Lake Victoria had a fish fauna dominated by 500+ species of haplochromine cichlids that made up more than 80% of the fish mass. The five main trophic groups caught with bottom trawlers in the sub-littoral areas of the Mwanza Gulf were: detritivores, zooplanktivores, insectivores, molluscivores and piscivores. The detritivores (13+ species) formed the most important guild, making up 60–80% of the number of individuals, followed by the zooplanktivores (12+ species), which comprised 10–30%. In the 1980s the haplochromines from the sub-littoral and offshore areas (estimated at some 200 species) vanished almost completely. Commercial trawl fishery, the upsurge of the introduced Nile perch, and an increase of eutrophication were potential causes of this decline. In the 1990s, when Nile perch was heavily fished, a recovery of some haplochromine species was observed. We studied the decline and partial recovery of the different haplochromine trophic groups in the northern part of the Mwanza Gulf. The rate at which the trophic groups declined differed; the relatively large piscivores, insectivores and molluscivores were the first to disappear from the catches. The small detritivores and zooplanktivores declined at lower rates, especially the latter group. From the beginning of the 1990s a resurgence of both groups was observed. By 2001, the zooplanktivores had reached their previous level of abundance, but their diversity declined from more than 12 species to only three. Though four detritivorous species began being regularly caught again, they constituted only about 15% of the number of individuals, while the zooplanktivores made up more than 80%. The patterns of decline and recovery indicate that, though fishery played a role locally, predation by Nile perch and eutrophication were the main factors determining the fate of the haplochromines. However, it has so far been impossible to establish the causal relationship between the two, and the relative impact of each of these phenomena separately. The potential effects of the changed trophic dominance, and the importance of the haplochromines for the ecosystem and a sustainable fishery, are discussed.
The first lake-wide fish survey in Lake Victoria (Graham, 1929) revealed a dominance of haplochromine cichlids. In his report, Graham wrote: “The number of individuals is almost incredible. … So great are their numbers that I have contemplated suggesting trawling for them, in order that they may be used for manure … .”. He also mentioned that it had been suggested to him frequently, that a large predator should be introduced that could convert these small fish into a commercially more interesting fish. However, as a warning he wrote: “The introduction of a large predatory species from another area would be attended with the utmost danger, unless preceded by extensive research into the probable effects of the operation.” Despite this warning, Nile perch (Lates niloticus L.) was introduced into the lake, after the catches of other major food fishes had declined. The first introduction, that for a long time remained unknown, was made in 1954 (Amaras, 1986). Later, after extensive debates, new introductions were made in 1962 and 1963 (Fryer, 1960; Anderson, 1961; Gee, 1964).
A second lake-wide survey from January 1969 through May 1970 revealed that haplochromine cichlids made up at least 80% of the demersal fish mass, and Nile perch less than 0.5% (Kudhongania and Cordone, 1974a). At that time, Greenwood (1974) estimated the number of species in the lake at about 150 to 170. According to him, these species were exploiting “every major food source in the lake except for zooplankton.” He suggested that insects, especially larval and pupal chironomids and Ephemeroptera, were the most important food organisms.
Based on these lake-wide surveys and on experimental trawling in the Mwanza Gulf from 1973 onwards, a commercial trawl fishery for a fish meal factory in Mwanza was started in 1976. The trawlers caught 10 to 15 tonnes of haplochromines per day, mainly just outside, or in the entrance of the Mwanza Gulf.
