Oilspill Hindcast Simulation of the IXTOC 1 Gulf of Mexico Spill

An oil spill model developed at the University of Rhode Island was used to hindcast the Ixtoc 1 oil well blowout using three pairs of wind and current field inputs. The sensitivity of the model trajectory predictions to these environmental inputs is discussed and comparisons made to overflight field data collected by the United States Coast Guard and National Oceanic and Atmospheric Administration. Oil mass balance model predictions and field data derived mass balance estimates are compared for sea surface, water column, and atmospheric partitions. Surface oil trajectory and subsurface elevated hydrocarbon water masses are mapped using the best of the three trajectory simulations: a geostrophic current field derived from seasonally averaged hydrographic data and a wind record recorded at Brownsville, Texas. With oil input parameters and the URI Oilspill Model routines fixed for· the simulations presented, it is shown that none of the sets of environmental data used have adequate scales of resolution to drive the model in a ballistic simulation and achieve trajectory estimates which match the observed trajectories of Ixtoc oil. Model mass balance estimates for oil in several environmental partitions fall reasonably within the bounds of field-data-derived mass balance estimates.

is a critical step in the efforts to improve our capability to predict the fate of spilled oil in the marine environment (Stolzenbach et al., 1977;Raytheon, 1982).
Few spills have been studied extensively enough to warrant their use in hindcast studies. The URI Oilspill Model (Cornillon and Spaulding, 1978a, b;Cornillon et al., 1979) has previously been used to hindcast the Argo Merchant spill using a three dimensional numerical hydrodynamic model to estimate the wind-driven flow field in the area of the spill (Spaulding et al., 1982a). The Ixtoc 1 spill was selected as the next intensive hindcast effort because it was a reasonably well studied spill; it would test the oilspill model's use in a new geographical area; and an extensive three dimensional numerical model for the Gulf of Mexico (Blumberg and Mellor, 1981) had recently been completed. It was hoped that the flow field generated by this hydrodynamic model would give the basis for an accurate prediction of the surface trajectory of oil from the Ixtoc 1 blowout. Application of the URI Oilspill Model to a completely new geographic area would give experience in the data collection effort needed to apply the model to new environments. The URI Oilspill Model has been used extensively in the Georges Bank -Gulf of Maine region (Cornillon and Spaulding, 1978a, b;Cornillon et al., 1979;Spaulding et al., 1982a, b) and it was felt important to extend the geographic extent of its application.
In addition, the URI Oilspill Model's capability to give ffiass balance estima~es for atmospheric, water surface, water column (subsurface), and beached oil; and to give a prediction of the temporal and spatial extent of elevated subsurface hydrocarbon levels would contribute to the understanding of these as yet unreported simulation estimates.

