5. THE EFFECTS OF FISHING AND ENVIRONMENTAL VARIATION ON THE REGENERATION OF FISH STOCKS

5.1 Introduction

Typical explanations for the decline or collapse of fish stocks are that the fishery puts too much pressure on the stocks or that the environment changes, changing the productivity of the stock^[17]. In the management of the African freshwater fisheries examined here, the traditional focus has been on fishing pressure with only limited considerations for environmental causes of changes in stocks. There is a long history of regulations against the allegedly detrimental effects of various fishing methods. Alerts about the effects of the ever-increasing fishing pressure are of a more recent date, and attempts to set sustainable levels of effort have been the goal of much research. Although the various attempts to regulate fishing effort have often proved to be fairly ineffective (Malasha, 2003) many African fisheries continue to thrive. Changes in catch composition and of target species have often taken place, but few stocks have collapsed or declined severely (e.g. Oreochromis spp. in Malombe; Labeo altivelis in Mweru; large Lates spp. in Lake Tanganyika), while in other cases, stocks previously thought to be endangered returned. For example, Oreochromis mweruensis, after very low stock levels in the 1970s is once again one of the most important components of the total catch of Lake Mweru despite an increase in effort of a factor of 5 (fishermen) to 10 (gillnets) over 30 years. The question then is why in African freshwaters do many fish stocks not collapse or yields do not decline, despite substantial increases in fishing pressure within the ranges observed in the past 50 years? How come some of these fisheries seem to be extremely resilient to increased effort while others are not and are certain fishing methods more harmful in some situations than in others?

This chapter will start by discussing the results and value of classical stock-assessments, where effort is the only variable, in the freshwater systems studied. Then, taking into account the often overlooked aspects of environmentally driven natural variability in fish stocks, we will examine the evidence on long-term effects of increased effort and discuss whether observed changes in fish stocks are a result of increased fishing pressure or are mainly environmentally driven. These observations lead to a holistic top-down approach with which the dynamics of effort development in African freshwaters can be related to both anthropogenic and environmentally driven changes in stocks. This approach entails addressing the impact of fishing on ecology and stock regeneration in three steps:

• System variability: the effects of environmental drivers on different scales of the fished ecosystems, indexed by changes in water level, can be used to classify systems based on fundamental ecological principles along a gradient of environmental stability from constant to pulsed systems.

• Susceptibility of fish species to fishing: the typical environmental variability each of these systems undergoes will be reflected in the biological responses of the fish species to it. Based on this, predictions can be made about the potential effects of fishing.

• Selectivity and scale of operation of fishing patterns: all fishing methods are selective and fishermen have various options to stabilize day-to-day catches. These options are related to system variability and the multispecies character of the fisheries studied. By discussing the scale of fishing operations relative to the ecosystem harvested, the development of fishing effort in different African lakes and its impact fish stock regeneration can be characterized.

5.2 Classical approaches

Rational exploitation of fish stocks involves the control of fishing mortality (effort and fishing methods) in such a way that annual catches of specific stocks can be continued indefinitely according to pre-determined objectives related to the productivity at different stock levels. The catch-effort curve of sustainable yields (Schaefer, 1954) exemplifies this approach: at any level of fishing effort up to the level where the ‘surplus yield’ is maximized, a yield can be found that is theoretically sustainable and stable (Appendix section1). Which level of fishing effort is chosen depends on a number of strategic objectives (Salz, 1986) such as securing a minimum biomass, maximizing food production (MSY), maximize the resource rent (maximum economic yield, MEY) or employment. Of these objectives, the concept of maximum sustainable yield (MSY) at which effort levels should be set in order to maximize food production has gained most prominence. Various models estimating MSY, or maximum yield per recruit, have been used extensively in African freshwater fisheries, and the concept of MSY has formed part of the research goals in many fisheries development projects as well (Kolding, 1994).

It is important to reiterate the biological assumptions of the surplus-production model most often used in the African context, the Schaefer model, because of the many policy implications that have arisen from these assumptions:

• Density-dependent logistic growth under environmentally stable conditions, which means constant carrying capacity (see Appendix 1).

• It is a single species model, with man the only predator.

• In the traditional application, it presupposes a continuous steady state, which means instantaneous equilibrium between catches and surplus regeneration.

These biological assumptions have been questioned for a long time in an extensive literature (e.g. Larkin, 1977; Sissenwine, 1978; Kolding, 1994), and are questioned again by the various case studies in this report. For now we will just note that natural variability in fish stock levels due to environmental variation results in changing values of the underlying biological parameters of the model: the intrinsic growth rate (r) and the carrying capacity (K) (see Appendix section 1). This will result in considerable uncertainty around the estimated sustainable yield curve, as constant conditions are not present and surplus production is not regulated by effort alone. Systems that experience less environmentally induced interannual variability may conform better to the underlying steady-state assumptions of the model, possibly resulting in more informative results from stock assessments. However, in systems with a large environmental variability, attempts to relate trends in fish stock levels to fishing alone may be more difficult. Such trends will be hidden in environmental variation or, in statistical terms, noise or error, resulting in what is often called ‘process error’ (Caddy and Mahon, 1995), that is variability caused by the unknown states of nature. Moreover, the multispecies and multi-gear situations encountered in most tropical small-scale fisheries, in particular in freshwaters, make the application of standard models even more problematic.

In all the lakes of our case studies, except Lake Mweru, previous attempts have been made to estimate MSY in order to set the level of effort at which the fishery would operate at maximum efficiency, or other sustainability levels (Mwakiyongo and Weyl, 2001) sometimes by gear type. Both Malombe and Kariba are highly instructive examples with regard to the limited informational value that such assessments have (Boxes 5.1 and 5.2). In Malombe, MSY from surplus production models has been estimated for the Chambo (Oreochromis spp.) and Kambuzi (haplochromine cichlids). The application of models was successful in the sense that they conveyed the message of limits to the system overshot severely. However, additional information was needed to explain the particular time sequence of catch and effort levels, as their behaviour did not conform to standard model predictions.

Table 5.1 The relative effect of hydrological changes and effort on catch rates in three systems studied The statistical regression model used is: Annual mean catch rate_ijk = overall mean + effort_i + lag(hydrological variable_j) + effort_i x lag(hydrological variable_j) + residual variation_ijk. Only significant effects are retained in the model and shown here as positive (+) or negative (-) effects. In some instances the hydrological variables and effort were confounded meaning that both parameters were significant alone, but in the total model one or the other became non-significant depending on the order they were entered into the model. In such situations it is not possible to quantify the relative effect of both parameters simultaneously. N = number of years. See text and Zwieten and Njaya 2002, 2002a and Kolding et al., 2003 for further explanation of the model.

Box 5.1 illustrates how the systems have evolved under increasing fishing pressure, and that, where stocks have collapsed, MSY in fact was found empirically. In Lake Kariba, the various attempts of estimating MSY and corresponding effort levels have all been unsuccessful in retrospect, partly because of model limitations and partly because of data limitations (Kolding 1994, Box 5.2). With hindsight, it can be maintained that several misconceptions and mistakes have been made in the past when evaluating and explaining the observed variations in the yields from Lake Kariba. These errors, however, are partly understandable when working under the conventional ‘stability’ paradigm, and using the established yield models.

Both cases represent situations where a good biological information base is present. In other cases (e.g. Lake Chilwa) some of the monitored information, such as fishing effort, was considered so unreliable that attempts to fit simple assessment models were regarded as completely unsatisfactory (Tweddle, 1995). Clearly, under such circumstances, and additionally in fluctuating systems, the very concept of ‘MSY’ as a management objective becomes highly questionable (Kolding, 1994, 1995). Still, the MSY concept remains deeply rooted in planners, administrators and research personnel, but so far it has mainly created misunderstandings or false expectations. Attempts to determine MSY in environments with a large, but to some extent predictable variability in productivity levels have failed. On the contrary, it can be generally said that yield (or Yield Per Recruit), models have not been of much use in any of the freshwater fisheries in Africa (Coulter, pers. com.).