In 1977, the Haplochromis Ecology Survey Team (HEST) of Leiden University, the Netherlands in co-operation with the Tanzania Fisheries Research Institute (TAFIRI), started to study the cichlid communities in the Mwanza Gulf and its vicinity (Figure 1). Many new species were discovered, including a large group of zooplanktivores, and by 1985 the total number of cichlid species in the lake was estimated at more than 300 (Witte et al., 1992b). In the sub-littoral and offshore waters of the Mwanza Gulf, species feeding on detritus, together with species feeding on zooplankton, dominated the haplochromine ichthyomass (Witte, 1981; Goldschmidt et al., 1993). Surveys made in the Kenyan and Ugandan parts of the lake (Figure 1) between 1989 and 1992, revealed approximately 30 new species (Kaufman and Ochumba, 1993). Finally, sampling along rocky shores and islands in the south-eastern parts of Lake Victoria, in the first half of the 1990s, produced some 100 previously unknown species of strictly rock dwelling cichlids (Seehausen, 1996; Seehausen et al., 1998). Currently, it is estimated that more than 500 cichlid species formerly lived in Lake Victoria.
In the 1980s, Nile perch suddenly boomed in Lake Victoria and, concomitantly, the haplochromines in the sub-littoral and offshore areas vanished almost completely (Barel et al., 1985, 1991; Ogutu-Ohwayo, 1990; Witte et al., 1992b). At about the same time, it was noticed that eutrophication in the lake strongly increased. This resulted in algae blooms (Ochumba and Kibaara, 1989; Hecky, 1993; Mugidde, 1993), decreased levels of dissolved oxygen (Kaufman, 1992; Hecky et al., 1994; Wanink et al., 2001) and decreased water transparency (Seehausen et al., 1997a; Witte et al., 1999). About a decade later, after a decline in Nile perch due to heavy fishing, a slow resurgence of some haplochromine species was observed (Witte et al., 1995, 2000; Seehausen et al., 1997b; Balirwa et al., 2003). However, the majority of species did not recover.
The Nile perch upsurge, the dramatic decline of the haplochromine cichlids, and the other ecological changes in Lake Victoria triggered many studies and debates about the causes of these changes (e.g. Barel et al., 1985, 1991; Balon and Bruton, 1986; Ribbink, 1987; Acere, 1988; Harrison et al., 1989; Ogutu-Ohwayo, 1990; Kaufman, 1992; Kudhongania et al., 1992; Witte et al., 1992a, b, 1995; Hecky et al., 1994; Bundy and Pitcher, 1995; Seehausen et al., 1997a, b, 2003; Verschuren et al., 1998, 2002). A major problem hampering these discussions is the lack of sufficiently detailed biotic and abiotic sampling, covering long time-series in the same area. In this paper we describe the decline and partial recovery of different trophic groups of haplochromine cichlids along a research transect across the Mwanza Gulf during the period 1979 to 2001. We discuss the possible causes of the differential decline and recovery of these trophic groups. This may help unravel what happened in the past. We also address the potential effects on the ecosystem of the resurgence of the haplochromines and their changed trophic composition. Understanding the causes and effects of the differential decline and recovery of the haplochromines is essential for proper ecosystem management.
Material and techniques
Densities and trophic compositions of haplochromine cichlids were studied in the sub-littoral waters of the Mwanza Gulf. Two types of bottom trawlers were used for sampling: (1) a small trawler powered by a 20 or 25 hp outboard engine (trawl headrope 4.6 m, cod-end mesh initially 15 mm and later 5 mm, see Witte, 1981); (2) large trawlers (80–250 hp, headropes 18–25 m, cod-end mesh 20 mm).
With the small trawler, 11 stations on a research transect across the Mwanza Gulf were monitored between 1979 and 2001. The research transect was approximately 5 km wide and ranged in depth from 2–14 m. Here, we only use the four most frequently sampled sub-littoral stations (G-J) with a depth range of 7–14 m (Figure 1). Between 1997 and 1999 we made only one trawl shot at the transect (station J). Therefore, we added data of a sample collected during the same period at the entrance of Nyegezi Bay (station N, 7–8 m; Figure 1). Trawl shots at each station lasted 10 minutes.
