem (NRC, 1985: Boesch and Rabalais' 19871. Transport of the oil-contaminated sediment or the ding mud and tailings is an important consideration in terms of the spreading of potentially harmful contaminants to a wider area or another area, and also in terms of spreading and dilution with cleaner sediment. The transport of drilling muds and cuttings is most likely limited to the immediate vicinity of drilling activity for the coarse fraction, which makes up about 90% of the effluent (see, e.g., NRC, 1983~. The fine, suspended fraction is carried further downstream by ambient currents and is rapidly dispersed; this dispersion is in agreement with both theoretical predictions and direct observations (NRC, 1983~. In relatively quiescent environments, such as much of the deeper continental shelf and upper continental slope, the fine suspensate may settle out in detectable concentrations within several kilometers of transport during storms (EG&G, 1982; Neff, 1987~. In more active environments, the fine fraction is often dispersed to less than detectable concentrations within a very short distance (NRC, 1983~. Uncertainties still exist for low-energy, depositional environments that are exposed to repeated discharges over long periods of time, and for extremely sensitive environments (NRC, 1983~. Boundary-layer processes have significant influence on circulation in the continental margin areas and cannot be neglected in studies and models of circulation important to oil-spill fate considerations.