POWER and McCLEAVE: SIMULATION OF THE NORTH ATLANTIC DRIFT OF ANGUILLA 



(1977) was followed for determining the weighting so 

 as to provide increased numerical stability and to ap- 

 proximate more closely the solution to the continu- 

 ous equation. Further details on the numerical meth- 

 ods used were presented elsewhere (Power 1982). 



In approximating the advection-diffusion equation 

 by finite differences, the region under study is par- 

 titioned by a grid, and a difference equation is 

 derived for each cell formed by the grid. The differ- 

 ence equations express the concentration of the sub- 

 stance at the center of each cell formed by the grid in 

 terms of fluxes between adjoining cells. The end 

 result is a large system of simultaneous (difference) 

 equations, which can be repeatedly solved to obtain 

 the cell concentrations at successive time steps. The 

 region included in this study was the Gulf of Mexico, 

 Caribbean Sea, and the North Atlantic Ocean be- 

 tween lat. 10° and 50°N and west of long. 40°W (Fig. 

 1). Coastlines were approximated by cell boundaries, 

 as were the Caribbean islands and shoal waters of the 

 Bahamas. Flux of leptocephali across these bound- 

 aries was prohibited. Leptocephali approaching the 

 Bay of Fundy, Gulf of St. Lawrence, and the long. 

 40°W boundary of the model were permitted to be 

 transported out of the modeled area (dashed lines in 

 Figures 2-9). 



Currents for the model were calculated using ships' 

 drift data obtained from the National Oceanographic 

 Data Center (Fig. 2). A ship's drift observation is the 

 inferred surface current calculated by comparing the 

 ship's true position after a given period of steaming 

 with the navigator's dead reckoning position. Surface 

 current charts of the North Atlantic are derived from 

 the same data base used in this study. Each ship's 

 drift observation was resolved into an east and north 

 component, and the current component at the inter- 

 face between two cells was calculated as the mean of 

 all the appropriate current components recorded in 

 the 1° of latitude and longitude bisected by the cell 

 interface. The means were calculated by calendar 

 months, so for each month in the simulations a dif- 

 ferent current regime was used. Using June as a repre- 

 sentative month, the median number of observations 

 used to calculate a current component was 10, and 

 75% of the components were calculated using five or 

 more observations. The number of observations was 

 greatest within 5° of the North American coast, with 

 sample sizes >100 commonly occurring. Sample 

 sizes were poorest in the southeast portion of the 

 modeled area. An average of 3% of the cell interfaces 

 had no associated ships' drift observations. These 

 points where data were completely missing were 



46 



30 



19 



FIGURE 1. — Main portion of the geographic region included in the simulations, along with the 1° 

 by 1° grid and coastline approximation. Lettered cells are the starting points for American eel 

 leptocephalus drift simulations discussed in the text, and cells with stars are the starting points 

 in other simulations not presented here. Note for comparative purposes that points A and C and 

 points B, D, and F are the same meridians, while points A and B and points C and D share the 

 same latitudes. Cell with an X is the starting point for the European eel drift simulation. 



485 