With the large trawlers, trawl shots of 30–45 minutes duration were made. Apart from the periods 1978/79 and 1984–1987, the trophic composition of haplochromines in the catches of the large trawlers was only analysed occasionally. Therefore, no complete data set from these trawlers is available over the study period. Moreover, MV Mdiria, which operated from 1973 until 1986, had no fixed sampling stations. It covered a large area of the gulf, ranging from the entrance up to 25 km southwards and mainly fished where catches were profitable. In contrast to Witte et al. (1992b, Figure 4), in this study only haplochromine catches were used from trawl shots that approximately covered the transect area. They comprise seven hauls that were made between March and May 1978 by MV Mdiria (headrope 25 m), 32 hauls made throughout 1987 by MV Kiboko (headrope 18 m), and one haul made in November 1997 by MV Lake Victoria Explorer (headrope 22.6 m). Nile perch densities in the period 1978 to 1983 are based on records of average catch rates by MV Mdiria, in the northern half of the gulf, and, in the period 1984 to 1990, on catches of MV Kiboko at the research transect. The catch rates of these trawlers were transformed to catches per hour with a 22.6 m headrope.
Of the catches of both the small and large trawlers, samples of 2–3 kg (400–600 fishes) were stored on ice and afterwards analyzed in the laboratory for species and trophic composition. Only adult or sub-adult fishes (generally > 4 cm SL) were used for calculating the trophic composition.
Data on diet, size range and distribution of the main trophic groups in the study area (Table 1) were obtained from literature (Greenwood, 1974, 1981; van Oijen, 1982, 1991; Goldschmidt et al., 1990, 1993; Witte and van Oijen, 1995; Goudswaard et al., 2004). For this study, we have combined the benthic detritivores and the pelagic phytoplanktivores into one group. The detritivores fed on bottom debris, but also included phytoplankton in their diet. The diet of the phytoplanktivores consisted mainly of cyanophyta (blue green algae) that were collected from the water column (Goldschmidt et al., 1993). These pelagic phytoplanktivores were rare in the bottom trawls (< 1% of the number of individuals; Goldschmidt et al., 1993). In this paper the combined group is referred to as detritivores. Paedophagous cichlids, feeding on haplochromine eggs and larvae, are included in the piscivores.
The trophic composition is based on numbers of fish in the catch. When considering biomass at the onset of the study period, the proportion of the insectivores, molluscivores and piscivores should be raised with a factor of 3.5 because of their larger size. If one wants to consider the total haplochromine biomass in the lake, further corrections should be made for the distribution of the haplochromines over the entire water column (Goldschmidt et al., 1993; Witte et al., 2005a). As data of fish weights and vertical distributions were not available over the entire period, comparisons were restricted to the numbers of individuals in bottom trawls, which still give a good impression of the relative densities of the major trophic groups.
Percentages per trophic group in Table 2 and Table 3 were calculated from the mean numbers of fish in the catches over periods of one to several years. These data did not meet the assumptions of analysis of variance, therefore the non-parametric Jonkheere-Terpstra test was used to test the significance of the observed trends. Differences between years were tested with Bonferroni corrected Mann-Whitney U tests (SPSS 11.5 for Windows).
Apart from differences in diet, in morphology and in distribution in the water column, the trophic groups also showed differences in size (Table 1; Witte and van Oijen, 1995). With a maximum size of up to 25 cm standard length (SL), the piscivores included the largest species. Insectivores and molluscivores had intermediate adult size ranges (6–18 cm SL), whereas detritivores, phytoplanktivores and zooplanktivores comprised only small species (5–9 cm SL).
At the end of the 1970s, the five main trophic groups at the four stations were: detritivores, zooplanktivores, insectivores, molluscivores and piscivores (Table 2, Table 3). The first two groups, comprising the smallest species, made up more than 90% of the numbers, and more than 80% of the weight, of the bottom dwelling cichlids (Table 2, 3; Goldschmidt et al., 1993). In bottom trawls the detritivores were at least twice as abundant as the zooplanktivores (Table 2, 3).