How large the effect of environmental variability on stock levels in fact is can be seen in Table 5.1. In this table we have compiled the results of the relative effect of hydrological changes and fishing effort on catch rates in three studied systems. In all cases the analysis was performed in two steps (see Zwieten and Njaya, 2003; Zwieten et al., 2003a and Kolding et al., 2003 for further explanation). It takes time for any species to ‘grow’ into a gear before it can be caught (see below). This means that an effect of higher productivity on stock levels will only be observed after some time - usually between zero and four years depending on the size of the species caught in the fishery. The effect of hydrology on such ‘recruitment variation’ is the lag phase between annual hydrological variables and catch rates. If there is no time lag this means that effects of hydrological changes is felt by the fishery within the same year. In the second step the statistical model (stepwise multiple regression) is performed according to the lag phase found. The relative effect of effort or water level on catch rates is examined through the proportion (%) of the total variation explained by each independent variable in the model. It is clear from Table 5.1 that in nearly all cases hydrological changes had a highly significant positive effect on annual mean catch rates, meaning that catch rates decrease or increase according to similar changes in water levels. Increased fishing effort has in many cases the expected negative effect on catch rates. However, the table makes clear that such an effect only explains part of the total variation in catch rates, and in any case this part will often be difficult to distinguish directly from the effects of the environment on stock levels.

5.3 System variability: water level as environmental driver

In the long history of research on fisheries it has been debated whether fishery collapses are due to overfishing or due to environmental change (e.g. Corten, 2001). Large changes in productivity of fish resources may be due to environmentally driven processes, in particular where large changes of nutrient input occur. The most conspicuous external drivers of nutrient inputs in all of the systems in this study are long-term, interannual and seasonal fluctuations in water level and river inflow. Biological productivity, from algae to fish, depends essentially on the nutrients introduced by the annual flood regime, including the nutrient mobilization through flooding of lake margins or associated floodplains and swamps. Such systems are also called allotrophic riverine systems (Kolding, 1994). The water budget, monitored through water levels, can thus be used as a proxy for abiotically driven changes in productivity (Kolding, 1992; Karenge and Kolding, 1995a). Most lakes have rivers flowing into them, or are at least affected by rainfall runoff from land. It thus depends on the ratio of the inflow to the size (volume) of a system - the flushing time - how much a system is externally driven by water and nutrient inflow. For example, the average flushing time of Kariba is 2.6 years and of Mweru approximately 3 to 4 years. In contrast the flushing times of the large lakes Tanganyika, Malawi, Victoria and Turkana are respectively 7 000, 750, 140 and 12.5 years (see Table 5.2). In these large systems (except Turkana) biological productivity is predominantly determined by an internal supply of nutrients through vertical mixing and nutrients from the atmosphere. External drivers in the very large systems refer to wind stress (e.g. Spigel and Coulter, 1996; Plisnier et al., 1999; Sarvala et al., 1999) and seasonal changes in solar radiation (Talling and Lemoalle, 1998) that affect mixing processes. The much smaller lakes of our study are fully mixed, at least for most of the year, while the large lakes (except Turkana) are all permanently or seasonally stratified (that is with different layers of water that do not mix).

The magnitude of changes in productivity, in other words the relative stability, thus depends on the size of the lake relative to the inflow rate. This in its turn is a function of the catchment area, which determines how much a system depends on local patterns of rainfall. But stability also depends on the temporal scale (periodicity) of the fluctuations relative to the life history characteristics of fish species and communities affected by them. Long-term climatic changes over large areas are reflected in long-term fluctuations in lake-levels (Nicholson, 1996; Nicholson and Yin, 1998). Such long-term changes may be associated with large-scale abiotic and biotic changes, and determine the range over which a system has oscillated historically and could be expected to do so in the future. Knowledge about longer-term changes that took place in the past, even the recent past (100 years or less) is important. It gives indications about the long-term dynamics of a system that is not human-induced and it can give clues about the persistence of trends and states over longer time spans. In statistical terms, this part of the total variance in time series of catches and catch rates would be called ‘red noise’ in analogy with the low periodic cycles in spectral analysis.

However, in a fisheries management context the important scale is the interannual variability over much shorter timescales, in general from one year to a decade. Interannual variability gives information about the relative stability (and hence predictability) of the productivity of a lake and with that the persistence of stock levels of longer lived species in particular. This is the timescale over which year class variability of longer-lived species as a result of interannual variability becomes important. In a tropical environment, the period over which cohorts of longer-lived species are under exploitation is about three to four years, which will be seen in time series of catch rates as medium-term fluctuations or ‘blue noise’. Lastly, intra-annual variability, or seasonality in water levels gives the extent of the variability in nutrient pulses that are generated each year. Stock levels of short-lived species such as the pelagic clupeids of Mweru and Kariba (but also those of Lake Tanganyika (Zwieten et al., 2002c)) and the small barbs in Lake Chilwa (Furse et al., 1979) are particularly driven by bottom-up processes resulting from seasonal variability in nutrient pulses^[20].

Thus, the effects of lake level fluctuations on productivity will be operating on a range of timescales. The biomass of fish species or a of community could be expected to co-vary over similar scales. Long-term historical fluctuations in water level due to climatic changes can give a measure of relative stability over that time. Over shorter time spans, externally driven conditions may still be highly variable, which will characterize changes in the fish communities and fisheries operating on them. We thus need a measure of variability of a particular water body over different timescales. In addition this measure should include the scale of the system, as it will be clear that a similar range of lake level fluctuations in large systems will have a smaller or a much more localized effect, compared to small systems. Therefore we have devised an index of Relative Lake Level Fluctuations (RLLF), defined as:

By calculating this index both for the average interannual (change in annual mean = RLLF-a) water level and the average seasonal pulse (RLLF-s), we have an index both for the average inter-annual stability of a system and the average strength of the seasonal pulse with which we can scale different systems. With each system, important measures would then be: (1) the persistence of conditions on a decade scale, and (2) the extent and duration of extreme high or low levels. Persistence is important to predict short-term trends while peak levels are expected to be important to predict year class strength. Here, we will only discuss these by visual examination of the time series of water levels. A third measure of importance would be the change in surface area of the lake as a result of changing water levels. We do not have such information, but it would be of considerable importance for lakes with low sloping shorelines as for instance in large parts of Kariba, for the smaller lakes, but also for local conditions in larger lakes (e.g. Fergusson’s Gulf in Lake Turkana (Kolding, 1993a)). These measures can be used to construct empirical relations with time series of catches or catch rates.

Historical to long-term fluctuations in lake-levels could indicate large-scale ecosystem changes and resets. Large fluctuations in lake-levels have taken place over long periods of time in the Great Lakes of Africa (Table 5.2 and Figure 5.3). These lakes act as integrators over long-term changes in rainfall patterns, and in a feedback loop partly influence rainfall patterns as well (Nicholson and Yin, 2001). Historical long-term fluctuations of the large lakes are remarkably correlated over the past 200 years^[21]. Smaller lakes follow the same pattern of climatic change and some of these may even dry up completely during extended periods of low rainfall, e.g. Lakes Chilwa, Malombe, Mweru Wa’Ntipa, Nakuru and Stephanie. Such droughts mean complete resets for the ecosystems of these lakes. Species diversity is related to the variability and the size of the system (Figure 5.4). For somewhat larger systems like Mweru and Kariba, climatic changes could also mean long-term changes in relative species composition, as can already be observed on a shorter timescale in large shallow pulsed lakes, for instance Lake Chad (see citations in Lévêque, 1995 or Lévêque, 1997). However, time series of catch rates of Mweru and Kariba are too short to be able to confirm such species successions. Those found in Lake Kariba appear to be influenced by lake-level changes (Kolding and Songore, in prep). Still, in this case succession may be largely a function of the age of the system and not of historical periodicity in changes in inflow of the Zambezi River.

