3. RESULTS AND DISCUSSION

3.1 Growth zones

The majority of the examined opercular bones indicated zones or inter-zonal rings. Each zone started with an abrupt inner edge which probably represents a sudden increase in the rate of growth after a period of slow growth leaving a transparent area on the bone equivalent to a winter ring (Lowe, 1952). In some cases in between the zones there were numerous striae which showed up as opaque rings. These rings were distinguishable from the zones by being sharply demarcated lines, whereas the zones, apart from an abrupt inner edge, fade away gradually on the outer side.

3.2 Age determination

Although it has still to be established, the assumption has been made in the following pages that the zones are laid down annually.

681 pairs of opercular bones were examined of which about 533 pairs (78.3%) had visible zones that were relatively easy to read. The oldest fish age group in the sample was estimated to be 6+ years.

Relationship between fish length and age of chambo is shown in Figs. 1–3. Analysis of covariance revealed significant difference in length-at-age between species (Table 1), thus each species was analysed separately.

Figure 1: Relation of fish length and age for (a) O. lidole, (b) O. squamipinnis and (c) O. karongae

Table 1: Analysis of covariance (ANCOVA) of length-at-age by species (P>0.05)

NB * = F1, Tabulated F(1,492) = 3.86

F1>F2

Table 2. illustrates mean lengths, annual increment and number at each age group for the chambo species.

Table 2: Chambo spp. Mean lenghts with the corresponding standard deviation and confidence limits (C.L.), annual growth increment (AGI) and numbers (no.) at each age group

Table 3. indicates the mean back-calculated total fish length at the time of annulus formation from 302 pairs of chambo opercular bones. The equations used to determine the relationship between fish length and opercular radius were as follows:

Table 3: The back-calculated total mean lengths (cm) of Oreochromis spp. at each age group

(a) O. lidole

(b) O. squamipinnis

(c) O. karongae

In both data sets the annual growth increment suggests faster growth for each species in the first year and much slower growth in subsequent years. Rapid linear growth in the first year for chambo juvenile may be necessary to reduce predation as they grow larger than most potential predators and also large sizes increase their swimming efficiency and speed.

Observed and back-calculated lengths at age (Tables 2 & 3) corresponded within each species. Comparisons of observed and back-calculated lengths did not indicate the occurrence of Lee's phenomenon. The absence of Lee's phenomenon suggests the presence of size selective mortality operating most heavily on the smaller fish of an age group (Bagenal and Tesch, 1978; Duncan, 1980; Francis, 1990). Therefore small differences in observed and back-calculated lengths at each age group were attributable to the growth after mark formation.

A general comparison of mean lengths at age obtained here with those reported in a previous chambo study (Table 4) was made. The mean lengths at age 1 were slightly larger and at ages 2–6 slightly lower than previously reported (Lowe 1952). The changes in mean length seems to be caused by the combination of various factors such as heavy fishing mortality, gear selectivity, environmental conditions, recruitment etc. (Turner, 1977; Toresen, 1990). The effects and interaction of such factors on growth of fish is beyond the scope of this paper.

Table 4: Chambo spp. Mean length at age group (After Lowe, 1952)

NB. * = maximum size seen

The mean instantaneous rate of increase in length and weight, G, at each age from this study and the previous one were computed using the appropriate growth equation (Table 5). There were no statistical differences in growth rate between species (Tables 6 and 7). This implied that all chambo species grow at the same rate and its growth rate has not changed for the past 40 years.

Table 5: Estimates of the average instantaneous growth rates in length and weight

NB. There is no difference in growth rates of chambo statistically (P>0.5) i.e. the growth rate of chambo is almost the same (see Tables 6 and 7)

Table 6: Analysis of variance of data from Table 2

Estimated values:

Tabulated F values (5%, 1%)
1. F(5,10) = 3.33, 5.64 F1>F(5,10)
2. F(2,10) = 4.10, 7.56 F2<F(2,10)

Table 7: Analysis of variance of data from Tables 2 and 4

Estimated values:

Tabulated F values (5%, 1%)
1. F(3,15) = 3.29, 5.42 F1>F(3,15)
2. F(5,15) = 2.90, 4.56 F2<F(5,15)

3.3 Theoretical growth curves

Values estimated for the parameters of the von Bertalanffy growth curves for chambo species using opercular bones are listed in Table 8. The values of Linf and K from both data sets for each species are mutually compatible. The Linf values for O. lidole ranged from 41.2–44.22cm, O. squamipinnis from 35.64–37.7cm and O. karongae from 39.13–46.1cm. The estimates of growth coefficient (K) for O. lidole were between 0.17–2, O. squamipinnis between 0.24– 0.31 and O. karongae between 0.16–0.21. Indeed samples collected by author and by fishermen show lengths above 40cm, 35cm and 38cm for O. lidole, O. squamipinnis and O. karongae respectively rather tending to confirm the values of Linf for all species. Thus, the growth parameter estimates reported here for chambo appear reasonable.