Decline and recovery
Since the 1970s, the catch composition in the sub-littoral area of the Mwanza Gulf changed dramatically. In 1978, haplochromines dominated the catches of the large bottom trawls (92% of the total catch weight) and Nile perch was almost absent (Table 2). In spite of intensive sampling, haplochromines were hardly present in the catches in 1987, and 97% of the catch weight consisted of Nile perch. Concomitantly, the mean total catch rate had decreased from approximately 1100 to 200 kg h− 1 (Table 2). In the sample taken in 1997, the total catch rate (199 kg h− 1) was similar to that in 1987. However, the contribution of Nile perch had decreased to 76%, while that of haplochromines had increased to 21%. Highest average catch rates of Nile perch in the study area were found in 1986 (Figure 2). In the following years (1987–1990) the catch rate of Nile perch in this area decreased significantly (Jonckheere-Terpstra test P = .000, N = 242).
In 1978, the detritivores made up about 60% of the number of the haplochromines in the catches of the large trawlers, and the zooplanktivores comprised 30% (Table 2; Witte et al., 1992b). Nine years later, in 1987, the contribution of detritivores had declined to 3.6% and that of the zooplanktivores had increased to 96.4% (based on 562 specimens caught that year). In the catch of 1997, the detritivores comprised only 15% of the number of haplochromines, whereas the zooplanktivorous species dominated the catch with 84%.
The survey with the small trawler confirmed the results of the large trawlers and revealed details of the decline and the partial recovery of the haplochromines (Figure 3; Table 3, 4). Between 1979/80 and 1981/82 the decline of the mean number of haplochromines was about 20%, but subsequently this decline amounted to 70% or more, every two years (Table 3). The insectivores, molluscivores and piscivores were the first to disappear from the catches. Among the piscivores, the decline of intermediate to large species at station G was significant over the period 1979/80 to 1983/84 (average numbers per 10 min trawling: 1.11 ± 1.13, 0.55 ± 0.82 and 0 in subsequent years; Jonckheere-Terpstra test P = .000, N = 48). Over the same period the small piscivores did not decline significantly (average numbers: 0.61 ± 1.24, 0.64 ± 0.67 and 0.12 ± 0.24; Jonckheere-Terpstra test, P = 0.153, N = 48). However, by 1985/86 they had also vanished. Both the numbers of detritivores and zooplanktivores decreased significantly at the research transect in the period 1979/80–1987/88 (Figure 3; Jonckheere-Terpstra test P = .000, N = 96, for each group), but only after 1985/86 did the zooplanktivores begin to decrease significantly between subsequent years (Table 4). Though at the onset of the study the detritivores were much more abundant than the zooplanktivores (Mann-Whitney U test, P = .000), their decline began earlier and was steeper. In 1983/84 the number of detritivores in the catches was already significantly lower than in 1979/80, and by 1987/88 no individuals of this trophic group were caught at stations G-J (Fig. 3; Table 4), although one was caught at nearby station E (Witte et al., 1992b). In 1987/88 the average catch rate of zooplanktivores was 0.7 ± 1.5 individuals per haul, significantly lower than in each of the previous years (Figure 3; Table 4).
Because the relative decline varied between trophic groups, the trophic composition of the cichlid community changed (Table 3). Until 1983/84, the contribution of the detritivores was more or less stable, whereas that of the zooplanktivores showed some increase. After1983/84 the fraction of the detritivores decreased strongly, leading to a reciprocal increase in the fraction of zooplanktivores.
From 1990/92 onwards, an increase in the number of zooplanktivorous and detritivorous haplochromines was observed (Figure 3; Table 3; Jonckheere-Terpstra test 1987/88–2001, P = .000, N = 60, for each of the two groups). The increase was slower than the decline in the previous years. For detritivores this increase can be located to between 1987/88 and 1993/95 (Table 4), whereas no further increase was observed after that time. For the zooplanktivores, significant increases were observed between 1987/88 and 1993/95, and again between 1993/95 and 2001 (Table 4). By 2001, the absolute density of the zooplanktivores was approximately as high as at the beginning of our study (Fig. 3; Table 4). Zooplanktivores were now more abundant than detritivores (Mann-Whitney U test, P = .018) and the catch rates of the other trophic groups, as well as the total catch rate of haplochromines, were much lower than in the 1970s (Fig. 3; Table 3, 4). As a consequence of these changes, the proportional contribution of the zooplanktivores to bottom trawl catches was much higher in 2001 than at the end of the 1970s (82% v 10% respectively).
