FISHERY BULLETIN: VOL. 75, NO. 4 



peratures and high precipitation at Canton Island 

 in the central equatorial Pacific with intensifi- 

 cation of the Hadley circulation and changes in 

 the "southern oscillation." Quinn (1974) related 

 an index of the southern oscillation (the difference 

 of atmospheric pressure between Easter Island 

 and Darwin, Australia) with El Nino phenomena 

 and abnormally high rainfall in the equatorial 

 Pacific. One of the latter is the failure of the Peru- 

 vian anchovy fishery. The large interyear dif- 

 ferences of equatorial sea-surface temperatures 

 undoubtedly affect the biota in as yet undescribed 

 ways. 



It is of value, therefore, to bring the results 

 of monitoring into a form that is useful for fishery 

 applications. Toward this objective we have 1) up- 

 dated our previous Koko Head and Christmas 

 Island analyses through 1973; 2) analyzed the 

 long series (18 yr for Koko Head, 20 yr for Christ- 

 mas Island) and separated changes into long-term 

 variability, the annual cycle, and the short-term 

 variability that characterizes individual years; 

 and 3) speculated about the processes that affect 

 the changes evident in the data records. 



THE 1970-73 UPDATE 

 Sampling and Processing 



Koko Head, where bucket samples for tempera- 

 ture and salinity determinations were taken twice 

 weekly, is located at the exposed, eastern shore 

 of Oahu. At this location, cliffs extend into the 

 water, and temperature and salinity samples 

 have been found to be representative of offshore 

 conditions. At Christmas Island, bucket tempera- 

 tures were obtained daily near the plantation 

 village on the ocean side of the lagoon entrance. 

 Measurements were made during the morning at 

 each location. 



The procedures used to derive the harmonic 

 coefficients for the 1970-73 observations were the 

 same as those described by Seckel and Yong 

 (1971). Fourier analysis was performed on the 

 residuals from a linear fit so that the temperatures 

 and salinities are expressed as a function of time, 

 t, by 



S - K + bt + £ C n cos co (nt — a n ). 



(1) 



n=\ 



where K =F(t ) + ^, to = 2?, and k isthe highest 

 z 7/ 



768 



harmonic in the series. F(t ) is the first observed 

 value, A is the Fourier coefficient for n — 0, 

 C n are the coefficients for n =£ 0, and tx n are the 

 phase angles, b is the slope of the straight line 

 joining the first and last observations of the funda- 

 mental period, T. 



The fundamental period for the Koko Head 

 analyses was 365 days. For the Christmas Island 

 analyses the fundamental periods were 120 which 

 for a full year followed in sequence with a 30-day 

 overlap from Julian day 1 to 20, 91 to 210, 181 

 to 300, and 271 to 390 extending 25 days into 

 the following year. 



Results 



Results of the analyses for the update years 

 are presented in the appendices. Coefficients and 

 phase angles for the Koko Head temperatures and 

 salinities are found in Appendix A, Tables 1 and 2. 

 Figures of the expected values computed from the 

 harmonic functions together with the observed 

 values for the Koko Head temperatures and salin- 

 ities are found in Appendix B, Figures 1 and 2. 

 The coefficients and phase angles for the Christ- 

 mas Island temperatures are found in Appendix C, 

 and the plotted functions together with the ob- 

 served values are found in Appendix D. 



Standard errors of estimate for the fitted Koko 

 Head temperatures and salinities and Christmas 

 Island temperatures are listed in Appendix E, 

 Tables 1, 2, and 3, respectively. 



Christmas Island Data Problems 



Observer problems at Christmas Island caused 

 the sea temperature sampling to be interrupted 

 from May 1972 to April 1973. The data gap was 

 reduced by Hawaii Institute of Geophysics (HIG) 

 bucket temperatures obtained daily since Novem- 

 ber 1972 near the airport on the northeast shore 

 of the island. Although NMFS (National Marine 

 Fisheries Service) sampling resumed in April 

 1973, HIG data were used in our analysis for the 

 entire year. In our long-term analysis the remain- 

 ing data gap between May and November 1972 

 was closed by linear interpolation. Mean monthly 

 temperatures obtained from the two sampling 

 sites indicate that NMFS temperatures are on 

 average about 0.5°C lower than the HIG values 

 (Table 1). The HIG data have not been adjusted 

 to reflect this temperature difference. 



The large scatter of data at Christmas Island 



