7. CATCH AND PRODUCTIVITY

7.1 Productivity

Kapetsky (1974) derives the productivity of five species of fish from the Kafue River using a “floodplain” model for mortality and growth as detailed in Section 5.6. The total figures for fish of age two and older range from 76 kg/ha/year and 56 kg/ha/year for S. andersoni and S. macrochir, two herbivorous and long-lived species, to 12 kg/ha/year and 6 kg/ha/year for Hepsetus odoe and Serranochromis augusticeps two short-lived predators. A very large percentage of the production (85–99 percent) of fish older than age two occurs during the third year of life. As a general rule, the production during the first two years of life contributes a substantial portion of lifetime production. Thus, the total lifetime production of the five species investigated by Kapetsky is probably considerably superior to the 199 kg/ha/year estimated for fish of age two and older.

7.2 Standing Stocks

A limited number of standing stock estimates have been made either by poisoning or exhaustive fishing in several river and floodplain areas.

Main river channel: During high water the mean standing stock of fish in a section of the main Kafue River channel was assumed to be similar to that of the open water lagoon of the floodplain (University of Michigan, 1971), i.e., 337 kg/ha. During low water Kapetsky (1974) found 106, 386 and 576 kg/ha for three stretches of the Kafue River and a mean value of 348 ± 59.5 kg/ha for the whole river.

At the beginning of the dry season estimates of 341, 227 and 141 individuals/ha were made from three reaches of the Bandama River by Daget et al., (1973). The equivalent ichthyomass for one part of the river in January was 125 kg/ha and in May estimates of 50, 177 and 305 kg/ha were made.

Loubens (1969) estimates a range of 370–5 620 (mean 2 435) kg/ha of fish to be present in the same dead arm of the Chari river at different times of sampling. In other stretches of river standing stocks between 870 and 2 170 kg/ha were obtained. The river estimates are somewhat high as compared with the estimates from the Bandama and Kafue rivers, whereas those of the dead arms (which correspond to oxbow lakes) are of the same order as the ichthyomass found in lagoons in the Kafue system.

Lagoons and floodplain: During high water the University of Michigan (1971) report an ichthyomass of 2 682 kg/ha from the vegetated waters of the lagoon, 337 kg/ha from open waters and 64 kg/ha from grass marsh.

During the dry season when the water had fallen and the lagoons separated from the main channel, estimates of 3 306 and 2 693 kg/ha were made for two different lagoons. Within three months these had fallen to 684 and 501 kg/ha respectively. Kapetsky (1974) found the largest concentration of fish (879 ± 267.9 kg/ha) in vegetated areas of the lagoons and lower concentration of 575 ± 122.8 kg/ha in open water areas of lagoons.

Holden (1963) found that the standing stock in floodplain pools of the Sokoto River varied widely with bottom type. The intermediate (mixed sand-mud) bottomed pools supported 1 029 ± 536 kg/ha; sand bottomed pools 475 ± 148 kg/ha and mud bottomed pools 144 ± 83 kg/ha. The mean ichthyomass from all types was 415 kg/ha which spread over the whole floodplain (of which the pools comprise 3 percent) amount to 12–17 kg/ha. Differences in pool form were also found to influence ichthyomass in the Senegal River (Centre Technique Forestier Tropical, 1972a). Here yields were high from elongated channel type pools (375–200 kg/ha) and much reduced from circular pools 1.5–20 kg/ha. Mean yield was low (37.5 kg/ha) from this area and a total tonnage of 260 t was estimated for the 6 600 ha of floodplain pools.

In one open water pool of the Yaérés floodplain a standing stock of 700 kg/ha was estimated by Loubens (1969).

The most complete studies carried out on floodplain pools have been done on the artificial whedos of the Ouémé delta. Here observations by a number of workers have been summarized by FAO (1971) and are here presented (Table VII; Fig. 12) for comparison with other areas. Ichthyomass is generally high although it is of the same order as that found in the intermediate pools of the Sokoto and some Kafue lagoons.

TABLE VII

Standing stock of some “whedo” floodplain pools of the Ouémé floodplain

Figure 12 Scatter plot of the relationship between surface area and yield in the “whedo” floodplain pools of the Ouémé delta

The total ichthyomass in the floodplain pool system of the Ouémé is estimated at between 2 255 and 1 719 t for the 1 095 ha of whedos.

As mentioned by Holden (1963) the high yields of floodplain and pool fisheries, and particularly the whedos, are achieved by the “drain-in” principle which concentrates fish from the whole floodplain area. If the ichthyomass of the fish from the pools is extended over the whole area of the plain, the standing stock is between 61.8 and 47.1 kg/ha.