In total, about 24 haplochromine species were caught in the period 1991–2001 in the sub-littoral areas of the Mwanza Gulf, where formerly more than 110 species were present (Witte et al., 2000). At the stations G-J the number of detritivorous species declined from 13+ to four, and that of the zooplanktivores from 12+ to three.
At the end of the 1970s, detritivores were the most abundant group in the catches of both the large and the small trawlers, while zooplanktivores were the second-most abundant. Even after corrections for differential vertical distribution in the column and for differences in mean individual weight among the trophic groups, the combined proportion of detritivores and phytoplanktivores in the 1970s was twice as high as that of the zooplanktivores (ca 50% vs 25% of the total haplochromine mass: Goldschmidt et al., 1993; Witte et al., 2005a).
In the 1990s, size differences between species of the recovering trophic groups were small. Therefore, the proportions of abundance by number or weight should be similar. The vertical distribution of the species was approximately the same as in the past (J.H. Wanink, University of Leiden, Leiden, the Netherlands, pers. obs.).
Causes of decline
During the past decades the causes of the decline of the haplochromines in Lake Victoria have been discussed frequently. Some authors mentioned over-fishing as the main cause (Acere, 1988; Harrison et al., 1989; Kudhongania et al., 1992; Bundy and Pitcher, 1995; Kudhongania and Chitamwebwa, 1995). Others attached the blame to the Nile perch (Barel et al., 1985, 1991; Ribbink, 1987; Ogutu-Ohwayo, 1990; Kaufman, 1992; Witte et al., 1992a,b, 1995; Goldschmidt, 1996). Eutrophication and the concomitant decrease in oxygen concentrations and water transparency were also suggested to be major causes (Hecky et al., 1994; Bundy and Pitcher, 1995; Seehausen et al., 1997a; Verschuren et al., 1998, 2002). In the following paragraphs we will discuss these potential causes in detail.
(i) Fishery. Since the start of the trawl fishery on haplochromine cichlids in the Mwanza Gulf in 1973, catch rates have shown an almost continuous decline (Witte et al., 1992b). In the first four years the average catches of haplochromines with a 90 mm cod-end mesh decreased from more than eight to less than one kg h− 1 (Kukowski, 1978; Witte and Goudswaard, 1985). With this cod-end, only haplochromines of more than 17 cm SL (viz. predominantly pisicivores) were retained. Between 1978 and 1982, the medium-sized species (8–17 cm SL) declined in the trawl catches with a 20 mm cod-end. For instance, the frequency of occurrence of two medium sized paedophagous species decreased from 80% and 63% in 1978 to 17% and 6% respectively in 1982 (X2 test P < .001 in both cases; Witte and Goudswaard, 1985). Moreover, even one of the smallest detritivorous species showed a decline in mean (and maximum) adult size, and in size at first maturity (Witte and Goudswaard, 1985).