Review of Previous Work:
Two previous oilspill hindcasting efforts of the Ixtoc 1 blowout have been reported. Grose, et al (1982) (Grose et al., 1982). Grose has kindly supplied the current field used for the EDIS model for this study. Galt (1981) headed the most intensive of the modeling efforts concurrent with the spill. Sources of information on the current field in the western Gulf of Mexico cited by Galt include: Nowlin and Mclellan (1972), dynamic topography; Bureau of Land Management (BLM) studies of the offshore Texas region, including the results of a number of current studies that used drift cards and current meters as well as hydrographic data; current studies and numerical modeling work done at NOAA/Atlantic Oceanographic and Meteorological 3 Laboratory (AOML); a paper by w. Sturges and J. P. Blaha (1976), hypothesizing that the strength of the Mexican Coastal Current is related to the large-scale curl of the wind stress over the Gulf of Mexico (see Discussion section under Chapter 4, Currents); and a Master's thesis by A. M. v. de la Cerda (1975). De la Cerda derived surfaces of constant thermosteric anomaly (Montgomery, 1954;Montgomery and Wooster, 1954) using a number of hydrographic data sets in the southwestern area of the Gulf, giving crucial information on the Campeche gyre and other permanent and non-permanent cyclonic and anticyclonic features of the region.
Because of the massive amounts of oil released from the blowout and the possibility of impacts on United States waters, considerable field observational resources were made available to the modeling group. GOES, TIROS, and ERTS satellite imagery were compared with model predictions of surface extent of oil transport within the first few weeks of the spill. Lower altitude aircraft overflights began on 3 July (Galt, 1981), leading to further flights by the State of Texas, Department of Transportation, the U. s. Coast Guard, and the National Atmospheric and Space Administration (NASA).
Surface ship data collection to quantify the strength and extent of the Mexican Coastal Current began in mid-July (Galt, 1981). A number of cross-shelf expendable bathythermograph (XBT) transects were carried out which delineated the axis of the Mexican Coastal Current and a smaller cyclonic feature off Tampico, Mexico. To define local small scale current features, helicopter-deployed Richardson current probes and radio frequency drogues and satellite-tracked drogues were deployed (Galt, 1981).
All of these field data were used to calibrate a two-component current modeling system consisting ,of a regional geometry first order geostrophic plus Ekman dynamics solution superposed with a streamline analysis (Galt, 1980) generating a mass-conserving flow field (Galt, 1981). The combination of the intensive collection of field data and multiple model simulations represents the work of a large group of investigators and support personnel with a considerable operational budget.
It is not reasonable to expect, given the inputs to the Galt (1981) ffiodeling effort, that a better job of trajectory modeling on the Ixtoc 1 spill will be done in the foreseeable future. The discussion of the Grose et al (1982) and the Galt (1981)  and c) the source and methodology for specifying the environmental driving forces for the model simulation. The major question to be asked is: "Which among these three sets of eletuents is limiting in the prediction?" The work described here· focuses on the use of existing environmental data inputs for the Ixtoc hindcast simulation.
Two relatively simplistic environmental wind and current datasets are contrasted with a reasonably advanced "state-of-the-art"   (Shuhy, 1979) ' .
Only Light ~ N Sheen Observed  and two, to contain the pressure of hydrocarbons in the formation with the hydrostatic pressure of the column of fluid, (Garmon, 1980 With mud circulation lost, it was decided to pull the drill string and remove the drill bit. The string was pulled out of the hole without incident and apparently without any sign of flow from the well until the drill collars reached the sediment-water interface. As they removed drill pipe, the engineers pumped drilling mud into the well to fill the space formerly occupied by the pipe and took measurements of the pressure in the mud column every 300 meters. They also tested the blow-out preventers (BOPs) on the sea floor and the drill-pipe safety valve on the platform. 13 Al much of the escaping oil was not contained by the Sombrero. so, Gas conveyed to the surface was flared, and dispersants were injected into the Sombrero piping system before the oil-water mixture was returned to the Bahia de Campeche. A PEMEX attempt in November to establish a subsea pipeline between the Sombrero and an onshore refinery was unsuccessful (OSIR, 1980a).

Relief Wells
In mid-June PEMEX began drilling IXTOC lA and IXTOC lB, both directional relief wells. Fluid communication between the IXTOC lB and the IXTOC 1 wells was established in late November, but capping was not successful until 27 March 1980.

