Does environmental variability matter? Lake Malombe presents an interesting and challenging case as it presents an example, still rare, of an African fishery that has collapsed as a result of overexploitation of its stocks. However, judging from the way fishing effort developed, it is also clear that it may represent a special case: a relatively small shallow lake has been fished with highly efficient gears that require a high investment. This is quite unlike many other similar small and medium African fisheries, where fishing effort develops more on the line of more of the same gears for which relatively low investment is needed. Unlike in Malombe many of these fisheries often still have under-exploited stocks. The answer therefore is no: environmental variability hardly matters where and when a complete mismatch between the scales of natural variability in fish production and human exploitation is reached through a highly efficient fishery.
But questions remains: did increased fishing effort on its own cause the collapse? Or was the collapse amplified by other factors? For instance, did habitat destruction and/or of changes in water level cause changes in productivity? Habit destruction is a side effect of the fishery itself, but has important consequences for the interpretation of our results. Sadly, not much is known about these effects as no information is available on the scale and rate of habitat destruction during the period of the collapse of Oreochromis. We also do not have direct evidence on changes in productivity but assume that fluctuating water levels causes them. In a small lake like Malombe the effect of surface runoff on productivity will be considerable. With highly seasonal water levels, with peaks during the rainy season, we could safely disregard the distinction between river inflow and surface, were it not that large changes in forest cover around the lake - as has occurred - not only increase the volume of surface runoff but also its sediment load, possibly bringing about structural changes in the bottom habitats of the lake.
We will now return to the two questions posed at the start of this paper and summarize our findings
is there evidence for possible changes in productivity of Lake Malombe as a result of environmentally driven factors, such as changing water level?
is it possible to detect the combined effects of productivity changes and increased effort through the existing monitoring of catch and effort and of water levels in Malawi?
4.1 Trend perception: The governance dilemma and the search for informative indicators
The high variability of the monthly catch rate information is, to a large extent, administratively induced (Appendix 3). After removing temporal effects (annual variability, seasonal variation) the remaining variability is about as high as in lake Chilwa (see Figure 2, Zwieten and Njaya, 2003). One of the results is that, while the high seasonal fluctuation in water level and runoff is expected to be reflected in the catch rates in particular of those gears targeting small species and juveniles, this is not the case in lake Malombe. Despite this conclusion, it is possible to detect long and short-term trends of statistical significance in the various catch rate series within 1.5 to 2.5 years of monthly aggregated data. These time windows to detect a trend are not too bad, though it indicates that there is a limit to the speed with which effects of management measures could be detected. This could be called a governance dilemma (Densen, 2001): the intended result of measures will take time anyway - a lag of 2-3 years -to take place (Pet, Machiels and Densen, 1996), but the number of years to detect an effect and causally attribute it to the measure taken will increase with increased natural and administratively induced variability. As the effect of a measure could be significantly detectable only after a long period - at least five years to a decade or longer, the proof of its effectiveness will be difficult to obtain, even if the necessity of the measures would be beyond doubt as in the case of Lake Malombe. Furthermore, though certainty on long-term trends could be obtained relatively fast - at least those of the magnitude discussed here, we have seen that short-term trends vary tremendously, in particular for the smaller haplochromines, which will make it difficult to decide on the causes of short run trends.
It should be noted that the analysis on trends was done on data aggregations by species. As a rule of thumb, observed variability expressed as CV will increase by Ön, with n = the number of species in an aggregation, to obtain the average CV for the separate species, assuming lack of co-variation (Oostenbrugge et al., 2002). With that assumption, the variability on a species level will be much higher. However, co-variation could be the rule in environmentally driven systems. In such a case species or species groups of which large amounts of information can be easily obtained could act as indicators for the state of all stocks.
4.2 Fluctuating water levels, effort and habitat
The effect of changing water levels on stock levels is large as it can be detected despite the high background noise in the data. Depending on the gear-target species combination the effect is detectable within 1-2 years for the small meshed seines and nets, and within 3-4 years for the larger meshed gillnets. Despite this, combined effects of fluctuating water levels and changing effort were difficult to detect, both due to problems of confounding of trends and of technical interactions between gears, in particular between gillnets and Chambo seines. The effect of Chambo seines on Oreochromis, exarcebated by the fishing on the juvenile part of the stock by Kambuzi seines, and in the early stages of the fishery with Nkacha nets as well, completely overshadowed possible effects of changing water level on the catch rates in gillnets. The catch rates in Kambuzi seines did show a clear combined effect of changing effort and fluctuating water level. This fishery was sensitive to short-term trends as it was fishing on the juvenile part of the Oreochromis stock. Both decreasing water levels and fishing effort may have caused the disappearance of submerged vegetation in the lake. If this habitat were important as nursery grounds for Oreochromis, it would mean that the level of the recovery of this species is dependent on the extent of restoration of this habitat.
The shift from gillnets (used in open water) and Chambo seines (used along the shore and in submerged vegetation) to Kambuzi seines (shore-based) and Nkacha nets (open water) is a shift from large species or specimens of species to smaller ones. The disappearance of Oreochromis from Nkacha nets indicates that the juvenile species disappeared from the open-water part of the lake - Mwakiyongo and Weyl (2001) found no change in species composition in Nkacha seines over the past ten years. The stabilization of juvenile Oreochromis in Kambuzi seines show that the species-group is still there and that juveniles are in in-shore areas.
Over the period examined only a limited increase in fishing effort in terms of people fishing is observed. Apart from that, two of the four gears, decrease in numbers and activity over this period, while a third - Nkacha seines - did so since 1987. The number of Nkacha nets and its activity levels has increased since the start of the series, but at present is decreasing and over the past year many operations have left Lake Malombe for the South-East arm of Lake Malawi or move between the two (Weyl, pers.obs.; Banda et al., 2002). Thus, the present situation of overexploitation has been caused mainly by the type of gear and their way of operation (fishing patterns), rather than demographic increase: in other words Lake Malombe does not present a case of Malthusian overfishing (Pauly, 1994), but is a case of over-investment in a fishery. With that the mode of exploitation of fish in lake Malombe seems fundamentally different from the mode of exploitation observed in the other fisheries reported on in this volume.