Similar effects of trawl fishery, having a strong impact on the larger haplochromine species, were observed in Lake Malawi. Commercial trawl fishery started in 1968 with a 25 mm cod-end mesh, and in 1975 a change in dominance from medium-and large-sized species (maximum sizes 180–200 mm and > 200 mm total length (TL) respectively) to small-sized species (140–180 mm TL) was observed (Turner, 1977). At that time, the shift in composition from large to small species did not appear to have affected the catch per unit effort or the size of the commercial trawl catch (Turner, 1977). In 1977 the cod-end mesh size was increased to 38 mm (Tweddle and Magassa, 1989), and in the early 1990s new surveys were made (Turner et al., 1995). In the areas where most trawling took place “large and medium-sized benthic arthropod feeders, molluscivores, benthic zooplanktivores, large sediment feeders and large piscivores” had all declined. The total standing stocks of the small sediment feeders and pelagic species remained largely unchanged, but their proportion in the catches increased (Turner et al., 1995). It should be noted that the trawlers in Lake Malawi used cod-end meshes that were considerably larger (25 and 38 mm) than the cod-end meshes mostly used in Lake Victoria (20 mm). However, the small-sized haplochromine species in Lake Victoria (maximum size 7–9 cm SL ≈ 90–115 mm TL) were much smaller than those in Lake Malawi (140-180 mm TL). In fact, the small sized sediment feeders of Lake Malawi would fall in the medium size category of Lake Victoria haplochromines.
Our data suggest that bottom trawling, had a strong negative impact, especially on the larger haplochromines in the Mwanza Gulf, resulting in changes consistent with the “fishing down” model (Balirwa et al., 2003). However, one has to be cautious with this interpretation, and the data cannot rule out alternative factors driving the initial decline for two reasons: (1) there are no comparative data from before 1973, the year that trawling began; (2) there are no comparative data for the period 1979 to 1984 from other areas in the lake, where trawling was not performed. Such data would be required to rule out the possibility that the haplochromine decline began before commercial trawling was introduced.
Irrespective of that, fishery cannot explain all patterns of decline observed in the 1980s. In the 1970s and 1980s commercial trawl fishery only occurred in a very restricted area of the lake, mainly in the northern part of the Mwanza Gulf and its entrance. In contrast, the haplochromines declined lake-wide (Arunga, 1981; Okemwa, 1981; Acere, 1988; Ogutu-Ohwayo, 1990; Witte et al., 1995), also in deep offshore areas, where no fishing was done, and in lightly fished shallow areas such as the Emin Pasha Gulf (Goudswaard and Ligtvoet, 1988; Witte et al., 1995). In the Mwanza Gulf, trawl fishery for haplochromines was stopped by the end of 1986, when average catch rates were approximately 50 kg h− 1. The decline of the haplochromines continued and by September 1987 hardly any haplochromines were caught with the trawlers, though they were still encountered at low frequencies in Nile perch stomachs (Ligtvoet and Mkumbo, 1990; Mkumbo and Ligtvoet, 1992).
(ii) Eutrophication. In the 1980s eutrophication and concomitant algal blooms lead to a rapid decrease in water transparency and oxygen concentrations in Lake Victoria (Ochumba and Kibaara, 1989; Hecky and Bugenyi, 1992; Kaufman, 1992; Witte et al., 1992a; Gophen et al., 1993, 1995; Hecky, 1993; Mugidde, 1993; Hecky et al., 1994; Ochumba, 1995; Seehausen et al.,. 1997a; Wanink, 1998; Wanink and Kashindye, 1998; Wanink et al., 2001). Though eutrophication had already started as early as the 1920s, it appears that its pace increased in the late 1960s and early 1970s due to the fast growing human population in the riparian region (Hecky, 1993; Scheren et al., 2000; Ntiba et al., 2001; Verschuren et al., 2002). In the 1970s, in the sub-littoral areas of the Mwanza Gulf, the biomass of detritivorous haplochromines near the bottom was estimated at approximately 125 kg ha− 1 (Witte et al., 1999). In the entire water column, detritivores and phytoplanktivores together were estimated at 270 kg ha− 1 (Witte et al., 2005a). The disappearance of these groups in the early 1980s most likely decreased the consumption of phytoplankton (Goldschmidt et al., 1993; Ogutu-Ohwayo, 1999). At the same time, the Nile perch boom exacerbated deforestation along the lake shore, because the large and relatively fat perches had to be smoked or fried (Riedmiller, 1994; Ligtvoet et al., 1995). This contributed to erosion and to further siltation and eutrophication of the lake.
