FISHERY BULLETIN: VOL. 84, NO. 3 



liters of seawater are filtered on 13 mm HAWP 

 Millipore filters, using a syringe and Swinnex type 

 filtering cartridges. The filters are then stored in 

 a dark place at ambient temperature. When the 

 observing ship reaches Noumea, the filters are taken 

 to the laboratory for fluorescence measurements. 

 A 3-wk minimum time lag is needed between filtra- 

 tion and measurement, after which degradation pro- 

 cesses lead to stable fluorescent chlorophyll by- 

 products on the filters. The fluorescence (Ff) of the 

 filters is then measured without extraction, using 

 a specially adapted sample holder. 



The measurement error e is proportional to the 

 chlorophyll concentration C and can be expressed 

 as e = |SSCC-C|/C where SSCC is measured by the 

 non-extractive method while C is obtained by a more 

 conventional technique (Holm-Hansen et al. 1965). 

 Ninety-five percent of e values are <0.6 (Dandon- 

 neau 1982, and confirmed by later tests). This value 

 is probably an overestimate of e since it results both 

 from the error on SSCC and from the unknown error 

 on C. Different phytoplankton populations can also 

 result in different fluorescence to chlorophyll ratios 

 for the dry filters. This ratio has shown no signifi- 

 cant change between winter and summer conditions 

 around New Caledonia where a mixed regime alter- 

 nates with a stratified one (Dandonneau and Gohin 

 1984). The risk of a variation of the ratio in other 

 environments has not been examined, and must be 

 kept in mind. The few SSCC data points at latitudes 

 higher than 30° were not taken into account for this 

 reason. 



Calibrations 



SSCC is estimated using SSCC = k Ff where k 

 is a calibration coefficient that must be corrected 

 from time to time. Twenty milliliters from a sea- 

 water sample are filtered giving a fluorescence Ff 

 after 21 d of storage. A larger volume V from the 

 same sample is filtered on a glass fiber filter, 

 ground, and extracted by a volume v of 90% acetone. 

 The fluorescence of the extract is Fe . Knowing the 

 fluorescence to chlorophyll ratio of the fluorometer, 

 R , determined from a known solution of pure 

 chlorophyll a, we can estimate the following chloro- 

 phyll concentration of the seawater sample: 



C = (Fe x v)l(R x V); 



we obtain then k = Ff IC . 



k is sensitive to detrital material in turbid 

 coastal waters, so these main calibrations are made 

 during offshore oceanographic cruises. As such op- 



portunities are infrequent, secondary calibrations 

 are made more frequently with known solutions of 

 pure chlorophyll a, giving R t instead of R . We then 

 assume that k t = k x RJRq. This procedure does 

 not consider correction for chlorophylls b and c, nor 

 does it consider correction for phaeopigments, which 

 has recently proven to be uncertain when the fluor- 

 ometer is fitted with a commonly supplied blue ex- 

 citation lamp (Baker et al. 1983). Although the SSCC 

 data presented in this work are expressed in milli- 

 grams of chlorophyll a, they should be considered 

 only as indices of phytoplankton abundance. 



Data Rejection 



The crew members who take the seawater 

 samples and make the filtrations are voluntary 

 observers. Errors may occur which are difficult to 

 detect because, unlike temperature or salinity, 1) 

 any SSCC value in the interval 0-1 mg-m -3 , which 

 covers almost the whole data set, is a possible one 

 anywhere in the tropical Pacific, and 2) the auto- 

 correlation of SSCC decreases very quickly with 

 time or space, so that surrounding data cannot help 

 in error detection. Therefore, all the data are ac- 

 cepted, unless the filter exhibits an obvious fault 

 (i.e., breaking, stain, extraneous material). Occa- 

 sionally, all the data from a ship's voyage were 

 evidently too high, by a factor 3 or 5. Contamina- 

 tion by a polluted sampling bucket was the cause, 

 and the data from the entire voyage were rejected. 



Other possible errors are more insidious, such as 

 insufficient care in keeping the filters out of light, 

 or using an oxidized sampling bucket. These errors 

 result in slightly lowered values, but there is no way 

 to correct them and, in most cases, no way to even 

 detect these biases. Such data are entered in the 

 data bank. As a resulting constraint, any estimate 

 from this SSCC data set must be developed from 

 many data, in order to minimize the effect of a few 

 possibly biased values. 



Mapping Techniques 



In a previous work (Dandonneau and Gohin 1984) 

 the principles of objective analysis were applied to 

 compute best estimates of SSCC at a given place 

 and time in the southwestern tropical Pacific. The 

 studied area in the current study is much larger and 

 more complex, and the density of data is not high 

 enough to allow good estimates of the statistics of 

 the field. Hence, the use of an objective analysis of 

 the SSCC data has been excluded. The SSCC 

 mapped here on Figure 1 have been estimated using 



688 



