The sampling of fish stocks is crucial for the management of reservoirs. It is important to know what fish species are found in the waters, what their densities are, and how the stocks react towards changes in management. Although the need for fish stock sampling is apparent, proper sampling is not obvious.
Multispecies sampling is difficult and requires a good balance of different sampling gears. A single fishing gear is seldom enough to adequately sample several fish species. Samples taken by gillnets in reservoirs in Zimbabwe were often dominated by O. mossambicus, while in samples taken by seine nets in the same reservoirs T. rendalli dominated (personal observations). Gulland (1980) considered gillnets, when not used in combination with other gears, unsuitable for multi species studies.
The Drottningholm method caught a wide variety of fish species, but since no other fishing methods were used in combination, it could not be established whether all species were identified. In reservoirs that were sampled on several occasions not all species were caught on every occasion, indicating that the method did not catch all catchable species. This was partly due to the low abundance of some species, but low catchability probably prevented some species from being caught regularly.
When ALCOM started the work on reservoir fishery in southern Africa, little was known about the species compositions of most of the reservoirs, about their fishery, and about the dominant species in the fishery. The Drottningholm method was easy to use with a minimum investment in manpower and equipment, and gave a good picture of the fish species present in the various reservoirs in the region. As a method to collect general data on fish populations the method was useful.
CPUE is proportional to the abundance and the catchability. To define the catchability is not obvious because it is not a single process. With the sampling nets the pooled CPUE was estimated with a high precision at most sampling occasions. However, in order to be useful as an index of abundance a period had to be found with minimal fluctuation in catchability. From the results it has to be concluded that this was not possible. Furthermore the duplication of sampling of the Mbvoniha and Chizombezi dams proved that the catchability fluctuated significantly from day to day. These daily fluctuations made that the CPUE as obtained with the Drottningholm method could not directly be used as indicators of abundance for the separate species nor for the whole fish population.
It proved more difficult to estimate a CPUE with a certain precision for the individual species. Large numbers of nets were needed to precisely sample most species. B. paludinosus could be sampled with good precision in most reservoirs, but no period for accurate sampling could be identified. Fluctuations in catchability remained high throughout the year.
It has to be stressed that the catchability is species specific, and that the influence of outside factors varies between species. The CPUE for one species is a different proportion of its biomass than for an other species. Using the CPUE data for the individual species as a quantitative description of the species distribution of the fish population, as done in several studies, is therefore not justified.
The CPUE of the reservoirs in Zambia was significantly lower than in the other countries, indicating a lower biomass in these reservoirs. It is well possible that the reservoirs held fewer fish since other studies had shown that the reservoirs in Zambia were intensively exploited (van der Mheen-Sluijer, 1991), while the exploitation of the reservoirs in Botswana (Sen, 1990) and Lesotho (Westerlund, 1994) was minimal. As such the method was capable of providing information concerning the exploitation level of a reservoir. However, it has to be mentioned that the differences in exploitation between Zambian reservoirs and others were considerable and it remains questionable whether the method can identify changes in the biomass as a result of less significant changes in management.
The method used the CPUE as an index of abundance. The logarithmic transformation of the CPUE data was useful to dissociate the means and the variance. However, it did not bring the distribution any closer to a normal distribution.
Gillnets are not only species selective but also size selective. To compensate for this effect the Drottningholm method used a variety of mesh sizes. The direct estimation of selectivity for O. shiranus showed that the gang of mesh sizes was still selective, favouring larger specimens. In the mesh range of 16.5 to 43 mm. a fish with a total length of 250 mm. had a ten times greater chance of being caught than a 100 mm. fish. For the whole range of meshes from 6.25 to 75 mm. it can be expected that the difference in vulnerability to the gear between the largest and smallest fish was even greater.
The combination of data concerning two or more fish species to make a common estimate is always difficult to justify. However, the pronounced selectivity of the Drottningholm nets for size, made it debatable whether data of different sizes of the same species could be combined. Since the gear was much more efficient for larger fish than for smaller, a reduction of the CPUE from one year to another may not be the result of a decrease in biomass, but may be caused by a change in population structure.