Long-term changes would result in long-term fluctuations of stock levels (‘red noise’). Time series of catch rates or of other indices of fish stock size are generally too short - even in temperate zones often not longer than 60 to 80 years - to show the presence of red noise. On the other hand, if over large areas and systems, similar patterns exist, then the possible existence of long-term variation can be pointed out. For instance, in the later half of the previous century, in particular during the 1970s, all lakes and river systems throughout East and Central Africa had exceptionally high water levels (Nicholson, 1999; Laraque et al., 2001), in contrast to West Africa and the Sahel (e.g. Le Barbe and Lebel, 1997). Though productivity changes in the Great Lakes will, as discussed below, largely reflect changes in internal dynamics, the rise in water levels did reflect a change in climatic conditions. It coincided for instance with a remarkable rise in mean catch rates and in a more pronounced seasonality of clupeids in Lake Tanganyika (Zwieten, 2002c). Catches of the whole SADC area, dominated by the output of the Great Lakes, show a fluctuation around the main trend, that is significantly positively correlated with lake levels of Tanganyika and Malawi with a time lag of six to eight years (Figure 5.3). Such a lag is too long to be explained merely by fluctuations in stock levels, as these would have to be visible within one to four years, more so now that small pelagic species are dominating catches in most of these lakes. However, this lagged persistence in catch levels related to lake-levels could indicate a time lag in relative effort levels as well. Good catch conditions could attract fishermen over a period longer than these conditions exist and vice versa possibly resulting in fluctuations in total output. Such results highlight the need for good information on trends and fluctuations in numbers of people fishing to be able to understand their significance. If our inference is correct, it could mean that catch rates are in themselves a regulating factor of population-driven levels of effort.

Interannual lake levels indicate medium-term trends and persistence of conditions directly affecting the variation and persistence of fish stocks. In this shorter time perspective, local rainfall conditions in the catchment area dominate (Nicholson, 1998b) and interannual changes in water levels in Lakes Mweru, Bangweulu, Kariba and Chilwa reflect these. In the lakes of our study, the range of annual mean levels over the period examined is between 0.8 and 3.7 m, but mean change in water level (D WL) and variability (CV = Coefficient of Variation = standard deviation*100/mean) in water level is very different. Lake Kariba, Lake Mweru and Lake Chilwa have the largest mean D WL (1.30, 0.53 and 0.58 m) and CV (135%, 93% and 94%) of all the lakes examined. But the effect on the system differs considerably as is reflected in the RLLF-a: 4.3% (Kariba), 7.2% (Mweru) and 17.8% (Chilwa). Malombe and Bangweulu have the lowest mean D WL (respectively, 0.32 and 0.26) but Bangweulu is a much more variable system (CV=78% compared to Malombe CV= 56%). Lake Malombe can be considered as a satellite lake of Lake Malawi and largely follows its changes in water levels. Compared to other lakes in this study, it is much more stable in terms of mean annual change, variability and RLLF-a.^[22]

The next element to consider is the level of persistence of environmental conditions. Examining short-term trends in lake-levels is one way to do this. A cursory examination of the time series of relative water levels and river inflow in Figure 5.3 indicates that all of the lakes examined exhibit upward and downward trends and stable periods generally not longer than five to six years, followed by peaks in water levels. As noted already, persistence is reflected in catch rates in different ways for different species depending on their longevity (Zwieten and Njaya, 2003; Zwieten et al., 2003a paragraph 5.5) and the peaks are important for the year-class strength of longer-lived species. These short-term trends in environmental behaviour of the system will be reflected in short-term trends in catch rates (‘blue noise’).

Table 5.2 Lake size and lake level variability ordered according to increasing RLLF-a. (See text for further explanation.)

Based on Kalk et al., 1979; Crul, 1993; Evans (1978); FAO, 1993; Kolding, 1989, 1994; Bos, 1995; Crul, 1995a, 1995b; Nicholson, 1996; Nicholson, 1998a, 1998b; Nicholson and Yin, 1998; Nicholson, 1999; Laraque et al., 2001; Nicholson and Yin, 2001; van den Bossche and Bernascek, 1990; Verheust and Johnson, 1998; Department of Hydrology, Zambia.

Figure 5.3 Relative water levels of lakes Tanganyika, Victoria and Malawi and the five lakes of this study. In blue are deviations from the long-term mean of annual mean levels over the period for which data were available. Green bars are the deviations of the 89 year mean annual total inflow of the Zambezi at Victoria falls (mcm = million cubic meter)(Langenhove et al., 1998). Arrows indicate the years Lake Chilwa was reported to be dry. Malombe (orange) is hydrologically considered a satellite of Malawi: when this lake has low water levels Malombe completely dries up. The relative levels of Malombe and Malawi is shifted to be able to show their respective lake level developments. Lake levels are based on Crul, 1995a (Victoria and Tanganyika), Nicholson, 1998b (Chilwa), the Department of Hydrology in Malawi (Chilwa and Malombe), the Department of Hydrology of Zambia and own measurements (Bos, 1995 and van Zwieten (unpublished) (Mweru) and Kolding, 1994, Songore, 2001 (Kariba)). The top panel shows the variability around the trend of the total catches of the SADC region (see also Figure 2.1), with grey bars comparing periods of low catches relative to the trend.

Figure 5.4 Number of fish species in ten SADC lakes related to the seasonal Relative Lake Level Fluctuations (RLLF-s).

Lastly, intra-annual lake levels represent seasonality in productivity. The actual (variation in) size of nutrient pulses is conditioned by the seasonal variability in water inflow and levels. Variation in size and duration of floods is important in regulating the fish productivity in a cascade of effects through various trophic levels in an ecosystem (‘wave of productivity’ - see Pope et al., 1994), where interactions between species may dampen such effects (see below). Mean seasonal change in lake levels is highest in Lake Mweru (2.1 m) with considerable variability (CV = 38%), indicating that the size of the seasonal pulse varies considerably between years. By comparison, Lake Malombe exhibits a lower seasonal variability (CV = 23%), again indicating that also on this temporal scale the lake is more stable than other lakes in our study (Kariba CV = 49%; Bangweulu CV = 35%; Chilwa CV = 36%). Taking into account the scale of the lake, the effect is highest in Chilwa (RLLF-s = 39.7%), followed by Bangweulu (34.3%), Mweru (25.7%), Malombe (20.4%) and Kariba (10.1%) (see Table 5.2).

The potential impact of changing water levels is thus relative to the scale and size of the system:

· From long-term trends and fluctuations and complete resetting of systems acting on the composition of whole fish communities (‘red noise’),

· via short-term trends and interannual fluctuations operating on fish species within communities on a short term depending on lifespan and response of species to changes in externally driven lake productivity (‘blue noise’),

· to variations in seasonal pulses acting on the recruitment of species with short lifespans, triggering fish migrations and, being the starting point of the annual productivity pulse, the year-class strength of longer-lived fish.

By examining different scales of variability, we can thus position a system within a general classification of lakes in terms of system stability. Within the class of systems externally driven by water inflow, or pulsed systems, we may also have an important general indicator for changes in stocks, easily measurable, available and easily communicable. The results of our study on the effect of water levels on catch rates (where both time series are available) are summarized in Table 5.1, which shows that trends and variability in annual mean or change in water level is dominant in explaining catch rates in all of these. It is also clear that effects of changes in water level are observed in catch rates of individual species for periods of between one to four years. In the next paragraph we consider how such variability affects individual species and fish communities on different timescales in order to explain these results. It is important to know to what degree fish yield is determined by abiotic factors (such as environmentally driven productivity changes) and by biotic factors such as predation, competition and life history characteristics of individual species.

5.4 Susceptibility of fish stocks and species to fishing under environmental variation

We will now draw up a framework to help explain why fish stocks in pulsed systems appear to be more resilient under increasing fishing effort compared to those of more constant systems. Fish stocks, like all natural populations, fluctuate in abundance with or without exploitation. A full understanding of the causes of the variations is still lacking (Rothschild, 1986), but Regier (1977) developed a conceptual framework for classifying the dynamics of fish stocks according to a number of biological and ecological research traditions. The framework defined four broad classes based on two variables along which a resource could be positioned:

• size of the resource (small to large) and

• temporal and/or spatial variability of a system (low to high).