Total system: From the above it seems reasonable to expect standing stocks of about 300–500 kg/ha from the river component of the system and in excess of 1 000 kg/ha from the floodplain lagoon and pool component.

The University of Michigan (1971) estimates were combined to give estimate of total ichthyomass in dry and wet season for the Kafue system (Table VIII).

TABLE VIII

Estimates of ichthyomass for dry and flood seasons for the Kafue Flats system (after University of Michigan (1971))

7.3 Catch

The theoretical relationship, catch for km₁= 0.0064 (distance km₁ from source)^0.954 calculated for African rivers by Welcomme (in press) indicates that the catch per kilometre of river increases with distance from its source. This increase is probably due mainly to the increasing area of floodplain that forms part of the normal evolution of the channel as it approaches the mouth. That catch for any reach of river depends on the area of floodplain of that reach is shown in Fig.13 which plots floodplain area against catch for different kilometres reaches of the Niger and Benue Rivers. The relationships-Catch = 2.65 (Floodplain area) - 0.98 correlates well with the data (r = 0.91; S_y·x = 2.5).

The catch from 12 floodplains is listed in Table IX.

The mean catch from all floodplains was 37.48 ± 23.06 kg/ha/year. There is however reason to believe that the Yaérés, Barotse and Okavango floodplains are considerably underfished. Similarly Motwani (FAO/UN, 1970) reports that the present levels of yield are only moderate with respect to potential in the Niger and Benue Rivers in Nigeria. Of the other floodplains mentioned in the table, all appear to be fished at a high level. Considering only these, a mean catch of 49.53 ± 18.98 kg/ha is estimated. It seems, therefore, reasonable to expect catches of this order from most floodplain areas (i.e., a potential catch of 834 000 t for the floodplains listed in Table I).

Figure 13 Catch per kilometre reach of river related to the floodplain area in km² of the river reach

TABLE IX

Commercial catch from 12 African floodplains

7.4 Fluctuations in Catch between Years

Three sets of data exist upon which an assessment of the effects of differences in hydrological regime upon the fish catch can be based. None of these sets of data include any reliable measure of fishing effort, which may itself be responsible for year to year, or season to season variations. Extremely good correlations between some hydrological indices and catch can nevertheless be obtained despite the reserves expressed by Williams (1960) and Muncy (in press), as to the validity of such a procedure.

In the Kafue River fishery water level and catch records are available from 1954 until the present. However, the fishery is not judged to have reached maximum expansion until 1958 (Muncy, in press) and the flood regime of the river has now been altered by the construction of a downstream dam in 1972. Calculations have therefore been based largely on the years 1958–71 although they have also been carried out less successfully for longer period. University of Michigan et al., (1971) found a correlation (r = 0.77) for catch against the amount of water retained in the system, represented by a Drawdown Factor (DDF) in the preceeding season, with the following relationship: Y'_n = -6 630 + 1 830 log DDF_n-1. University of Idaho (1971) found correlations of a similar order using Flood Index (FI) as a representation of intensity of flood. These were T = 0.72 on the preceding year and 0.71 on two years previously. Muncy (1973 and in press) presents a complete analysis of these factors. In summary he found a very weak negative correlation (r = -0.171) for water stage (derived from September-December water levels) and catch in the same year. This is attributed to a lessened efficiency of fishing during more intense floods. Good positive correlations exist between catch and flood intensity of the preceding years (y-1 and y-2).

Sets of data for water level and catch also exist for the Shire River (1969–1973) and the Central Delta of the Niger (1966–1974)¹ which have not before been analyzed. The following analysis of the data from the three systems uses Kapetsky's HI1 (see 4.1) to investigate the effects of flood regimes on catch in these river systems.

¹ Based on landings of fish at Mopti only.

TABLE X

Correlation coefficients for linear, semi-log and log-log regressions of catch against HI 1 in the same year (y) one year previously (y-1) and two years previously (y - 2) for three African floodplain systems

( ) Significance
* Not calculated

From Table IX it appears that highly significant correlations exist between catch and flood regime in the previous years (y-1) in all three systems. Better correlations may be obtained in some cases using Semi-log or log-log transformations although the improvement is not outstanding and linear relations have been used for further analysis throughout. Correlations with the HI of the same year are not significant in two systems and are less significant than both HI_y-1 and HI_y-2 in the Niger. The flood regime of y-2 is significant only in the Niger.