Sea Surface Oil Recovery
Equipment from several manufacturers was exercised at the spill area, and at different times throughout the clean up operations was hindered by: 1) breakdown or improper deployment, 2) shifting plume trajectory, 3) oil sucked into working vessels' cooling systems, 4) suspension of activities for the safety of diving operations, 5) high winds and seas, and 6) low oil/water recovery ratios (OSIR, 1980a The URI Oilspill Model Spaulding 1978a, b, cornillon et al., 1979) is one of the most advanced of the oilspill models now in use. It is notable for its modular construction and high level of within-code documentation. Particular attention has been paid to the bookkeeping tasks of dimensional units, variations in the grid spacing and angles of inclination, and usable printed, plotted, and machine-readable simulation output. Huang and Monastero (1982) in their Review Major Routine Descriptions: The URI/OSFM formulation and development have been presented and documented adequately elsewhere cornillon et al., 1979;Reed et al., 1979;Reed, 1980). A brief description of the major routines follows. The spreading mechanism is described by the three-regime spreading model proposed by Fay (1971) and is shown in the following where: a, q, and~ are empirically derived constants A is spillet area, and U is windspeed These fractional evaporative losses are then summed and subtracted from the spillet mass.
Surface Advection: The movement of the spillet is determined by the vector sum of the surface water movement and a percentage of the wind speed. The wind-driven response of surface oil has been a consistent feature of oilspill models to date (Huang and Monastero, 1982), and is one of the modeled properties open to interpretation. A drift factor of l.l% of the observed wind for the surface slick over the surface r was reported from observations during the Argo Merchant spill wate (Grose and Mattson, 1977). Because many of the oilspill models iJllplemented thus far have not included in their formulation a methodology for estimating wind-driven surface currents, there has evolved a so-called "3% rule" (Huang and Monastero, 1982). This empirically derived "rule" has the surface advection of the spill 'defined by around 3% of the over-water wind velocity, with a deflection of between 0 and fifteen degrees clockwise, in some has been modified in three ways. First, a exponential decay with a folding time of one day has been added to address the expected decrease in entrainmen~ of oil as it weathers, or changes in chemical and physical properties when exposed to the external environment (Huang and Monestero, 1982). Second, a maximum value for U of 12 m/sec has been set, beyond which no further entrainment is generated.
This value was estimated ' from standard deepwater-wave forecasting tables for unlimited fetch, fully developed sea (Ippen, 1966) and also from an estimate of the approaching maximum of the probability distribution o~ breaking wave heights (Nath and Ramsey, 1976).
Third, a minimum wind velocity for entrainment is set at 5 m/sec (Nath and Ramsey, 1976). Breaking waves are assumed to be the major mechanism responible for oil entrainment, and breaking waves do not occur below the 5 m/sec wind speed threshold. The modified formulation is defined as: The mass of oil Me ' is partitioned into n particles according to a user-set mass-per-particle parameter, and each of these particles is injected into the water column by a uniform spatial random scheme across the spillet-water interface.

SUBSURFACE PROCESSES
The subsurface subroutine calls the seven main subroutines which advect and diffuse the subsurface droplets in the water column. The method is based on the Water-Advective-Particle-In-Cell method developed by Pavish (1977). A brief overview of the numerical scheme, along with a discussion of those factors which have been modified specifically for this application, is presented below.
The three-dimensional mass transport · equation is solved using a particle-in-cell method. The volume which these "marker" particles occupy is then divided into a number of rectangular cells. The concentration distribution of the particles is determined by calculating the number of particles in each cell, yielding an effective concentration positioned at the center of that cell. The model then obtains the concentration gradient within this field and calculates the resulting diffusive velocity, which it adds to the advective velocity input to the program, to obtain the total particle 24 velocity. Finally, the particle velocity is used to move the d 1 P articles for the time step in execution. indivi ua The fundamental equation being solved is the transport diffusion equation: The quality of environmental data, especially the wind and current fields, is of critical importance to successful simulation.
Optimally, one would have error-free real time sampled wind and current data from the spill site and affected advection areas. Such data is not available, nor can we expect it to be in the future. The modeller is presented with several forms of inadequately described wind and current fields and must make some determination about which combination is the most reasonable to use.