Regier’s four classes were:

• Large stocks generating large amounts of revenue in relatively stable environments. The traditional stock assessment models have largely been developed for this type of problems. However, it has long been recognized that the usefulness of advice resulting from this approach is reduced with systems that are changing under growing or highly variable stress, or where numerous resources within an ecosystem are utilized simultaneously, as in multispecies, multi-gear situations. Here this is illustrated for Malombe and Kariba (Boxes 5.1 and 5.2).

• One or a few large fluctuating resources - examples are clupeids and floodplain fisheries.

• Relatively small but numerous resources in stable environments - lakes, large reservoirs, near shore marine environments, or reefs.

• Systems of small interacting resources that fluctuate erratically - small lakes, ponds, lakes under highly variable abiotic stresses (anthropogenic or otherwise).

Assessment approaches related to the last three classes range from monitoring and stepwise forecasting, to the development of simple indicators addressing whole-ecosystem variables and models based on a community-level approach. In terms of research and regulations the cost of single-species management would become excessive for small fluctuating resources. Instead, management should concentrate on real-time decisions for the immediate future with expected outcomes based on continuous evaluation of time series of yields, catch rates or other indicators and possible explanatory variables (effort, water levels) of the fluctuating system in a comparative and stepwise management approach.

The complexity of the fisheries and the ecology and dynamics of the small interacting resources, exacerbated by the abiotically driven intra- and interannual variability, and the general small economic value of the stocks thus precludes use of models based on steady-state assumptions. However, even disregarding environmental variation, a number of generalizations can be made on the responses of multispecies fish assemblages to fishing that for all practical purposes may be used for predictions on reaction to stress. For instance, a systematic sequence of changes in a species composition as a response to increased stress has been observed on the North American Great Lakes (e.g. Rapport et al., 1985 and Regier and Henderson, 1972) and termed the ‘fishing down process’. Since this early work, much has been learnt from a diverse array of aquatic ecosystems. Welcomme (1999) presents a synthesis based on inland fisheries, and contributes to a framework for the development of indicators for management and conservation purposes based on the use of explanatory variables in a largely statistical context.

We will take these approaches as a starting point to address the differential impact of fishing mortality on fish communities that are subjected to different conditions of environmental stress. Based on the gradient between pulsed and constant systems emerging from our discussion of environmental drivers, we can use Regiers’s dimension of temporal and spatial variability to typify the fish communities present in those systems in this dichotomy (based on Lowe-McConnel, 1987; Kolding, 1993 and expanded by Coulter, unpublished). This does not mean that the scale or extent of the systems is not important: we will return to that when discussing the dynamics of fishing effort later in this chapter. After discussing in some detail the character of fish assemblages in constant or variable systems, and the generalized reaction of fish communities to stress (see also the expanded theoretical considerations in the Appendix), we will discuss how a multispecies fishery may interact with these communities. We will then develop a conceptual model in which all of this information can be summarized in the context of the gradient between variable and constant systems, and which could be used to identify variables that can be used as indicators of change as a result of increased fishing mortality.

African freshwater systems differ in their temporal variability (Table 5.3) and can be arranged in two broad groupings: pulsed and constant systems. This arrangement of course represents extremes in a continuous gradient, and various systems will occupy different positions along this dimension. Furthermore, larger freshwater systems may have subsystems with either characteristic. For instance, Lake Tanganyika and other African Great Lakes have a pelagic ecosystem that is highly variable, while various other parts of the lake system have constant characteristics (Zwieten 2002c; Allison et al., 1995). Variable systems encompass a large range of environments, many of which are associated with large African rivers: floodplains, swamps and allotrophic riverine lakes. Lake Mweru and Lake Kariba fall into the latter category. Furthermore, the class of variable systems includes endorheic lakes (i.e. lakes that have no river outlet) - e.g. Lake Chilwa and Turkana - or those that are parts of larger constant systems. Lake Malombe, for example, is grouped here as a constant system because it is a satellite of Lake Malawi’s South-East arm, with which it has many characteristics in common. Littoral sections of the Great Lakes are considered constant, though the littoral is seasonally most affected by runoff from land. FAO (1993) asserts that the productivity of Lake Malombe is probably highly affected by runoff, while, as we have seen, in terms of water-level fluctuations, the system is much more constant than most of the other small systems discussed.

Table 5.3 Examples of (seasonally) pulsed and constant African freshwater systems, with a comparison of typical fish community attributes and the implications for fishery in these systems. P/B ratio = Production/Biomass; F/Z ratio = Fishing mortality/Total mortality, where Total mortality (Z) = natural mortality (M) + fishing mortality (F). M2/Z ratio is Predation mortality/Total mortality. The arrow indicates that the system characteristics are a gradient over the dichotomy. ++ = high, rapid, many, long; --- = low, slow, few, short.

Table based on (Lowe-McConnell, 1987; Kolding, 1993 and Coulter, unpublished).

The typical environmental variability experienced by a system is reflected in the biological responses of fish species to it. The various attributes of fish species presented in Table 5.3 represent generalized life history characteristics found in a community typical for each class of environments. This resource characteristic is more or less determined by the ‘ruling pattern of mortality’ (Appendix section 2). In other words, the probability of dying from either abiotic or biotic causes determines the typical assemblage of fish species in a system. Under environmentally stable conditions, biotic causes such as competition and predation will prevail, and organisms will tend to develop mechanisms to escape mortality from other organisms (grow big, parental care, specialize), leading to the attributes listed under constant systems. If, however, the chances of dying are more unpredictable in terms of abiotic disturbances (floods, droughts, fire, etc.) then mortality will hit all size groups equally. In this case it could be a disadvantage to be a big long-lived fish and an advantage to be a small short-lived fish with attributes for a rapid recolonizing as listed under pulsed systems^[23]. The basic question of whether fish yield is determined mainly by abiotic or biotic factors in these systems can now tentatively be answered, at least on a generalized system level. Fishery implications, that is the effect on the interaction between resource character and fishery, can be generalized over the two classes of systems^[24] in five dichotomies (Table 5.3). Where the proportion of fishing mortality is low compared to the predation mortality from the community, and the biological turnover (P/B ratio, see Appendix section 3) is high, growing fishing effort is less likely to lead to overfishing or stock collapse. Species and communities having these characteristics have adapted to a high natural mortality, and possess a high regenerative capacity, so the susceptibility to increased fishing mortality is low. As fishing mortality can grow without much impact and the natural turnover is fast, the yield potential of these species is high. However, interannual variability in the landings of a fishery will also be high in these instances. Examples are: Limnothrissa (Kapenta) in Kariba, Microthrissa (Chisense) in Mweru and Barbus (Matemba) in Chilwa.

However, regardless of environmental variation, all fish communities contain species or segments of populations that have a differential vulnerability to exploitation. We distinguish three categories of susceptibility to fishing. Large groups of fish species and segments of fish communities can be fitted into these three groups based on rather diverse considerations of ecological and behavioural specialization and size characteristics (see Table 5.4). Under intensifying fishing pressure, which initially typically targets the large individuals in a community, the composition of a catch shifts (Regier, 1973; Hoggarth et al., 1999a, 1999b; Welcomme, 1999). In general, in the successive development of a fishery, the more susceptible components of a fish community will give way to the more resilient species. Susceptibility here is considered as an intrinsic characteristic of a species or ecological group, by which we mean the relative ease with which they can be subjected to high exploitation rates.

The spawning concentrations of species migrating up-river are very susceptible to fishing pressure. Examples of rapid declines in stocks of such species abound. In our case studies, the example comes from Lake Mweru where the large cyprinid Labeo altivelis, the mpumbu, disappeared quickly from the fishery when demand for ‘caviar’ was high in the 1950s and spawning migrations were heavily hit (Kimpe, 1964). Spawning runs of the cyprinids of Lake Tana, Ethiopia are presently endangered through fishing (Nagelkerke et al., 1995; M. de Graaf, pers.com.). Older age classes of all species generally disappear fast from a fishery in the early stages of its development.

