temperature profile was obtained with a bathy- 

 thermograph. Samples of the bottom waters were 

 obtained with a Nansen bottle; the salinity of 

 surface and bottom waters was determined on a 

 conductivity bridge in the laboratory. Meteoro- 

 logical observations were recorded in World 

 Meteorological Organization codes. 



LABORATORY TECHNIQUES 



Plankton sorting was carried out under wide- 

 field, stereoscopic microscopes fitted with variable 

 magnification. 



All penaeid shrimp larvae were removed from 

 the samples and identified to genus by using keys 

 developed from descriptions by Pearson (1939), 

 Gurney (1942, 1943), and Heldt (1938). Identi- 

 fication of different stages of Penaeus larvae was 

 based on the work of Dobkin (1961). P. duorarum 

 is the only member of this genus reported in the 

 Tortugas area (Eldred, 1959); we found no larvae 

 of other species of this genus. 



Stages of other penaeid larvae also were identi- 

 fied. Genera encountered included Trachypeneus, 

 Sicyonia, Parapenaeus, and Penaeopsis, as well 

 as three unidentifiable species. Several species of 

 Trachypeneus and Sicyonia occur in the Tortugas 

 area (Eldred, 1959), but the specific identity of 

 these larvae was not determined. 



One-half of one of the two samples from each 

 station was usually sorted. The second sample was 

 retained for counting if ambiguities in the flow- 

 meter readings or other data should be encoun- 

 tered. Samples were divided into equal parts by 

 means of a two-compartment, rotating Plexiglas 

 cylinder similar to that designed by Folsom 

 (McEwan, Johnson, and Folsom, 1954). 



COMPUTER CONVERSION OF DATA 



On the basis of the methods of Ahlstrom (1954) 

 and Sette and Ahlstrom (1948), we designed our 

 data analysis to produce estimates of the abun- 

 dance of P. duorarum larvae and to put the infor- 

 mation in a form that could be reproduced easily. 

 The plankton analyses were punched on computer 

 data cards and processed at the Institute of 

 Marine Sciences. 



The original counts of the number of larvae in 

 each stage were transformed in two ways. In the 

 first, the numbers of each stage were divided by 

 the volume of water filtered through the Gulf V 

 sampler, giving an estimate of the number of 



larvae per cubic meter of water. In the second 

 calculation the numbers of each stage under 10 m.^ 

 surface area were estimated by multiplying 10 

 times the number per cubic meter by the depth 

 of water (in meters) at the station. Unless other- 

 wise specified, values in this report are based upon 

 the numbers of each stage under 10 m.- surface 

 area. All physical and biological data gathered 

 during this investigation were pubhshed by 

 Munro and Dimitriou (1967). 



VARIABILITY OF ESTIMATES OF 

 LARVAL ABUNDANCE 



Reproducibility of plankton hauls and confi- 

 dence Unfits to be appHed to plankton estimates 

 based upon counts of one-half of a single sample 

 were computed. For estimating this variabihty, 

 we counted the protozoeae in 47 pairs of repHcate 

 samples. After reference to the records of flow- 

 meter volume and water depth, counts were ad- 

 justed to give estimates of numbei's of protozoeae 

 present under 10 m.- of surface area (table 1). 



Following the methods of Winsor and Clarke 

 (1940) and SiUiman (1946), we carried out an 

 analysis of variance to determine the variability 

 of these estimates. Data were subjected to a 

 logarithmic transformation, resulting in a moder- 

 ately skewed normal distribution. The results of 

 the single-classification analysis of variance are 

 given in table 2. 



The mean square between replicate estimates is 

 0.07319 = (7E^. The standard deviation of a single 

 estimate i<rE) is VO.07319, or 0.27054 (a logarith- 

 mic value). The 95-percent confidence hmits are 

 the antilog of the 2aE value, \'iz, 2X0.27054 

 = 0.54108, and the antilog is 3.476. Percentage 

 confidence limits for the estimates are 1/3.476X 100 

 and 3.476X100, or 28.8 and 347.6 percent. 



These confidence limits are wider than those 

 obtained by SiUiman (1946) and Strasburg (1960), 

 who found that plankton counts at the 95-percent 

 level should be considered significantly different 

 if they were less than half or more than double 

 the sample compared. In the present study, the 

 "half-or-double" rule appUes only at the 70-percent 

 level of confidence. 



The confidence limits dealt with here, however, 

 include a greater number of sources of variation 

 and error than those of the above-mentioned 

 authors, because the present data were computed 

 numbers of larvae under 10 m.^ surface area; thus, 



PINK SHRIMP ON FLORIDA TORTUGAS SHELF 



167 