Environmental Parameters:
Winds: A continuum of complexity of wind field data exists for use in oil spill simulation. Examples from this continuum include: single station land site weather station data (National Climatic Center); a few long-term fixed-position continental shelf buoys; pressure-inferred two-dimensional wind fields computed on coarse three degree latitude-longitude grids on a world-wide basis and much smaller grids for special areas by the Fleet Numerical Weather Center (FNWC) in Monterey, California; and full three-dimensional 26 wind-field models including topographic effects and surface boundary layer formulations. Sophisticated analyses at offshore wind fields Cu rrently under development for certain areas (Mooers, 1978;are Weisberg and Pietrafesa, 1983). At present, however, no estimates are generally available which are directed at obtaining fine-mesh wind fields for high-risk oilspill areas.

currents:
A similar continuum of complexity exists for current estimation.
Data collection efforts along the continental shelf regions of the continental United States which focus on high-risk oilspill areas are becoming commonplace. These efforts are funded by the U.S. Minerals Management Service in support of offshore oil lease sale environmental impact studies. These data collection efforts suffer relative to their atmospheric analogs from few data stations at irregular sampling intervals. On the U.S~ East Coast, historical drifter studies (Bumpus and Lauzier, 1965) have served to give overviews of the general patterns of continental shelf flow.
First-order geostrophic models with dynamic topographic surfaces inferred from many years' hydrographic data (e.g. the U3 dataset introduced below) give pictures of the seasonally stable flow fields Which exist over continental shelf areas. More ambitious three-dimensional time-dependent hydrodynamic models incorporating Wind, tidal, and density forcing, with various numerical solutions 27 b ndary condition specifications, are now in the development and ou ( Blumberg and Mellor, 1981). An important advantage to these stage S ophisticated modeling approaches is that the wind, tidal, and more density forcing functions are coupled explicitly within the model.
When modeled currents used as environmental input for oilspill models are not coupled within the specification of the hydrodynamics model, arguments of superposition must be used in sunaning the numerical estimation of a wind-driven flow field with a separate estimate of a density-driven flow field (see discussion on current field U3, below).

Spill Parameters:
Oil Fractionation: When oil is released into the environment, weathering processes begin to change its composition (Overton, · 1981). Processes included in this weathering include evaporation, dissolution, emulsification, absorption onto suspended sediments and detritus, photochemical oxidation, and microbial degradation. These weathering processes alter the physical and chemical properties of the oil, transforming it into several distinctly different types of petroleum residues. An oilspill model must address this weathering process through some partitioning of the spilled oil into physically and chemically different sub-classes. The approach taken in the URI Model is based S molecular weight and hydrocarbon species classification (see on gros Table 1). Spectral absorbtion analyses can give much more detailed information about the relative abundances of specfic compounds and Pou nd amalgams (Boehm et al., 1982;Gundlach et al., 1983) which com are of particular use in "fingerprinting" particular oil mixtures.
'.[be simple eight-class approach employed here matches the levels of complexity of the physical and chemical changes needed to describe the environmental partitioning of the spilled oil mass.

Oil Spill Rate:
Estimation of the rate at which oil is released into the marine environment from a well blowout hinges on instantaneous oil volume estimates derived from sea-surface oil area and thickness estimates.
Spatial extents are measured in the hundreds of thousands of square kilometers across the ocean surface and hundreds of meters in the water column. The hydrocarbons present on the sea surface may be from on the order of a few molecules to tens of centimeters thick, will assume irregular patterns, and often will undergo some dynamic subsurface entrainment as droplets which are entrained are displaced upw~ru to th~ ~urfa~~. Phytoplankton blooms and cloud shadows can be •isL~k~u for oil slicks (OSIR~ 1983), further confounding the estimation procedure.

