From 1966 to March 1968 the sampling was intermit- 

 tent (Fig. 1). Subsequently, until January 1972, daily 

 sampling became nearly continuous with isolated gaps 

 of up to 20 days. In addition to the temporal gaps, sta- 

 tions were occasionally occupied outside the nominal 

 location, a 10 nautical mile (nmi) square centered on 

 lat. 34°N, long. 164°E (Fig. 2). We, however, used all 

 observations within a 60 nmi square centered on this 

 location. Sixty-seven percent of the stations were oc- 

 cupied within the 10 nmi square and 95% within the 60 

 nmi square. 



The oceanographic station data were obtained from 

 the National Oceanographic Data Center (NODC) of 

 the National Oceanic and Atmospheric Administration 

 in two different formats: 1) temperature and salinity at 

 observed and standard (interpolated) depths, and 2) 

 temperature, salinity, and depth at increments of 0.2 

 sigma-t (ff, ) units between the surface and the deepest 

 observation (isentropic format). A cubic spline interpK)- 

 lation function was used to obtain the values of temper- 

 ature, salinity, and depth at the desired a, values.^ 



The Isentropic Format 



The study of changes in water properties on surfaces 

 of constant potential density or a', called isentropic 

 analysis, was introduced to oceanography by Parr (1938) 

 and Montgomery (1938) and has been a valuable tool in 

 descriptive oceanography. The assumptions underlying 

 isentropic analysis are that mixing or interchange of 

 water masses on a constant density surface proceed with 

 a minimum of change in the potential energy and 

 entropy of the system and that surfaces of constant den- 

 sity are the preferred surfaces along which mixing oc- 

 curs. The method has been used in the identification 

 and tracing of water masses and also lends itself to the 

 analysis of temporal changes in the water structure as 

 was done, for example, at OWS-P by Tabata (1965). In 

 order to pursue the latter aspect of isentropic analysis 

 the temperature, salinity, and depth are presented as a 

 function of a, in this report. 



Quality Control of Isentropic Data 



The initial step of quality control is concerned with 

 the interpolated temperatures and salinities at 

 designated o^ values for each station received from 

 NODC. The spline interpolation function calculates the 

 temperature and salinity at designated a, values from 

 the two relationships: 1) observed temperature versus a, 

 and 2) salinity versus o; with a, as the independent 

 variable. A a, value was computed from the interpolated 

 temperature and salinity at each designated (t, using the 

 relationships (1) and (2). When the calculated a, dif- 

 fered from the designated by >±0.02 a, units at more 

 than one level, both interpolated and observed data 

 were used to plot two temperature-salinity (T-S) dia- 



grams for the station. The T-S curve drawn through the 

 observed values together with the temperature-depth 

 curve was used to correct the interpolated values ob- 

 tained from the spline interpKjlation or to determine 

 whether the station should be rejected. These quality 

 control procedures frequently had to be used for values 

 that were erroneously produced by the spline interpwla- 

 tion because of an inadequate sampling interval in the 

 vicinity of the salinity minimum (a, 26.8). Out of a total 

 of 1,067 stations, 496 stations were corrected during this 

 quality control procedure. 



Harmonic Analysis of Isentropic Data 



Harmonic analysis was used as a curve-fitting 

 technique to summarize the daily oceanographic data in 

 a manner that lends itself to studies of temporal changes 

 in the water structure. The resulting analytical expres- 

 sions yield smoothed estimates of the water properties 

 for any time of the analysis period. 



Harmonic (Fourier) analysis was performed on the 

 isentropic data for each year from 1966 through 1971 

 with a fundamental period of 366 days beginning on 1 

 January of one year and ending on 1 January of the next 

 year. In these data sets, values at the beginning and end 

 of the annual series were usually not equal and to 

 facilitate the curve-fitting procedure the harmonic 

 analyses were performed on the residuals produced by 

 subtracting the linear trend (day 1 to day 366) from the 

 observed value: 



Rit) = fit) - [/(I) + M-(t -1)] 



(1) 



'Hamilton, D. 1973. Isentropic analysis and spline interpolation. 

 Natl. Oceanogr. Data Cent. Tech. Rep., 29 p. 



where fU) is the value of the property (temperature, 

 salinity, or depth) at a constant (t, at day t, and t = 1 to 

 366. The slope of the straight line fitted to the values at 

 day 1 and day 366 is 



M = [/(366) - /(I) 1/365. (2) 



The expression for the fitted curve then becomes 



E(t) =/(l) + M -(t - I) + F^{t) (3) 



where FAt) = A^ + x(a^cos^SI^ + B„ sin ?^\ 



n = 1,2, . . .k (4) 



is the Fourier series derived from the residuals, R(t). 

 The coefficients A„, i4„, and B„ were evaluated by use of 

 a standard computer program. 



Data gaps in the time series were filled by linear inter- 

 polation between observed values in order to satisfy the 

 program requirement of equal time intervals between 

 data. When no observations were available for the first or 

 366th day of the year, data from the first (last) actual 

 station of the year was extrapolated backward (forward) 

 to the missing day. For the 1969 and 1971 analyses, the 

 values for the first day were computed from a harmonic 



