Figure 7.— Standard errors of estimate for harmonic fit of the 1971 depths at o, 25 to a, 25.8 (left panel) and 



at (T, 26 to i;, 27 (right panel). 



the depth is relatively large at all levels for low har- 

 monics. At the 26.0, 26.2, and 26.4 <t, levels significant 

 improvement in the standard error continues up to n = 

 10. Relatively high variance at (t, 26.6, 26.8 and 27 for 

 n = 6 indicates that low frequency baroclinicity, proba- 

 bly caused by eddies and current meanders, penetrates 

 deeper than 800 m. S^ approaches an asymptotic value of 

 about 20 m in the deeper a, layers indicating the magni- 

 tude of the residual variance associated with short-term 

 variability and sampling error. 



One would assume that the variance of temperature 

 and salinity decreases with increasing a, . Figures 5 and 6 

 and the tabulated variances of the temperature and 

 salinity (Table 2) show that this assumption is not cor- 

 rect. There is a tendency for S^ to decline from higher 

 values in the upper layers to lower values at o, 25.8 or 

 26.2, then to rise again to higher values at o; 26.4, 26.6, 

 or 26.8 before finally declining to the lowest values at cr, 

 27, 27.2, and 27.4. In other words, the T-S relationships 

 are less variable below the level of seasonal influence 

 than in the deeper layers at and above the salinity 

 minimum. The reason for this curious tendency may be 

 the uncertainty in the T-S relationship introduced by 

 the inadequate vertical sampling interval (200 m at a, 



Figure 6. — Standard errors of estimate for harmonic fit of the 1971 

 salinities at a^ 25 to cr, 25.8 (left panel) and at a, 26 to 7, 27 (right pan- 

 el). 



26.4-26.8) that was previously mentioned. It is also pos- 

 sible that the relatively large variance at the ct, 26.4-26.8 

 levels reflects active mixing of water masses upstream. 

 Depths derived from the harmonic curve fitting can 

 produce density inversions in the seasonal pycnocline 

 during spring and summer months. These inversions are 

 small and are due to the uncertainty of the harmonic fit 

 being of the same order of magnitude as the depth incre- 

 ment between cr, levels. The depth intervals between 

 successive <t, levels from the surface to the a, 25.2 level 

 range from 1 to 5 m. The standard errors of estimate 

 listed in Table 2 generally are larger than these values 

 and range from 5 to 14 m. These uncertainties in the 

 depth estimates are unavoidable with the method 

 employed, but the errors produced in computations of 

 the integrated properties of the water column, e.g., 

 mean temperature or heat content, will not be large 

 since the depth increments in the upper layer are so 

 small. The levels producing inversions can be deleted in 

 these computations without significantly affecting the 

 results. For example, consider the mean temperature of 

 the water column from the surface to the 0; 26.0 level on 2 

 days at OWS-V representing summer conditions, 1 

 September 1969 (Fig. 8) and 1 July 1971 (Fig. 9). The ex- 

 pected depth values for 1 September 1969 revealed an in- 

 version at (T, 22.8 and 23.0. Deleting these two levels, the 

 mean temperature was calculated to be 17.20°C, com- 

 pared with a value of 17.24°C calculated from the 



11 