CHAPTER 4
Ixtoc 1 Hindcast Input Data Description Table 1 summarizes the input data for the simulations discussed below.
Spill Paramaters: Oil Volume Input: TEMPERATURE: CURRENT (2) CURRENT (3) SUR.FACE OIL ADV!:CTIOfil    The three separate sets of wind and current fields are described because they are representative of the kind of environmental information which might be readily available for hindcasting a spill event. Comparison of the results obtained by use of each of the three sets is the major work of the thesis here presented. Figure 4 gives a schematic overview of the environmental data used.
Wind Fields: (Wl) (FNWC pressure-inferred wind field): The Pacific Environmental Group pressure inferred wind dataset (Wl) is an atmospheric pressure-derived monthly-averaged wind dataset on a three-degree latitude longitude grid spacing (7 x 5 nodes) which was generated by A. Bakun according to a methodology developed for application to the eastern Pacific Ocean (Bakun, 1973 Stolzenbach et al (1977) secondarily from Wu (1969); and correspond also to an empirical rule of thumb value based on observation of ocean winds under a variety of conditions (Stolzenbach et al., 1977).
The 7 x 5 pressure-inferred sea surface wind data for the Gulf Of Mexico have been bili_nearly interpolated (Wendell, 1972) onto a 34 X 24 grid system for model input. The primary usefulness of these wind field predictions is seen to be in defining the large scale spatial and temporal features of the wind field across the western Gulf. Figures 5A-D show four representative (Wl) wind fields at the l ied three degree spatial mesh. supp (W2) (Climatological "synthetic" yearly wind): The Dynalysis climatological wind dataset was constructed through a data-intensive interpolation procedure from National Climatic Center (NCC) TDF-11 data files (Blumberg and Mellor, 1981). The raw data (consisting of over a million Gulf surface ship observations) were edited and converted with the aid of standard bulk aerodynamic exchange formulas to produce monthly estimates of the wind stress statistics. The stresses were then interpolated onto a finer numerical grid by a statistical interpolation technique (Kantha et al., 1981). Since the (W2) wind record is designed to drive a hydrodynamic model, the higher frequency component of wind stress from the passage of weather systems would be lost by a simple monthly averaging technique. The model used has three time-varying parameters representing wind energies in the seasonal, cyclonic weather system passage, and diurnal frequency partitions as follows:  I  I  I  I  I  I  I  I  I  I  I  I  I   I   I
1 ·-· 2 ( ;. I 9-  I  I  I  I  I  I  I  I  I  I  I .  ....  -(Ul) (Classical Ekman wind-driven): The sea surface wind from the 3 degree mesh PEG wind field (Wl) were used to vectors a surf ace current using the Ekman solution for surf ace compute currents (Neumann and Pierson, 1966). No density forcing is included in this current field estimation procedure.
Ekman's solution for wind-driven transport assumes: 1) no boundaries, 2) infinitely deep water, 3) a constant vertical eddy viscosity, and 4) barotropic condition (l • F(p) only). All  Dynalysis turbulent closure current dataset is the product of an ambitious modeling effort by Blumberg and Mellor (1981). The model is driven at the surface by winds and surface heat flux derived from climatological atmospheric surface data from an intensive data analysis study (Kantha et al., 1981). were generated by a dynamic-topography-derived geostrophic model simulation by Blumberg and Mellor. Blumberg (personal communication) confirmed that the hydrographic data used as input to the current simulation was the same data set as that reported in Blumberg and Mellor (1981). This hydrographic data set is referenced as "the complete set of Gulf temperature and salinity data files maintained by the National Oceanographic Data Center (NODC) ••• archi.ved prior to 1979, and consists of over half a million temperature and salinity observations." (Blumberg and Mellor, 1981). Despite the large number of observations, the temporal and spatial coverage of the data set justified only four seasonal flow field estimates. Blumberg 3.5% downwind surface slick drift rate was used when the current field estimation procedure did not include wind forcing. The forcing mechanisms included and the coupling of these forcing mechanisms between the wind and current fields used as environmental inputs is discussed below.
Successful oilspill trajectory modeling is dependent on adequate description of both the hydrodynamic flow in the near-surface waters and the wind field. Most oil spill models simulate wind-driven surface water currents by some variant of the so-called "three percent rule" (Stolzenbach et al., 1977), an extremely simplified model for wind-driven surface water movement. Some approximation of n -wind-driven components of the water flow field is superposed the no with a wind-driven translation component to achieve a resultant Sl ation of the oil spill element within each time step. tr an previous hindcast experience (Spaulding et al., 1982) used several wind-driven simulations of a three-dimensional numerical hydrodynamic model to give a more precise description of wind-driven water movement for a nearshore spill. It was hoped that the use of the predicted current field from the Blumberg and Mellor (1981)  The location of the (W3) wind record is far from the spill site and inland. Both these characteristics make it a less than optimum wind record to force an Ixtoc hindcast. It would have been preferrable to acquire a wind record from the area of the spill site itself, but this was not possible. The NCC Brownsville wind record was acquired and used in this analysis because of its ready availability. In retrospect it is clear that the wind record from Buoy 42002, located at 26.0 degrees North, 93.5 degrees West, and supported by the National Data Buoy Center (NDBC), NOAA would have been a better choice. Efforts to make use of quickly obtainable wind statistics for Buoy 42002 (National Data Buoy Center, 1979 54 collected over the time span of the Ixtoc 1 spill to modify the wn sville wind record to more accurately reflect open ocean aro conditions were not successful using the technique of Williams and Godshall (1977). The monthly wind statistics computed for the Brownsville wind record were not suf f icently similar to those recorded by the NDBC for Buoy 42002 to enable this statistical technique to yield satisfactory results. The methodology uses a monthly mean wind matrix with several windspeed intervals and directions to characterize both the shore and the offshore wind records. A two-parameter transform for the shore wind record results from a comparison of the two mean wind speed matrices. The transform is subsequently applied to the shore wind record values. The methodology gave reasonable results for some months, but completely unreasonable results for others.
The use of the wind record to approximate the "three percent rule" wind-driven surface transport is consistent with the seasonal low band-pass filtering of the geostrophic ·solution of (U3).  ,, . . .  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  . . ,,. ·' ·"' .. --.
--------··· ·'· ' · • \ \ ' -. .  I  I  I  t  1  I  I  I  I  I  I  I  I  I  I  ....... Figure 9C), we see that 1xto the current field predicted by Figure   To the extent that the model is successful, one could expect the Dynalysis current field to give the best estimate of the Climatological features of the flow field. It is apparent from re 9C that the cyclonic and anticyclonic features of the figU southwestern Gulf described by de la Cerda (1975) are not in strong evidence in the (U2) current field.

