launched from the ship site, with initial wind velocities based either on 

 National Weather Service reports or on random selection from historic wind 

 data. Wind velocities for subsequent steps were generated from a wind trans- 

 ition probability matrix, as described earlier. Table 2-12 shows probability 

 estimates for 2 different days in the period following the grounding. The 

 sensitivity of probability estimates to initial wind conditions is evident. 



Significant quantities of oil may still be contained in the hull of the 

 Argo Merchant, and how the probability of this oil coming ashore will change 

 in the months to come remains a question. Model runs were therefore con- 

 ducted using different starting dates and initial wind velocities randomly 

 selected from historic seasonal wind data. Table 2-13 demonstrates the 

 sensitivity of probability estimates to seasonal conditions. 



In January, the spill left the area for which model input had been pre- 

 pared for the risk analysis. On the basis of data in the U.S. Navy Marine 

 Climatic Atlas for the North Atlantic (Meserve, 1974), model inputs were 

 prepared to continue forecasting oil trajectories in the Gulf Stream, al- 

 though various modifications to the model were required. 



The length of time steps used in the model was increased to 1 day. Winds 

 at each step were randomly drawn from local historic frequency tables, rather 

 than from transition matrices. Frequency charts of winds by month and loca- 

 tion were obtained from the climatic atlas, as were prevailing currents at 

 each step by season and location. Either a null current or the prevailing 

 current was chosen at random at each step according to the reported "persist- 

 ence" of the prevailing current. 



Trajectories are located in a latitude- longitude coordinate system with 

 displacements adjusted for earth sphericity. Model output takes the form of 

 displays of simulated trajectories on maps of the North Atlantic. 



Updated starting sites for model runs are determined by reported loca- 

 tions of a satellite-monitored NOAA drifting buoy deployed in the slick on 

 December 31. As yet, no account has been taken of long-term weathering 

 effects on the oil or the differential transport of the oil and the buoy. 



Figure 2-13 shows the locations at 30-day intervals of 100 simulated 

 trajectories launched from the reported position of the drift buoy on January 

 30. In this area of the Atlantic, the Gulf Stream divides into northern, 

 eastern, and southern flows. This accounts for the wide spread in trajec- 

 tories. In the future, when trajectories are launched from one division of 

 the Gulf Stream, model runs should show much tighter clustering. Figure 2-13, 

 indicates the northern flow to be the most likely of the three. 



2.3.5 University of Rhode Island, Department of Ocean Engineering 



Surface Currents 



C. Noll, P. Cornillon, and M. Spaulding of the Ocean Engineering Depart- 

 ment at the University of Rhode Island developed a simple computer model to 

 predict the surface drift of spilled oil shortly after the Argo Merchant went 



56 