The fact that larger mesh sizes were more efficient than smaller is commonly known, and is influenced by the reduced visibility of the larger mesh sizes and the increased activity of larger fish. The leading effect of a gang of mesh sizes, leading the larger fish along the smaller meshes towards the larger meshes may also have contributed to this effect. This leading effect can be minimized by leaving gaps between the meshes. However, it is best to set different mesh nets at separate locations, far enough apart that they do not compete for the capture of the same fish. Hamley (1975) mentioned that because nets of smaller meshes are less efficient, their surface areas must be correspondingly larger. This recommendation may be viable in a mono species study. However, when more species are studied, all with different gillnet selectivity curves it becomes impossible to adjust the sizes of the nets for selectivity.
The pronounced selectivity for larger O. shiranus of the nets as determined by Mattson (1995) might, at least partly, have been caused by the siting of the nets. The sampling nets were 1.5 metres high and were set in water of at least that depth. The small and young specimens of the Oreochromis species are mainly found in the shallow parts of a waterbody and often reside between the weeds. Hakkens (1991) showed that in Zambia small O. andersonii were only caught in areas with aquatic vegetation and were not caught in areas with a depth above 2 metres. This would mean that the gear was not selective in itself but is set in a habitat favouring the catch of larger fish.
Degerman et al. (1988) found a positive relation between the size of a lake and the number of nets needed to sample perch (Perca fluvialitis) with a certain precision, which was possibly an effect of the high habitat heterogeneity in larger lakes. For roach (Rutilis rutilis) this relation was not found. In the reservoirs sampled during this study no relation between size of the reservoir and the number of nets required to reach a certain precision could be detected. The smaller range of surface area (0.3–400 ha) and the sampling in the 0–3 m zone only, reducing the habitat variation, can be an explanation for this. The fact that only the maximum surface area of a dam was recorded and not the surface area at the time of the sampling, which may have been significantly different, reduced the accuracy of testing the relation between surface area and number of nets needed for sampling. The catches in Semarula dam showed that less nets were required to sample the fish stock at the end of the dry season then during the rainy season. The larger surface area combined with a consequent higher habitat heterogeneity could have caused this.
The reservoirs in southern Africa are formed by the construction of a dam across a river. These rivers are in most cases seasonal. As a result the water level in the reservoirs fluctuates between seasons and fluctuations differ between years. An enormous increase in area under water can occur in one year while the next year water fluctuations are only marginal. Merron and Bruton (1993) showed that the floods have a great influence on the biomass of a fish population and on the species distribution. Irregular floods or non-existing floods resulted in a skewed biomass distribution favouring a stunted population of O. mossambicus, while flooding favoured riverine species.
The abiotic factors influencing spawning success, growth, and the sampling itself vary from year to year in a reservoir and great fluctuations in numbers, biomass, and CPUE will occur as a result of these fluctuations. It will be extremely difficult to identify to what extent the changes in the CPUE are a results of changes in the biomass, and whether the changes in the biomass are a result of changes in the management of the reservoir.
Now that information about the fish species for the various countries is known, further and more detailed work should concentrate on the target species. It is important to establish reliable estimates of the growth and mortality parameters of the most important species. In order to estimate these parameters, it is necessary to carry out a length based assessment. Gears should be adapted to sample the most important species adequately. For instance, B. paludinosus has been mentioned in reports as a target species for improved yields from reservoirs, while still only the two smallest mesh sizes of the sampling gear catch this species efficiently. For a study of this species and for monitoring the effect of management strategies on this species, it has to be sampled with a wider range of gears. This also applies for C. gariepinus. This may necessitate the use of other gears. For the riverine species sampling and monitoring of the spawning migrations will be necessary.
Since the reservoirs in southern Africa are not stable environments it is important to determine to what extent the fluctuations in water levels and water quality influence the fish production. Future work should concentrate on how the management of the reservoirs can be adapted to optimize the production, both biologically and socio-economically, of these fluctuating water systems.
Besides the sampling by the project itself, more effort should be made to collect catch data from those reservoirs where an active and accessible fishery operates. Fish stock monitoring often only gives good results after a long period of recording. For the project it is difficult to commit itself to sample reservoirs over a long period, and many governments in the region do not have the means to do it. Catch recording by the fishermen themselves, with periodic supervision and support, offers the best chance of collecting longterm, detailed and regular data on reservoir fishery. This may require that some fishermen are allowed to fish with a range of mesh sizes, and not only with the legalized larger mesh sizes.