Table 5.4 Resource character and susceptibility to fishing of various components of a fish community.

Based on Coulter, unpublished.

Large specimens of Lates sp. disappeared quickly from the catch both in Lake Turkana (Kolding, 1995) and in Lake Tanganyika (Coulter, 1991). Large tigerfish (Hydrocynus) and catfishes (Clariidae; Bagridae) also became rare in the catch in Mweru (Kimpe, 1964; Zwieten et al., 2003b) and Kariba (Kolding et al., 2003a).

The species of the intermediate category are more resilient to intensified fishing pressure. They have a wide distribution in rivers and lakes, and in many cases furnish the bulk of the yield of African freshwater fisheries together with the fisheries on small pelagic species. They are ecologically flexible, are relatively unspecialized in food preferences and behavioural patterns and are able to withstand high natural mortalities while living under highly variable abiotic conditions, such as in the swamps of Lakes Chilwa, Bangweulu and Mweru. The reproductive capacity of many of these species is high. Although many Tilapia-like species have rather specialized spawning behaviour and exhibit parental care (mouthbrooders, nest builders) they are very productive and can withstand highly varying ecological conditions^[25] through adaptations such as stunting^[26]. Narrow trophic specialization is generally not present: many species have a varied diet and are unspecialized omnivores, detrivores, herbivores or opportunistic piscivores (Bowen, 1988).

The third category is species that are highly resilient to fishing pressure. These are the small pelagic species that have become important in the African fisheries in the past three or four decades. Table 5.4 lists these species, that all have high P/B ratios (see Appendix section 3) and high natural mortalities, which together with the ‘down fishing’ of their predators, enables them to withstand high exploitation pressures.

Total yields in multispecies fisheries are, in general, surprisingly stable over a wide range of fishing effort^[27], though such stability may often be obscured by interannual variation in environmentally driven systems. This overall stability can be explained by the effect of aggregation of catches over numerous species, which reduces variability. In a recent discussion on the stabilizing effect of bio-diversity on the resource outcome, this phenomenon has been called the ‘portfolio-effect’, in reference to the stabilizing effect of a more diverse investment strategy on the stock market (Tilman et al., 1997; Doak et al., 1998; Tilman et al., 1998; Lehman and Tilman, 2000; Densen 2001). However, as different segments of a fish-community have a differential susceptibility to fishing and other stress, it has long been recognized that fish communities tend to react in broadly predictable ways to increased stress (Regier and Henderson, 1972; Regier, 1973 Rapport et al., 1985). The succession in exploitation of more susceptible to more resilient components of a fish assemblage under increased fishing pressure is a well known phenomenon, and has been documented for a range of freshwater and marine systems (e.g. Regier, 1973 and references in Welcomme, 1999; Zwieten et al., 2003b). An expectation that can be derived from these generalized observations is that the size of species caught in a multispecies fishery will reduce over time, while the species composition of the catch will change to components of an assemblage that can withstand higher mortalities. This process has also been linked to the trophic level of each of the components of a fishery, and an annual mean trophic level of the total catch of a fishery can be calculated. A downward trend in the mean trophic level indicates that a fishery is moving down the food web as new small-sized resources are increasingly being exploited, while the large-sized species are reduced in abundance.

Not only biological characteristics intrinsic to a species determine susceptibility to fishing pressure. Other factors are changing environmental conditions and the susceptibility of species to specific gears: in other words the accessibility of stocks to the gears in use. Changing environmental conditions may change susceptibility to fishing, by changing the accessibility to gears. Drying pools in floodplains where fish are concentrated and easily fished out are an obvious illustration of this effect. Conversely, high water levels in floodplains and swamps will provide more space and habitat for shelter, and catch rates will decline because of the resulting reduced accessibility of fish to gears. In Lake Chilwa, under lowered water conditions, all species become more susceptible to fishing without a change in fishing methods (Zwieten and Njaya, 2003). Similarly, in Lake Mweru water draw down after the seasonal peak in May results in extremely high catch rates of Oreochromis mweruensis in the shallow lake area immediately adjacent to the swamps, while catches of this species are reduced under seasonal flood conditions.

Figure 5.5 The trophic levels in a community at which a fishery intervenes. Triangles represent trophic pyramids of animal communities with predators at the apex and animals feeding on primary production and detritus at the bottom. The width of the triangle at any level represents the relative biomass of that level. Black lines represent selective exploitation, arrows the direction of increased pressure. The three triangles could each represent a different fishery, for example: a sport fishery on tigerfish (“Hunting lions”), a gillnet fishery on tilapiine fishes such as the Oreochromis fishery in Mweru comparable to grazers in wildlife (“Hunting antelopes”) and a fully developed fishery in which all trophic levels are harvested proportionally to their biomass (“Hunting all that moves”).

The accessibility of a stock to a gear is related to catchability and selectivity (see Chapter 2). Considering that all fishing gears are selective means that each gear affects only a certain portion of a fish community, that is the fishable stock. Basically there are two types of fisheries: those that employ only one gear, usually gillnets, and those in which a high diversity of gears are employed. As a consequence, the intervention of a fishery will either be selective to certain components of a fish assemblage or will to a greater or lesser extent target all components present. Sport fisheries on predatory fish such as Hydrocynus vittatus in Lake Kariba, commercial fisheries on Lates niloticus in Lake Victoria or on large Lates spp. in the early stage of the pelagic purse seine fishery of Lake Tanganyika (see Box 5.3) are examples of the first type of fishery. A second example of this type, but with a completely different impact, is the early stages of gillnet fisheries of Lake Mweru or Malombe, when larger mesh sizes were used that mainly (though not exclusively) targeted large Oreochromis species. The difference between the two examples is that the fishery intervenes on entirely different trophic levels within an ecosystem - predators vs. detrivores/herbivores (Figure 5.5). Lake Mweru also gives the example of a developing fishery diversifying into targeting successively more components of the fish community. Mesh size in stationary gillnets decreases, while more and more active methods are employed, such as seining (open water and beach). Lastly, fish attraction through lights on the pelagic species Microthrissa (zooplanktivore) or through Fish Aggregating Devices (FAD’s) on Alestes macrophthalmus (facultative piscivore/insectivore) develops.

The result is a maturing small-scale and subsistence fishery, in which many components of the assemblage are targeted, without any of them disappearing from the fishery, though some species reduce to low stock levels. Thus, the conceived negative image of “fishing down the food webs” is not necessarily ecologically bad, but just represents a fishing pattern illustrated in Figure 5.5 where the fishery develops to exploit all trophic levels - hunting everything in proportion to the natural P/B ratios. In principle, such a fishing pattern could be much more harmonious for the overall natural mortality pattern than a selective fishery with gear and mesh size regulations (Kolding, 1994; Misund et al., in press). Floodplain fisheries in the large Asian rivers, for example the Mekong (Coates, 2001) are examples of highly complex fisheries in which a large number of species are targeted by an enormous diversity of gears, without evidence for any of these species disappearing from the system. The Bangweulu case study (Kolding et al., 2003b) is a similar example of how the different gear types (many of which are illegal) are targeting different parts of the fish community.

To understand the complex simultaneous effects of changing environment, fisheries impacts and trophic interactions (predation and competition) on the biomass and structure of a fish community, we will use a conceptual model that is derived from an approach used to analyse properties of whole ecosystems based on the distribution of biomass over the size of organisms (Sheldon et al., 1972). By including information on externally driven productivity on the one hand, and size-selective stress on the other hand, changes in the shape of the biomass-size distribution may indicate on a ‘whole’ system level what size and extent of effects these have on this particular group or assemblage.