On biological grounds and from statistical data in Table X, it might be expected that the flood regime in both preceding years (y-1 + y-2) might exert an effect on fish catch in year y. There is no significant autocorrelation between catch and FI in any year and years precedings it. A series of regressions were therefore carried out weighting the hydrological indices according to the formula (αHI_y-1+ [1-αHI_y-2). The resulting correlation coefficients and standard errors of estimate are shown in Table XI.

TABLE XI

Correlation coefficients and standard errors of estimate for various weighting of HI_y-1 and HI_y-2 for three African floodplain systems

The increase in correlation coefficient and decrease in the standard error of estimate in all three cases indicates that variations in fish catch in year y are better explained by a combination of flood history from the preceding years than by either year on its own. Best fit linear regression lines which are plotted in Fig.14 are:

Kafue: C_y = 2 962 + 70.5442(0.7HI_y-1 + 0.3HI_y-2) accounting for 57 percent of the variation

Shire: C_y = 5 857 + 38.1135(0.9HI_y-1 + 0.1HI_y-2) accounting for 82 percent of the variation, and

Niger: C_y = 3 239 + 32.1039(0.5HI_y-1+ 0.5HI_y-2) accounting for 92 percent of the variation.

Figure 14 Best fit linear regression lines for the relationship between catch and flood regime for
A. The Kafue flats fishery C_y = 2962+70.5442(0.7HI_y-1+0.3HI_y-2)
B. The Shire River fishery C_y = 5857+38.1135(0.9HI_y-1+0.1HI_y-2)
C. The Central Delta of the Niger C_y = 3239+32.1039(0.5HI_y-1+0.5HI_y-2)

The above results, while still being rather imprecise, indicate clearly that there is a relationship between the hydrological regime and catch although much further data on other systems and over longer time periods is needed to define the relationships more fully. The main interest here centres on whether the factor responsible for the fluctuation is in fact the intensity of the flooding or the severity of the drawdown period (or some complex of both). The similarity of the correlation obtained in the Kafue is hardly surprising as intensity of flooding (HI 1) is inversely correlated with severity of drawdown (HI2) (r = -0.78). The correlation between these factors in the Shire is much lower (r = - 0.45) which may account for the lowered correlation of catch with HI2 (log catch y = 4.3 + 0.21 log HI2_y-1: r = -0.54 or Catch y = 100.37 - 4.47 HI2_y-1: r= - 0.43). A further influencial factor may be the area of water remaining in the system relative to that at peak floods. In the Kafue this is about 34 percent whereas, in the Elephant and Ndinde marshes, it is 68 percent. It may be supposed that the more stringent the drawdown, as reflected by lessened percentage of water remaining in the system during the dry season, the greater the influence on catch of fluctuation in this factor.

Several attempts have been made to predict catches in future years from existing information on the flood regime using regression formulae similar to those described above. To check the validity of this procedure a series of analyses were carried out on each batch of data by calculating successive regressions and using them to predict the catch in the following year. It is assumed that if one can validly predict the catch of fish in the coming year y+1 from this year's flood data (HI_y and HI_y-1) then the accuracy of prediction would improve as an increasing number of sets of data is added to the regression. It is further assumed that to be of any use for the purpose for which such estimates may be intended - regulation of numbers of fishermen, anticipating supplies of fish to the markets, etc. - estimated catches should be consistantly better than ± 10 percent of the actual catch. In fact results presented in Table XII indicate that there is no improvement as additional data is added and that estimates are rarely better than 25 percent. In addition there is a slight tendency to estimate high as the series increases. From which it is concluded that at our present state of knowledge it is not possible to predict catches in river systems in coming years from regression analyses of the past performance of the fishery with the degree of accuracy required. Predictions may however be precise enough for assessment of the effects of dams, irrigation schemes, etc., where accuracies of ± 30 percent may suffice.

7.5 Summary

As may be expected from the biology of the fish (Section 5) the productivity and yield patterns of floodplain fisheries fluctuate considerably both during one year and between successive years. Populations of fish increase through the flood season reaching a maximum at peak floods. Populations then decline both numerically and by weight during the dry season reaching a minimum just prior to the next flood. Between year fluctuations in catch (and probably in population) can be correlated with differences in flood intensity in previous years, the most influential being that of the year preceding the current year. Similarly the degree of drawdown on the amount of water remaining in the system would appear to exert an effect. Thus, years of optimum catch follow one or more years of good flooding and where the amount of water remaining in the system during the dry season is relatively abundant.