5. PRACTICAL EXERCISES

Actual data from surveys by several CECAF countries were brought to the workshop and used to illustrate calculations of stratified mean abundance indices and to compare coefficients of variation for different species. Brief summaries of these exercises and an overall comparison of them are presented here.

5.1 Morocco - Mauritania

A trawl survey for octupus (O. vulgaris) was conducted off the coast of Sahara from 21° to 26°N during February and March 1980. A total of 14 transects and 43 stations were occupied outside the 12-mi security zone in depths of 20–100 m (Figure 2). Although maximum abundance of octopus is known to occur inside 12 mi, the results are useful for description of the offshore distribution and abundance of octopus.

There appeared to be no relationship between abundance and depth in the 20–100 m zone but three strata were defined according to apparent differences in abundance as against latitude (Figure 2). The individual catches (kg/hour) are shown in Table 3 and the stratum means and stratified mean catch per hour and corresponding variances and coefficients of variation (/) are shown in Table 4. Strictly speaking, these estimates of variance may be biased because the pattern of stations within each stratum approximates a uniform grid (i.e., a systematic rather than random design); however, the bias is probably small. Even though the number of hauls is relatively low the coefficients of variation are fairly small (16% to 38% for individual strata, and 18% for the stratified estimate) indicating that octopus are rather uniformly distributed beyond 20 m. The distribution and variability may be considerably different inside 20 m.

5.2 Ivory Coast

The entire shelf of Ivory Coast out to a depth of 120 m was surveyed with a commercial trawl in 1978 and 1979. The area was first stratified into 6 depth zones: 20–25 m, 25–35 m, 35–50 m, 50–70 m, 70–90 m, and 90–120 m. Then each depth zone was divided into rectangles as indicated on Figure 3¹.

Within this scheme each rectangle was considered originally as a stratum; one subrectangle was chosen at random within each rectangle and a single haul was made in that subrectangle. However, beyond 50 m only every other rectangle was sampled in 1978 to reduce the sampling intensity by half. This sampling design represents a mixture of systematic and random sampling within the depth zones; the strata themselves represent a systematic pattern but the location of a haul within a stratum is random. Since only one haul is made per stratum it is not possible to estimate within stratum variances directly, and the method of collapsed strata must be used.

The precision of abundance estimates was estimated for two species, Balistes capriscus and Brachydeuterus auritus, and for the total catch (all species combined). Since the basic stratification scheme precluded estimation of within stratum variances, the method of collapsed strata was used in four ways. The first post-stratification (after the fact) scheme keeps the six depth zones but defines only three longitudinal sectors called San Pedro, Grand Lahou, and Grand Bassam - resulting in 18 depth/longitude strata². The second scheme considers only the six depth zones as strata, and the three longitude sectors are ignored. The third scheme uses two depths and three longitude sectors for a total of six depth/longitude strata. The fourth scheme uses only two depths as strata, 10–50 m and 50–120 m.

The coefficients of variation for the abundance estimates are shown in Table 5 for the four post-stratification schemes. The data show the importance of depth stratification for Brachydeuterus. Paradoxically, depth stratification appears to reduce precision for Balistes, but this is probably an artefact and merely suggests that depth stratification is not really useful. For total catch of all species, stratification by either depth or longitude does not appear very useful, due to compensatory distribution of species; that is, distribution of all fish (regardless of species) is more uniform than any one species because different species fill different niches.

It was noted that use of a high-opening trawl would have sampled pelagic species more efficiently, and that larger bobbins would have made it possible to reduce the extent of the non-trawlable areas. However, for the sake of continuity, it does not appear reasonable now to change the gear.

¹ The original data can be found in Caverivière and Champagnat (1978 and 1979)
² Only one haul was available for three strata in the first scheme. An estimate of within-stratum variance for these three strata was derived from the relationship between mean and variance for all the other strata, and the results of the single haul

Fig. 2 Strata and area of survey for Octopus vulgaris (February-March 1980)

Table 4
STRATUM MEANS AND VARIANCES, AND STRATIFIED MEAN AND VARIANCES FOR O. VULGARIS CATCHES OFF SAHARA (FEBRUARY-MARCH 1980)

Statistics	Stratum Estimates			Stratified estimates
	A	B	C
(kg/h)	19.4	34.1	15.4	22.6
V ()	54.9	29.5	7.3	15.9
CV (%)	38.2	15.9	17.5	17.6

= mean;
V() = variance;
CV = Coefficient of variation

Fig. 3 Trawl survey sectors in Ivory Coast (from Caverivière and Champagnat, 1979)

5.3 Summary of exercises

A summary of the results for the three sets of data examined for CECAF countries is presented in Table 6. In general, the coefficients of variation (15–45%) were not very large considering the relatively small sample sizes (numbers of hauls) suggesting that spatial variability is not excessive for the species and areas represented. By comparison, it may be noted that coefficients of variation for cod, haddock, and yellowtail flounder abundance indices off the US east coast range from 20 to 30% for stratified estimates based on approximately 60 hauls (Grosslein, 1971). The general conclusion from the exercises was that trawl surveys for these species and areas of the CECAF region should be more than adequate to monitor major trends in abundance.