As there is generally a larger biomass of small fish (small species and juveniles of large species) compared to large fish, the overall shape of a biomass-size distribution is a descending curve over size (Figure 5.6). Variations in the shape of this curve, the slope and the intercept with the vertical (biomass) axis indicate systematic changes in the size structure of communities that can be related to the mortality pattern (see Appendix 1) and variables such as mean lake depth, lake productivity, water turnover rate and human impacts (Cyr et al., 1997). Variations in the intercept with the vertical axis indicate that average overall biomass in the system is a function of the availability of nutrients to the system, rather than a function of the structure of the particular biological community (Boudreau and Dickie, 1992).

Changes in the slope indicate changes in the size structure of the group or assemblage. The approaches derived from these insights deserve special attention in tropical fisheries, as they could provide promising ways to address the problems related to the assessment of multispecies resources. The size structure of a multispecies catch can provide a global interpretation of a fishery (Gobert, 1994) and relations between size structure and impact parameters can be used at a local scale to describe the size structure of freshwater communities. Though much work still needs to be done to turn this tool into a method of formal assessment of a fish community, it can already be used as a way of visualizing a large amount of complex information (Zwieten et al., 2003b). In this way, the information contained in the length structure of catches, nowadays extensively sampled in many small-scale fisheries, can be used to their full potential.

Size information tells us something about the species susceptibility to fishing. Due to the selectivity of most fishing gears (usually regulated in terms of the prohibition of fishing on small individuals), it can in general be said that highly susceptible large fishes are on the right - and those not so susceptible are on the left in a biomass-size distribution. Figure 5.6 visualizes the impact of the magnitude of the seasonal pulse on various parts of a fish community, which has been described as “riding the wave of annual productivity as it rolls through the extended size spectrum from phytoplankton to big fish” (Pope et al., 1994). The impact of this ‘wave of productivity’ is dampened along the size spectrum by individual life history traits of species and size groups (growth and mortality rates, that is productivity, generally decreases with increased size) and species interactions at higher levels such as predation and competition. In constant systems, the seasonal energy pulse, which acts on the smaller sizes of the fish community first, is much smaller. Together with the long-term stability of these environments, this allows for the development of life history strategies that result in narrow niche specificity (stenotopy, territoriality, trophic specialization). These adaptations towards efficiency rather than proliferation make susceptibility of fishing for these smaller species also higher. Variability in standing biomass of smaller-sized individuals is higher than larger-sized individuals, but again larger in pulsed systems compared to constant systems. Energy distribution, regulating the abundance through interactions of competition and predation, is larger. Lastly, Figure 5.6 indicates that the range in size of fish in constant systems is generally larger. In summary, the alternative life history traits of species as a function of different selective forces in the pulsed or constant environments, results in different biomass size-distributions. The biological attributes and implications for exploitation are the same as listed in Table 5.3.

Figure 5.6 Biomass-size distributions, variability, and energy pathways in A) constant and B) pulsed systems. Triangles characterize fish communities by biomass and size while variability (vertical arrows) and energy flow (horizontal arrows) indicate dominant mortality patterns. Biomass decreases with fish size and predatory fish (blue triangles) are generally larger than their prey. Energy flow through pathways starts from seasonal input of nutrients resulting in seasonal changes in productivity (red arrows). Energy in a fish community is partitioned through competition and predation (green and blue arrows). Variation in biomass caused by changes in energy input is larger with smaller-sized fish (black arrows). Increasing fishing pressure generally results in decrease in biomass of large fish and increased catches of smaller fish.

The impact of increased fishing effort is, in general, described as decreasing from large to small specimen of fish. However, as indicated in the discussion on the selectivity of fisheries and the preference of certain target species, it will be clear that the situation may be more complex. For instance, in Lake Mweru the decreased mesh size of the dominant fishery on the lake, the gillnet fishery, did reduce the size of fish in the catch but also targeted different sections of the fish community, and thus released pressure on the population of larger Oreochromis (Zwieten et al., 2003b). In the following section we will discuss the dynamics in fishing effort in a context of scale of the fishery, the level of externally driven variability of the system and its consequences for the fish communities.

5.5 Selectivity and scale of operation of fishing patterns: dynamics of fishing effort

Previous chapters emphasize and explain the small-scale character of fisheries in SADC freshwaters. “Small-scale” fisheries generally operate within geographically limited areas and intrinsically depend on local resources. This is in contrast to technologically and organizationally more developed fisheries with a broader spectrum of options in terms of fishing grounds, markets and alternative investment opportunities (Thomson, 1980; Panayotou, 1982; Bâcle and Cecil, 1989; Misund et al., in press). Since 1950, African freshwater fisheries have grown much more rapidly in the small-scale sector than in the technologically more advanced sectors, which have not thrived very well (see Chapter 3). How problematic is this growth in technologically less advanced but numerically abundant effort from a biological point of view, in particular in pulsed systems? In all our case studies, increased effort, with annual rates of increase of numbers of fishermen between 2% (Mweru) to 6% (Chilwa) had a significant effect on stocks (Table 5.1). However, only in one case, Malombe, has increased effort led to a collapse of an important stock. Increased effort in Mweru and Chilwa was mainly population-driven, but in Malombe growth of effort was concluded to be investment-driven. Does this mean that a population-driven increased effort, operated with a high variety of gears on a small-scale is biologically less harmful than increased effort as a result of accelerated investments in more effective technology in these pulsed systems?

Firstly, observed effort in terms of density of fishers is not uniformly distributed across the different systems. The average allocation of effort as number of fishermen per unit area in the various African freshwater systems is roughly correlated with the magnitude of the seasonal pulse into the systems as quantified by the RLLF-s (Figure 5.6A). In turn this could be a function of the overall high productivity of these systems (Figure 5.6B). Though catch rates and effort are not independent, this could indicate that the more pulsed systems in general support a higher effort than the more constant systems. There is little evidence from our case studies that this high effort has had a negative impact on the regenerative capacity of most of the stocks. This result supports the theoretical fishery implications for pulsed systems outlined in Table 5.2.

Secondly, to understand the dynamics of fishing effort within each system - i.e. the resultant effect of the choices made by all fishermen - it is important to look at the scale on which a gear is operated relative to the variability in abundance of the resource it targets. It is the scale of operation that makes a small-scale fishery distinct from technologically more advanced fisheries and determines the investment level and amount of labour needed (Misund et al., in press). The scale at which a fisherman can operate his gear determines how the variability of the resource will be reflected in his day-to-day catch. One could also say, it is the way a fisherman uses his gear to aggregate fish that determines how his daily catches are stabilized, as well as the effect it has on the stocks.

Figure 5.7 The relationship between annual relative lake level fluctuations (RLLF) and effort density (A) or catch rates (B) in 11 African freshwater systems. Lake Itezhi-tezhi is not included in the regression lines. Data for the number of fishers and landings are given in Table 2.2, data for RLLF are given in Table 5.1 and Kivu = 0.06, Turkana= 1.97.

If a fisherman wants to stabilize his income from fishing through stabilizing his daily catch he has three options (or combinations of options):

• Option 1: invest in more, larger or more efficient gear, in other words aggregate fish in space,

• Option 2: use the gear more intensively (extra labour), or aggregate fish in time,

• Option 3: diversify to other species and be less dependent on one resource, in other words aggregate fish in the fish community through catching a diverse assortment of fish.

Lastly, he also can stabilize income by doing something else besides fishing (Aarnink, 1997; Oostenbrugge et al., in prep.; Zwieten et al., 2003b).

Thus a high diversity of species in a catch will also reduce overall variability, even if the catch of each individual species will be low or highly variable. Most important regarding the effects of the choice of these different options in effort allocation, is that the variability in fish stocks, both in total biomass and in relative abundance of the component species, is determined, as argued before, by longer-term (interannual, decadal) changes in annual flood pulses. In other words, both the fishing pattern and the resource variability will result in a more or less variable outcome. Both are important variables for a characterization of the dynamics of effort in a fishery (Appendix section 5).

Small-scale fishing gear can be classified in three development categories with increasing scale of operation and more or less with increasing level of investment (Misund et al., in press):

Traditional fishing implements are locally produced and are operated from small canoes or rafts and by wading or diving fishers. Simple hook and line and handline, spears, harpoons, tongs, pots, traps, dip-nets, gillnets, small beach seines, weirs and barriers are included in this group.

