Welsh 



"potential vorticity," [A^V^ij; + f - fgh/hj,] . To implement the irro- 

 tational flow, the velocity potential A is computed at each time 

 step by solving (10). Then (11) is solved for the tendency 3(j;/3t . 



In order to solve for A , it is necessary to describe A on 

 the peripheral boundary of the prediction grid. Setting A to zero 

 on the boundary is not acceptable at the inflow to the prediction 

 area, so an experiment had to be carried out to determine reasonable 

 values. An area (see Fig. 2) was selected with the prediction area 

 inflow well surrounded. The 31-day mean analysis for March, 1960 

 was used for 1^ , and the velocity potential A was computed from 

 (10) using zero values on the boundary. The velocity potential is 

 shown in Fig. 3; the units have been converted to degrees Celsius. 

 Note that the rotational flow is directed along the surface iso- 

 therms, and that the irrotational flow is directed normal to the A 

 isotherms. 



Consider the velocity potential of Fig. 3. There appears to 

 be a boundary-value distortion on the lower boundary, and perhaps a 

 slight amount on the upper boundary, but the existence of a maximum 

 near 36N, 74W is definite. To the southwest, beyond the computa- 

 tional area, the velocity potential probably is shaped like a 

 trough which contains the Gulf Stream. Although the magnitude of 

 the irrotational flow is an order-of-magnitude less than the rota- 

 tional flow, its direction, more-or-less normal to the Gulf Stream 

 near Hatteras, allows its effect to be significant for surpressing 

 meander growth and stabilizing the Gulf Stream position. 



SEA-SURFACE TEMPERATURE PREDICTION* 



This section describes experimental sea-surface temperature 

 (SST) predictions made with the model. The first two experiments 

 were designed to test the flow from the SST conversion scheme. The 

 third experiment is a series of predictions from 5-day SST analysis. 



The initial field for the first two experiments was computed 

 from 31 days of ship reports for March, 1960. This long-term 

 analysis is smoothed some but still shows relatively small-scale 

 features, which may or may not be real. The first experiment uses 

 a conversion factor of 1000 m^ sec" °C , which gives a reasonable 

 mass transport for the Gulf Stream. The second experiment uses 

 3000 m sec" °C , which corresponds to a mass transport much higher 

 than generally suggested. The values for A^ and A. were taken 

 directly from Table 1; they are shown together with the flow con- 

 version factors in Table 2. For both experiments, predictions have 

 been carried out to 30 days after the initial map. 



The quantity predicted is ^ = k.T where Tg is the sea-surface 

 temperature. Because k is taken to be a constant, it is con- 

 venient to label i(j in °C and to refer to it as sea-surface 

 temperature. 



188 



