FISHERY BULLETIN: VOL. 75, NO. 4 



first and rose in the second of these years 

 (Figure 3A). 



Heat gain across the sea surface cannot produce 

 a temperature decline and, therefore, other pro- 

 cesses must affect the temperature. One of these 

 processes is heat advection that, at the Equator, 

 is the product of the zonal current and the zonal 

 temperature gradient. A raft designed for under- 

 water biological observations was set out in Feb- 

 ruary 1964 near the Equator at about long. 150°W 

 and drifted westward 1,084 km (585 n.mi.) in 

 194 h (Gooding and Magnuson 1967) giving an 

 average speed of 155 cm s _1 . A current with the 

 speed of the raft, given a zonal temperature gra- 

 dient of 0.5°C/10° of longitude, would produce a 

 temperature change of more than 1.8°C/mo. A 

 slower surface current, 30 cm s _1 , was observed 

 on the Equator at long. 140°W during April 1958 

 (Knauss 1960). This current with the same zonal 

 temperature gradient as before would produce a 

 temperature decline of about 0.4°C/mo. 



The South Equatorial Current indices pre- 

 sented by Wyrtki (1974) reflect large variability 

 in the zonal current such as cited above. Addition- 

 ally, monthly charts of sea-surface temperature 

 (Eber et al. 1968) show the zonal gradient at the 

 Equator to range from zero to >1°C/10° of longi- 

 tude. Advection, therefore, is expected to play a 

 large role in the temperature variations observed 

 at Christmas Island. 



Near the Equator the wind field is a key element 

 in the evaporative heat loss, the cloudiness (affect- 

 ing the radiation flux across the sea surface), up- 

 welling, and in driving the equatorial currents. 

 Quinn's (1974) southern oscillation (SO) index is 

 related to the central South Pacific trade winds. 

 It is not surprising, therefore, to find coherence 

 in the changes of the SO index, Wyrtki's current 

 index, and the Christmas Island temperature. 

 Selecting the pronounced features of Figure 3B, 

 declining SO index values during 1956, 1963, 

 1965, 1968, and 1971-72 correspond with rising 

 temperatures. Increasing index values during 

 1964, 1966, and 1970 correspond with declining 

 temperatures. During the first series of years 

 South Equatorial Current speeds are declining 

 and during the second series they are increasing. 



SUMMARY 



In this paper we have used harmonic analysis 

 to make Koko Head temperature and salinity 

 time series and Christmas Island temperature 



778 



time series available for descriptive as well as 

 numerical applications. 



Time series data can be treated by a number 

 of mathematical procedures in order to elicit 

 important information. Initially, however, the 

 presentation of the data in graphical form is most 

 useful. The graphs in the appendices indicate the 

 nature of the annual variations, and Figures 1, 

 2, and 3 indicate the nature of the long-term 

 variations. 



Although spectral analysis is not the objective 

 of our work, the curve-fitting procedure further 

 serves the descriptive purposes in that it permits 

 separation of the time series into different scales 

 of variability (panels B, C, D of Figures 1, 2, 3). 

 For example, at Christmas Island the interannual 

 temperature variation is as much as four times 

 the average annual variation (Figures 3B, 5C). 

 Equivalent figures of Koko Head salinity show 

 that the interannual change can be about three 

 times as large as the average annual variation. 



Results of our analyses are also useful in 

 numerical applications. Coefficients and phase 

 angles (Appendices A, C, F) rather than observed 

 values can be used for further calculations. In 

 this manner the sampling variability apparent 

 in the graphs of Appendices B and D is filtered 

 out and variations of undesired duration can be 

 omitted. 



The separation of the time series into different 

 scales of variability is a mathematical procedure 

 and physical inferences must be made with 

 caution. For example, Figures 2C and 3C show 

 an annual cycle during every year although no 

 annual cycle was apparent during 1957 in Fig- 

 ure 2A or during 1963, 1964, and 1965 in Fig- 

 ure 3A. The procedure does not indicate whether 

 during these years the processes producing the 

 annual cycle were absent or whether they were 

 present but obscured by other processes. In 

 another example, a 12-mo and a 6-mo sinusoid 

 combine to reproduce the mean annual tempera- 

 ture cycle at Christmas Island. Again, the proce- 

 dure does not indicate whether there exists a pro- 

 cess affecting the temperature with a 6-mo 

 periodicity. 



Available information indicates that advection 

 is an important process affecting the observed 

 temperature and salinity variations. At Christ- 

 mas Island large changes in the zonal component 

 of the South Equatorial Current appear to cause 

 large variations in advection. At Koko Head 

 changes in the North Equatorial Current (Wyrtki 