Intermediate fishing implements are nets (gillnets, seines and lift-nets) and much of the hook-and-line gear that are factory produced and made of synthetic materials.

Modern fishing implements are trawls and other mechanized active gears, including the mechanized haulers that increase the spatial coverage of a fishery. Modern accessories to fishing such as hydro-acoustic equipment (echo-sounders), and global positioning systems increase the efficiency of the fishing operation by increasing the spatial window over which the fish can be searched and found or that will give accurate site information.

Most African fisheries contain a mix of the first two groups of fishing implements, the first being associated more with riverine and floodplain fisheries, while the second are dominant in lakes. Modern fishing implements are rarely found in African freshwater fisheries. Increase in efficiency is generally reached by an increase in the scale of operation through more intensive use of gear - for instance through active methods such as seining or ‘beat’ fishing - and not by auxiliary technology. These two development categories with their lower capital investments and reduced maintenance costs that are used with simpler fishing methods, make small-scale fishers more adaptable to the exploitation of a multispecies stock that show large changes in stock levels of different species. In other words, in African freshwaters small-scale fishermen often opt to stabilize catches by increasing the number of species caught or switching between them by using a high variety of capture methods.

This adaptability could be seen as utilizing multispecies stock in a more efficient and biologically more sound way, compared to technologically more advanced or larger scale fisheries that are less adaptable to changing circumstances. Some of the following characteristics typical of the fishing patterns of small-scale freshwater fisheries, are indicative of adaptability:

Effort (overall fishing pressure) is unevenly distributed among different systems. Pulsed systems in general support a higher fishing intensity than constant systems suggesting that the productivity of the system to a certain degree is regulating the effort despite the overall population-driven growth (Figure 5.7)

Allocation of effort is unevenly distributed in time and space within a system. For example in Mweru, most fishers are concentrated in the swamps, lagoons and the highly productive southern part of the lake and effort is to a large extent a function of the physical environment of the coast, access to land and to markets. Fishing is hardly ever the sole source of income (see Chapter 3). Effort allocation is seasonal during periods of receding water levels and associated longitudinal or lateral migrations of fish, while weather conditions highly affect spatial and temporal fishing patterns on the pelagic fishery (Zwieten et al., 2003b). In Kariba, most fishing effort is in the inshore, in Zimbabwe on the eastern part of the lake around the main population centres. Lastly, migrational shifts in effort take place both within a fishery and in and out of a fishery (see Chapter 3 and Zwieten and Njaya, 2003; Zwieten et al., 2003b; Overå, 2003).

Fishing patterns are highly diverse and often labour-intensive: numerous people are active with an extraordinary variety of specialized capture solutions to different resources, environments and seasons resulting in labour-intensive multi-gear/multispecies exploitation patterns. This is not typical for African freshwater fisheries, but has been described for many situations in tropical freshwater and coastal systems (von Brandt, 1959, 1984; Nédélec, 1975; Nédélec and Prado, 1990; Coates, 2002; Christensen, 1993; Hoggarth and Utomo, 1994; Hoggarth et al., 1999a, b; Zwieten, 2002d; Medley et al., 1993; Misund et al., in press). River and floodplain fisheries often have limited numbers and sizes of intermediate fishing implements per fisherman. For example, in Mweru swamp fishermen have on average one gillnet compared those in the lake that have between 10-20 gillnets of the same size each (Zwieten et al., 1995). Gillnets and other intermediate fishing methods are used in most lakes and reservoirs for inshore and demersal species, often as both active and passive fishing techniques. The offshore and pelagic regions of these fisheries have become more important over the past three decades, where small pelagic species are caught in more specialized nets often in combination with light attraction.

These complicated fishing patterns and highly diverse fishing methods operated on a small scale, in general have a small daily output - a bucketful - for each individual fisherman, and each capture device is intrinsically associated with specific selectivity. In other words, the impact of fishing is distributed over large numbers of fishing units and all methods have a differential impact on the ecosystem. Much attention and many legal regulations are focused on the supposedly detrimental effects of the so-called less selective gears such as seines, small mesh sizes in gillnets, drive- or beat fishing, barriers and weirs. However, in practice little is known about the actual impact of these methods. In the few instances where the actual impact of these gears has been studied, it is an open question how “detrimental” they actually are (Kolding et al., 1996; Chanda, 1998; Bennett, 1993; Clark et al., 1994 a, b; Lamberth et al., 1995 a, b, c). Small-meshed beach-seines on small barbs and juvenile bream have been used for a long time in Chilwa without much apparent negative effect (Zwieten and Njaya, 2003), though by contrast the large beach-seines used in Malombe were highly detrimental to the stocks of Oreochromis. Furthermore, as many gears can be utilized in variety of ways and while the lifespan of most fishing gears is generally low, fisheries can readily react to changing circumstances and seasons. For example, in Lake Mweru gillnets are used in at least eight different fishing patterns: as stationary gears, in various types of open water and shoreline seining, drift-netting and in combination with other gears such as knobbed sticks for ‘beat fishing’ and Fish Aggregating Devices, all targeting different subsets of species (Zwieten et al. 2002b). Traps and gillnets normally do not last longer than two to four years and shifts in mesh sizes rapidly take place over a whole fishery.

As a result of these characteristic effort dynamics, multi-gear/multispecies fisheries, in particular in highly variable ecosystems such as floodplains, could be producing an overall species abundance and size composition that closely matches the size and species structure of a fish community. In principle, a fishery that harvests all species at all trophic levels at rates proportional to their natural mortality pattern during their lifespan will be non-selective on an ecosystem level. Non-selective overall harvesting patterns would conserve the fish community (see Appendix section 4) as the relative structure of the ecosystem would be maintained, only each component would be smaller (see Figure 5.5 “Hunting all that moves”). For instance, the Mekong floodplain fisheries seem to have persisted unchanged (although with high temporal fluctuations) with very high numbers of people fishing with a high variety of methods for as long as our observations can tell. Even the susceptible migratory “white fish” and predatory species still form important parts of the catch, while giant catfish still are present. Where biodiversity has been adversely affected, it is generally where ecosystem integrity was not maintained (Coates, 2002). In other Asian river systems, such as the Bramaputhra and the Ganges, migratory fish have disappeared and fishing is now completely dependent on small individuals, indicating that eventually adverse effects may occur (Hoggarth et al., 1999a,b).

Figure 5.8 Generalized development of fishing yield and catch rate of a fishery with increasing effort. The broken lines represent increasing variation around the mean yield and catch rates over time (vertical arrows). The six characteristics in the table below the plot change with increased effort as indicated (high/low). They refer to total fish biomass of a system (a, b, c and f) and to the outcome of the fishery (d, e and f).

However, this does not mean that increasing effort is without problems. Increased effort, population-driven or otherwise, invariably leads to decreasing catches per fishing unit (CpUE) (Figure 5.8). Some fish species, found mainly but not exclusively in constant systems, will be more susceptible to fishing pressure (Tables 5.3, 5.4). As an overall harvesting pattern of a fish community where the fishing mortality is strictly proportional to the natural mortality of its constituent species probably does not exist, selection will take place. Increased reliance on the smaller species or the smaller, juvenile, segments of a fish community leads to increased variability in the catch of individual species^[28]. This is caused both by the natural mortality patterns of the smaller species as they follow the boom and bust periods in variable systems, and as a result of a decreasing overall biomass. The time series of catch rates of the industrial fishery of Lake Tanganyika, described in Box 5.3 shows this, while this is also an example of option (1) of the choices for catch stabilization through spatial aggregation.