C spill site (marked by an asterisk in
The (U3) current set is derived from salinity, temperature, and pressure profiles ensemble averaged seasonally over many years.  (Galt, 1981;Merrell and Morrison, 1981), but in substance is typical of the kind of approximations which are commonly used in the modeling of coastal oil spills. Galt's methodology for description of the coastal · current regimes along the western boundary of the Gulf is differentiated from this crude approach by the collection of a large amount of near-synoptic field observations collected by several federal and state government agencies involved in the spill response and by a hydrodynamic modeling procedure which created a mass-conserving flow field with composites of analytically-derived current patterns combined to achieve a "best fit" to these obseved data (Galt, 1981). The general descriptions of the cyclonic and anticyclonic features of the th western Gulf in Sturges and Blaha (1976) and Merrell and sou Morrison (1981) give needed overview of the general circulation of the area, but do not give a basis for prediction of the magnitude, direction, or extent of the Mexican Coastal Current as it extends the continental shelf. Field observations and current meter over data as employed by Galt (1981) Capurro and Reid (1970) is 19 degrees to 30 degrees Celsius, or 2 92 degrees to 303 degrees Kelvin, yielding a four percent variation in the denominator. Windspeed shows as the exponent of the natural logarithm base, and thus for an expected range of zero to ten meters second, yields a variation of 1 to 8,000 in the numerator. per Because of the relative insensitivity of the evaporation algorithm to temperature, a value of 28 degrees Celsius (Blumberg and Mellor, 1981)

