er than the monthly mean of Ekman transport 

 computed at synoptic intervals by an amount 

 depending upon the variance of the wind vector 

 within the month. In order to evaluate this 

 effect, a 54-mo series (January 1967 through June 

 1971) of monthly means of 6-hourly computations 

 of offshore component of Ekman transport was 

 generated for comparison with the corresponding 

 values computed from monthly mean data. The 

 results of the comparison are summarized in 

 Table 2. 



The high values of the correlation coefficients 

 indicate quite linear relationships between the 

 two sets of series. Thus the distortion introduced 

 into the series at any particular location due to 

 use of monthly mean data is mainly one of abso- 

 lute magnitude. The relationship of each monthly 

 value to the other monthly values at the location 

 is affected only slightly. This indicates that the 

 series computed from monthly mean data gives 

 an indication of the monthly variations in inten- 

 sity of upwelling nearly equivalent to that given 

 by the corresponding series of monthly means of 

 values computed each 6 hr. 



However, Table 2 shows the slope of the re- 

 gression line for the two series to be variable 

 from one location to the other. This indicates a 

 change with location in the amount of variability 

 of the wind vector within a month. Thus the use 

 of monthly mean data in conjunction with the 

 nonlinear stress law (Equation 4) gives an erro- 

 neous impression of the relative magnitudes of 

 offshore Ekman drift when different locations 

 are compared. For this reason the indices tabula- 

 ted in this report which are designed to indicate 

 temporal variations at a particular location, 

 should be used with caution as indicators of spa- 

 tial distributions. 



This distortion of the spatial distributions in- 

 troduced by designing the computational method 

 to minimize distortion of the time-series at each 

 particular location can be illustrated by exam- 

 ining the long-term mean annual cycle of off- 

 shore Ekman transport computed in two differ- 

 ent ways. Figure 4(a) is a time-series isogram of 

 offshore component of Ekman transport for a 

 long-term composite year computed from actual 

 wind observations taken from the National Cli- 

 matic Center's file of marine surface observa- 

 tions. The figure summarizes over 75,000 indi- 

 vidual wind observations taken by ships at sea 



Table 2.— Results of comparison of a 54-mo series of 

 monthly means of offshore component of Ekman transport 

 computed 6-hourly to the corresponding series computed 

 from monthly mean data at the selected coastal grid- 

 points. 



Rank Product- ^ ^^'^P^^''^ ^ 



Location correlation moment correlation " 



coefficient coefficient ^ 



line 



over the 20-yr period, 1948 through 1967, within 

 the 1-degree squares shown in the accompanying 

 coastline plot. The Ekman transport was comput- 

 ed for each wind observation and the offshore 

 components were averaged by month for each 

 coastal square. The drag coefficient used with 

 these actual observed winds was 0.0013. Bakun 

 (1971) demonstrated that the spatial distribu- 

 tions obtained in this manner agreed well with 

 the mean annual cycle of sea-surface temperature 

 distributions. Reference to the slopes of the re- 

 gression lines in Table 2 indicates that since this 

 drag coefficient is half that used with winds com- 

 puted from monthly mean pressure fields to gen- 

 erate the indices tabulated in this report, rough 

 numerical equivalence between the two trans- 

 port sets should occur at about lat. 45°N. 



Figure 4(b) is a similar isogram constructed 

 from long-term mean monthly values for the same 

 period of the indices derived from monthly mean 

 pressure data. The locations at which the indices 

 were calculated are marked on the accompanying 

 coastline plot. Comparison with Figure 4(a) shows 

 an extreme southward shift and intensification 

 of the point of maximum indicated upwelling. 



Figure 4(c) displays the data from Figure 4(b) 

 after having been adjusted to the slopes of the re- 