Table 8: Estimated von Bertalanffy growth parameters and growth performance indices

1. Length-at-age data

2. Back-calculated data

In the analysis of length-frequency data, it was observed that in most cases five components could be successively dissected (see Table 9). The resolution of the mean lengths of cohort components becomes uncertain for larger length groups. Also due to the extended breeding season of the different species of chambo, the modes in some cases are very close together. This creates difficulty in trying to follow a mode on a monthly basis. Mean lengths taking apart on year to year basis proved useful in growth parameter estimation (Table 10). In general the Gulland and Holt plot over-estimated k compared to the von Bertalanffy method.

Table 9: Mean lengths of components dissected from Bhattacharya analysis (Seisay, 1992)

1. O. lidole

2. O. squamipinnis

Table 10: Estimated growth parameters and growth performance indices using length-frequency data (Seisay, 1992)

1. Length-at-age data

A comparison of the growth parameter estimates obtained from three wholly independent approaches (opercular reading, back-calculation and length-frequency analysis) shows reasonable correspondence with the exception of over-estimation of k by Gulland and Holt from length-frequency data. The von Bertalanffy growth coefficients K from all approaches were generally lower than the values Moreau (et al, 1986) reported and those estimated from Lowe's data (see Table 11) for chambo species. The lower values simply indicates that growth of chambo species towards asymptotic size occurs at a relatively lower rate than 40 years ago. On the other hand, the results from both data studies indicate that O. squamipinnis exhibit a relatively rapid growth to an asymptotic length; the differences noted between two periods however, might be an artifact caused by difference in age reading.

In contrast the estimates of Linf for chambo species were higher than those Moreau (et al, 1986) reported and those derived from Lowe's data (see Table 11). The estimates in this study however, are reasonable as indicated by the largest specimen observed in the field. Further evidence comes from Lowe herself who also noted a specimen of 38cm for O. lidole (see Table 4). In this case the difference in Linf values is attributed to the scarcity of old, large fish in samples of the previous study resulting in under-estimation of Linf values (Samuel and Mathews, 1986; Barry and McFarlane, 1990).

Table 11: Estimated growth parameters and growth performance indices according to Lowe's (1952) data and Moreua (et al, 1986) report

Gwyther and McShane (1988) have indicated that comparisons between Linf and K are often confounded by the highly correlated nature of these two variables. Consequently the resultant growth curves obtained (Figs. 2–6) were also compared by using the growth performance index, O (phi) = 2*log10Linf + log10K developed by Pauly and Munro (1984) and to what is known of chambo growth (Moreau et al, 1986). The phi values obtained for all data sets were close, with slightly lower and higher values for the back-calculated length-at-age and length-frequency data respectively (Table 8). On the other hand these estimates compare favourably with the values obtained from the previous study and those reported by Moreau et al (1986) (Table 11) indicating that the present results are in close agreement with existing information concerning chambo growth.

Figure 2: von Bertalanffy growth curves for chambo species from length-at-age. Linf and K estimated using (a) Walford plot, (b) Gulland plot and (c) Munro plot.

Figure 3: von Bertalanffy growth curves for chambo species from back-calculated data. Linf and K estimated using (a) Walford plot, (b) Gulland plot and (c) Munro plot.

Figure 4: von Bertalanffy growth curves for chambo species. (Linf and K Least Square Non-linear Regression method from length-frequency data).

Figure 5: The growth curves of Tilapia saka (O. Karongae) and Tilapia squamipinnis (O. squamipinnis) by various methods (After Lowe)

Figure 6: Acomparison of growth curves for chambo species. (Banda = B and Lowe = L).