Wind Drift Factor
For wind and current sets 1 and 2, a down-wind drift of the Spillet over the current field advection of 1% is used, based on experimental observations at the Argo Merchant spill site (Grose and Mattson, 1977). For wind and current set 3, a wind drift factor of J.5% of the wind speed downwind is used.
f.!edicted Spill Trajectories:   I  I  I  I  I  I  I  I  I  I  I  I  I  I I  I  I  I  I  I  I  I  I  I  I  I  I  I  forcing. Figure 13C was judged to be the best of the three in its prediction of the surface trajectory of Ixtoc oil (see Figure 14A, below). Figure 13A shows remarkably good prediction of the observed trajectories of Ixtoc oil, given its relatively simplistic origin.
The (U2)(W2) case, Figure 13B, is judged to be the furthest from the observed trajectories of the wind and current forced cases.  Comparison with Overflight Data: Figure 14A shows overflight surface oil data (Shuhy, 1979) from United States Coast Guard (USCG) flights on three consecutive days: 29, 30, and 31 August, 1979 (Julian days 241-243). The oilspill predicted trajectory for Julian 243 with (U3)(W3) forcing is overlayed with the USCG observations. These three days of coverage have been selected from the overflight data reports received from Shuhy (1979)   A second source of overflight information comes from a figure in the Galt (1981) paper, and represents an earlier overflight. Figure   14B depicts the (U3)(W3) spillet trajectory prediction and the mousse observations for this day. The observed trajectory is indicated to emanate from the spill site at right angles · from the predicted trajectory, showing a greater divergence of observed from predicted initial trajectory than for Figure 14A. Discussion: The poor predictions of both Figures 14A and 14B can be attributed to deficencies in both the wind and the current fields.
Inspection of Figure SC,  This comparison indicates the spatial differentiation of the wind field which has been mentioned above as a weakness of the use of the Brownsville wind data. Additionally, the cyclonic feature termed the campeche Gyre by Galt (1981) and described by de la Cerda (1975) would be likely to contribute to the southwesterly drift of the  Julian days 225 through 235 (see Figure 1). Heavy shoreline oiling was also reported by USCG overflight on Julian 235 between Latitudes 21 and 21.5 degrees North (Shuhy, 1979). These observed strandings of oil to either side of the simulation predictions again show the predicted spill trajectory to be less than satisfactory in describing the surface advection of the spilled oil.

Mass Balance Predictions:
Total Simulation Mass Balance Predictions:  water surface, water column (top 10 m), and shore. Column one is the continually increasing Julian day reference (see Figure 1). Column 2 is the cumulative mass input to the model, based on the estimates of  1980). Almost 30% of the derived mass. fraction is made up of the paraffins and aromatics with molecules of twelve carbon atoms or less. The low molecular weight fractions have a high vapor pressure and will very rapidly evaporate when the oil spreads out on the sea surface. Evaporation can account for as much as 60% of the mass balance of a spill of a light crude (Huang and Monastero, 1982).
Discussion: Beached Oil Entrapment: The routine used to entrap beached oil at the shoreline assumes a 100% loss of oil from the sea surface to the shore. This is a crude algorithm which will overestimate the oil mass lost to the shoreline (Gundlach et al., 1983). Of the 1.2 million gallons of Ixtoc 1 oil reported to have come ashore on Texas beaches, approximatedly 53 thousand gallons, or less than 5%, was found to persist near the shore in the form of sub-tidal tar mats (Gundlach et al., 1981). The heaviest oiling of the south Texas coast occurred between 29 August and 1 September 1979. A tropical depression crossed the shoreline on 13 September 1979, causing a 60 cm elevation of the usually less than lm tides and generating 1 to 2m waves.
Within two days over 90 percent of the oil on the shoreline was removed by wave activity (Gundlach et al., 1981). Hard-packed, fine-grained sand beaches characterizing most of the barrier-island coastline resisted oil penetration and in general were cleansed rapidly. A small section of mixed sand and shell beach near the center of Padre Island retained significant amounts of oil. (Gundlach et al., 1981).
The Water Surface prediction of Table 2 is seen to be an underestimate, and the Shore prediction an overestimate for the later days in the spill hindcast because of the simplistic algorithm of stranded oil discussed above.
Discussion: Wind-Forced Water Column Entrainment: The Water Column partition prediction is judged to be low,  .