In Malombe, fishing with large beach seines on Oreochromis spp. resulted in average monthly catch rates that, though highly variable^[29], showed no trend over the years. No effect of decreasing water levels was observed. In the same period, catch rates for Oreochromis in gill-nets exhibited fast declining trends and were significantly affected by changes in annual lake-levels, with expected time lags as fish “grow” into the gear. The fact that this effect was not seen at all in the large beach seines can only be explained if a change in the scale of operation - through a change in gear size, intensity of gear use or spatial allocation (options 1 and 2) - took place (Zwieten et al., 2003a). In Malombe, stabilization of catches with these large seines has been obtained through increased spatial allocation of highly efficient labour-intensive gear. In other words, in this case increased scale of operation overrode the natural variability induced by flood pulses. As flood pulses have a much more limited effect in this system compared to the other systems researched, stocks of both Oreochromis and the small haplochromine complex had limited opportunity to bounce back within the period of high fishing pressure.

The larger Oreochromis had little chance to occasionally produce large year classes. In Malombe thus a mismatch exists between the natural variability in fish production and the scale of exploitation.

Lastly, an example of Option 3: in Lake Mweru the effect of decreased catch rates and increased variability has resulted in an increasing diversification of fishing methods and target species, as previously unexploited stocks are now targeted. The shift to smaller mesh sizes has resulted in a higher diversity in the catch. In this case, increasing the number of species in the catch reduces variability in individual catches. Each of these species may have highly different stock sizes, since the opportunities to reproduce and the abiotically induced mortality patterns differ both within and between years and are different between species. Timing, duration and extent of flooding are important factors in this dynamic. For instance, large variation in water levels has resulted in strong year-classes of Oreochromis, which could partially escape the gillnet fishery in inaccessible areas during flooding. The real danger for these species would be if the occasional larger flood pulses did not take place over a long period of time. As with lake Chilwa, where flood pulses are reflected in the catch in the same year or one year later, increased effort in Mweru will mean that boom-bust periods will be seen in the catch of individual fishermen with much shorter time lags than previously when larger fish were targeted.

From this analysis we can conclude that:

Artisanal small-scale fisheries such as in Lake Mweru, by hedging the inherent variability in relative abundance of multispecies stocks, and opting to target many species simultaneously, are developing an overall fishing pattern that could in principle conserve the ecosystem. On the other hand, where operations override the inherent variability by increasing scale and maintaining catch rates at the same level, this will lead to problems.

In other words, the multi-gear (overall unselective) fishing pattern employed in many small-scale fisheries existing in pulsed systems, combined with the ability of fishermen to rapidly change their target, could be seen as an ecologically optimal fishing pattern over a very large range of fishing pressures. In these cases, the conflicts arising from the ‘management’ implementation of existing gear restrictive regulations (based on single-species models and considerations), are in many situations, at least from an ecosystem perspective, largely futile.

In situations where multi-gear operations exist with limited investment per operation many people in principle will be able to start fishing. At several points in our analysis so far we have seen indications that effort levels are dependent both on productivity and changes in productivity. Increased effort will invariably mean that catch rates will go down, while at the same time the inherent variability of pulsed systems can produce boom-and-bust situations. In that case, catch rates could be a regulating factor for fishing effort by itself if other options to produce an income still exist. This is in contrast to the idea of Malthusian overfishing that predicts that declining catch rates pose a trap from which fishermen cannot escape.

5.6 Information on catches, effort and environment in African freshwaters

As both the Malombe and Kariba cases indicate, generating the classical models and obtaining basic population parameters such as growth and mortality have been useful in interpreting biological developments and status of particular fisheries. However, taken as a scientific point of reference on which management measures could be taken, which for a long time was considered the ideal in many African countries, the model outcomes only gave a misguided sense of controllability. The inadequacy of the models to encompass the effect of the fishing pattern on the whole community may also create disillusion if enforcement of measures taken is unsuccessful. For instance, in 1999 all fishing methods used daily in Lake Malombe were in fact forbidden (pers. com. Dr. Mapira, Director of Fisheries in Malawi)! Uncertainty about the effectiveness of the measures taken and about causal relationships only makes matters more complex.

To overcome the problem of information for fishery management in the face of these complexities, a pragmatic approach should be adopted in attempting to translate uncertainties into the decision-making process and management practice. The classical models do help in thinking about complex data sets, and in generating intuitions about a system. However, the next step, which is often forgotten, is how they could help in defining information needs. The results of models, however crude, are just one piece of the evidence in the translation of uncertain information into knowledge relevant to management (Zwieten et al., 2002d). Observations outside the model invariably prove to be highly relevant to the issues at hand. Thus, learning about general system behaviour through evaluation of long-term information will aid in generating the necessary intuitions on what may or may not work. The important questions to be answered are: what is the gross abiotic and biotic behaviour of a system considered, what information can be obtained directly from evaluating time series of catch and effort data and other long-term data, to assess and predict changes in fish stocks in multispecies assemblages in response to fishing or other stress?

We have examined the evidence of long-term biological effects of increases in fishing effort, population- or investment-driven. Observed changes in fish stocks in the freshwater systems in question could be related to increased fishing pressure in all cases. But, size, extent and duration of flood pulses that act as environmental drivers are also invariably extremely important and cause many of these pulsed systems to be more or less resilient to increased effort. Growth in effort is discussed in the context of the scale of fishing operation relative to the variability of the ecosystem harvested, and the choices a fisherman could make to stabilize his catch. It has been shown that numerous small-scale operations using a variety of gears could conserve the fish community, and are probably less detrimental than increased effort as a result of increasing the individual scale of operation. For highly pulsed systems, the natural boom-and-bust will to a large extent regulate effort by itself, or as Beverton (1990) puts it “[some] fisheries cannot be driven to extinction because the fishermen will disappear before the fish”. And even if they are intermittently ‘overexploited’ they have the capacity to regenerate fast when conditions are favourable. For the more stable systems, a higher potential danger of over-exploitation exists since they need more time to recover. For these systems, it may be appropriate to monitor the catch rates to learn their limits empirically and possibly intervene with effort-regulating measures. Most important, however, in the transition from traditional single-species classical models to a more comprehensive community approach taking more than ‘effort’ into consideration, is the need to learn more about individual systems’ response to exploitation. From this perspective, the continued collection of catch, effort and environmental parameters is paramount.

The case studies on the five lakes aimed to address the possibility of detecting changes in fish stocks under increased effort and changing environmental conditions using the existing time series obtained through Catch and Effort Data Recording Systems (CEDRS), or experimental gillnet survey data, combined with time series on lake levels. It appears that much of the information present in time series of catch and effort monitoring and experimental surveys is severely under-utilized. Data obtained through the CEDRS’s are in many situations often used solely to calculate total yields in statistical reports. Combined with effort data, attempts are sometimes made at calculating sustainable catch levels through formal stock-assessments, but with limited success. The information used in such models is often considered too unreliable. However, we have shown that information that could be derived from the time series of catch and effort is not limited to these applications. An analysis of trends and variability in catches, catch rates, fishing effort and water levels can produce empirical relationships, which in the proposed pragmatic approach can be used to predict not necessarily when something will happen, which is the aim of stock-assessments, for example, but what could happen if something changes (Kolding, 1994). This can only be achieved through continuous and mandatory evaluation of time-series information, also in relation to environmental changes, within the administrative structures responsible for management. Such evaluation will lead to knowledge of what can be perceived, while on the basis of this, expectations on the effectiveness of measures can be formulated. With an explicit analysis of uncertainties - also as a function of data-collection procedures - such trends can be qualified (Zwieten and Njaya, 2003; Zwieten et al., 2003a), and tell how fast changes occur and how fast the effect of management measures directed to stock protection will become visible (Densen, 2001). Examination of long-term experimental surveys can be used to comparatively examine changes in the fish communities, possibly in relation to changes in fishery or the environment (Karenge and Kolding, 1995a, b). From such results, important indicators can be derived which will aid in management decisions. In the formulation of information needs, there is thus a need for the ‘long view’ - the longer the time series of catch, catch rates, effort and general system indices such as water levels, the more can be learned about the behaviour of a fishery and its effect on the regeneration of stocks (Bjorstad and Grenfell, 2001; Yoccoz et al., 2001).