A common obstacle to fitting a von Bertalanffy growth curve is a lack of sufficient growth data for both small and large individuals. Smith and McFarlane (1990) have shown that length-frequency data might provide good estimates for K and Linf when the younger and older fish are well represented in the sample. Similarily Francis (1988) indicates that length-at-age data gives the same results when the younger and older fish are fairly represented in the sample. In view of this, on the basis of the biology of chambo species (Turner and Mwanyama, 1992) and the author's experience the Munro plot (using back-calculated data) and the Least square non-linear regression method (using length-frequency data) have produced better growth parameter estimates. The advantage of both methods is that the variations in Linf and K estimates are minimised thereby giving the best value estimates of the growth parameter estimates. Consequently the growth parameters considered to be reliable in this case are Linf= 43.2cm and K= 0.18 for O. lidole, Linf= 37.7cm and K= 0.24 for O. squamipinnis and linf= 41.3cm and K= 0.18cm O. karongae. Thus, the general growth curves for chambo species following this method are as shown in Fig. 3c. and the relationships between total length and age fitted with a von Bertalanffy growth curve for each species are shown in Figs. 7a-7c.

3.4 Verification

The interpretation of the ageing results has relied on the assumption that each seasonal band or ring on the opercular bone is produced as a result of a sudden but temporary growth acceleration taking place due to high food intake during phytoplankton blooms. From literature however, it has be indicated that phytoplankton bloom in L. Malawi do not exhibit a regular circuannual periodicity and can occur any time throughout the year. Lowe (1952) also pointed out that in some cases there were two bands in a year. This implies that the identification of the so-called seasonal bands each year should be accompanied by a close study of phytoplankton blooms throughout the year (which has not been done in this study). Consequently there is a need to validate this method. The validation probably would involve the analysis of a large number of opercular samples taken on a monthly basis throughout the year to determine the period of ring formation which corresponds to the onset of phytoplankton bloom. In this study however, the ages were indirectly validated by comparing the observed results to those by length-frequency analysis results which was an independent study. From the comparison, the results of the two methods were similar as already indicated above.

Figure 7: The general growth curve pattern for (a) O. lidole, (b) O. squamipinnis and (c) O. karongae. TTML = theoretical total mean length by the growth curve model.

3.5 Length-age Key tables

Table 12 shows length frequency samples of chambo species which have been used to construct the age/length keys for each species (see Table 13). The age/length keys simply indicate the percentage or fractional age frequency distribution for each length class of fish. Such keys could be used to determine age in situations where the length frequency is the only data available or where age determination is time consuming and expensive. Since the age/length keys have limited applications, the above keys can only be applied to commercial landing catch sites of the South East Arm.

Table 12: Length-frequency of chambo in each age group

1. O. lidole

2. O. squamipinnis

3. O. karongae

Table 13: Age-length keys for a commercial fishery of Oreochromis species (Proportion of chambo in each age group

1. O. lidole

2. O. squamipinnis

3. O. karongae

3.6 Conclusion

The mean length of chambo species at each age group caught in the south-east arm has generally descreased. The observed difference is attributable to differences in sampling procedures during the two studies although further research may be required.

O. lidole, O. squamipinnis and O. karongae appear to have the same pattern of growth. Growth is rapid before maturity which then slows down thereafter. They also seem to have the same ralative instantaneous growth rate.

Differences between estimates of the parameters of the von Bertalanffy growth equation between this study and earlier studies could be attributed to samples of different size range and distribution, and possibly biased by fishing strategy.

Therefore, the results of ageing method for O. lidole, O. squamipinnis and O. karongae seem reasonable and to justify the use of opercular bones for age estimates of chambo species under difficult field conditions (it is quick and requires minimum preparation). Indeed attempts to use other methods such as otolith microbands are often frustrated by technical difficulties in the field. Independent growth estimates for the same two species obtained from length-frequency distribution (which can be considered as an indirect way of testing the reliability of the ageing method) and from ageing gave similar results, making these results plausible. The growth studies therefore give some evidence that the ageing method using opercular bones is reliable and imply that the bands found in the opercula are indeed annual.

Finally it is important to note the following: 1- The interpretation of rings on opercular bones was difficult, and the use of two separate sets of reading could not completely exclude subjectivity in decisions on zone counts. The difficult in zones reading was further aggravated by the existence of multiple bands, ossiculation and fat deposition in opercula especially from fish > 25cm. 2- Linf and K values reported in this work represent an overall values for both the males and females fish. Preliminary analysis seems to suggest that there no great differences in the growth rates between the sexes. 3- The difference of to values between ageing method and length frequency analysis could be the effect of estimating the mean length of the young fish.