Mass Balance Comparisons for 19 September 1979:
An environmental compartment mass balance estimate for 19 September 1979 based upon reports from the Researcher/Pierce Cruise has been reported previously (Spaulding et al., 1982). Only data collected during this cruise were comprehensive enough to allow an environmental partition mass balance estimate to be made on the basis of field-collected data. The values in Table 3 were generated on the basis of upper bound, lower bound, and best estimates. Specific citations of all data sources used in the estimation procedure are included in Spaulding et al (1982). Figure 15 shows a comparison of the model predicted mass balance in the partitions of atmosphere, water surface, and water column to the field-data-based estimate ranges for this day. Estimates of the amount of oil ashore were not possible with the field data available.
Model estimation of oil entrained in offshore sediments was not possible given the entrainment scheme used, which first introduces oil into the environment as a surface slick (Chapter 2).
The model prediction for oil mass entrained in the water column is less than the lower bound estimate based on field observations. This result is to be expected, given the two reasons discussed above.
The 5% of spilled oil mass estimated to be burned at the spill site was subtracted from the simulation spilled oil mass (Figure 3), and was an input, rather than a predicted variable in the simulation.
The best estimate value for the field-data-derived atmosphere partition was the common subjective judgement of several observers at the blowout site. Upper and lower bound estimates were derived from observed n-alkane concentrations in the air above the affected area.
The simulation prediction falls close to the upper bound of the field-derived estimate range. Questions of oil composition, subsurface entrainment, and oil weathering make these estimates

Conclusions
The URI/OSFh has been used to hindcast the Ixtoc 1 oilspill, using two reasonably simple and readily available wind and current dataset pairs and one state-of-the-art hydrodynamic model and its associated climatological wind field as environmental data inputs.
The simple and readily available environmental data gave better surface trajectory estimates than did the much more sophisticated The fact that the more simplistic environmental input datasets gave better trajectory simulations does not imply that the simpler approaches are better for use in equivalent oilspill trajectory simulations. Galt (1981) has demonstrated that successful trajectory modeling is achievable using a nested grid with a fine spatial mesh and coastal current observations taken in the field. The cost of the data-collection effort incorporated in the Galt (1981) modeling effort would be prohibitive for anything but a spill of the magnitude of the Ixtoc spill. The further development of coastal hydrodynamics modeling including finer spatial mesh grids and more complete specification of model boundary conditions with wind, density, and tidal forcing terms is the fundamental basis needed to improve oilspill trajectory formulations used in the modeling of coastal and continental shelf oil spills.

Wind Data Collection:
Underestimates of over-water wind speeds characterized the over-land collected wind data used in the simulations. An attempt to modify the land-collected wind data by the use of a statistical technique based on the monthly mean wind statistics of off-shore wind data gave reasonable wind speeds for only some of the months of the simulation. The monthly mean wind summaries for the Brownsville wind data were characteristically northerly and southerly for the winter months, while the buoy-collected wind dat3 had a more equal 112 distribution of wind directions around the compass.
Subsurface oil: Subsurface elevated hydrocarbon levels were predicted to exist near the spill source for the first two months of the spill. These predicitions were judged to be low estimates because: 1) low wind speeds from a land-collected wind record caused an underestimate in the wind-driven subsurface oil entrainment; arid 2) an overly simplistic oil-ashore routine in the simulation trapped 100 percent of all oil which touched shore onto the land, thus reducing the pool of oil on the sea surface available for sub-surface entrainment.
Environmental Partitioning of Oil Mass Balance: Mass balance predictions for the model simulation were compared with estimates derived from existing field data for a time 108 days after spill inception. Good agreement between the simulation atmosphere and water-surface predictions and field-data-based estimates was observed.