Several authors suggested that the decreased oxygen concentrations in deeper water (> 40 m) have been a major cause of the decline in haplochromine cichlids (e.g. Hecky et al., 1994; Bundy and Pitcher, 1995; Ochumba, 1995). According to Kaufman and Ochumba (1993) and Hecky et al., (1994) the hypoxic deep water area may have functioned as a refugium for the haplochromines because they tolerate lower levels of dissolved oxygen than do Nile perch. The progressive deoxygenation (< 1 mg l− 1) of the deep water over the years might have forced the demersal populations to the shallower water, where they were exposed to Nile perch predation. It has even been suggested that this was the cause of the Nile perch boom. However, it would imply that the original haplochromine densities in shallower areas were too low to sustain a Nile perch population, while in contrast the densities in the shallow areas were much higher than those in the deeper areas (Kudhongania and Cordone, 1974a, b; Verschuren et al., 2002). Moreover, the idea that the hypoxic layer in the open water is a potential refugium against Nile perch has been questioned by Wanink et al. (2001), who showed that Nile perch regularly explore the hypoxic layer. Similar observations have been made in Lake Tanganyika, where Lates mariae and some cichlid species were caught at depths where oxygen concentrations were less than 1 mg l− 1 (Coulter, 1967).
Laboratory experiments revealed that Lake Victoria haplochromines are relatively tolerant to low dissolved oxygen (DO) levels (Chapman et al., 1995). They seem to be more tolerant than Nile perch and the cyprinid Rastrineobola argentea (Wanink et al., 2001; Chapman et al., 2002), which are both flourishing in the lake despite the current hypoxia. Juveniles of several haplochromine species could be raised successfully at 10% air saturation (0.8 mg l− 1 DO), for more than a year, to adulthood (Rutjes et al., 2007). Negative effects of hypoxia are, however, still possible, as the scope for activity may have decreased in hypoxic areas and because severe hypoxia might hamper embryonic development (Witte et al., 2005b).
Decreased water clarity may pose a greater problem for haplochromine cichlids than hypoxia, as these fishes heavily depend on vision for feeding and reproduction. East African lakes with a high diversity of haplochromine cichlids have distinctly clearer water than lakes with a low diversity (Seehausen et al., 1997a). Also within Lake Victoria, species diversity is far lower in areas with low visibility than in clear water areas (Seehausen et al., 1997a; Witte et al., 2005b). A decrease in water clarity and light transmission interferes with visual mate choice and leads to hybridisation (Seehausen et al., 1997a; Seehausen and van Alphen, 1998), or even may completely frustrate breeding in some species (Witte et al., 1999, 2000). Further, low visibility may decrease prey selectivity and lead to increased interspecific competition as a result of loss of feeding specialization. This will have a negative impact on species coexistence (Seehausen et al., 2003). This latter mechanism would affect the highly visual predators of evasive prey in particular.
(iii) Predation by Nile perch. In 1983 Nile perch suddenly boomed in the Mwanza Gulf, mainly due to immigration of sub-adult fishes (Goudswaard and Witte, 1985). Concomitantly, the decline of some groups of haplochromines increased strongly in the sub-littoral and open waters, and shortly after the Nile perch peak in 1986–1987 haplochromines had virtually disappeared from the catches in these areas. Until the haplochromines had disappeared, they were the main food items of Nile perch (Ligtvoet and Mkumbo, 1990; Mkumbo and Ligtvoet, 1992). Scanty data from other parts of the lake indicate similar accelerations of the decline of haplochromines after Nile perch began to boom in those areas (Witte et al., 1995; Goudswaard et al., 2006). In shallow areas, with relatively low Nile perch densities (Goudswaard et al., 2002), and areas with structured bottoms, such as rocky shores, haplochromines were less affected (Witte et al., 1992b; Seehausen et al., 1997b).
In Lake Kyoga and Lake Nabugabo, where Nile perch had been introduced as well, the haplochromines also declined strongly with increasing Nile perch densities (Ogutu-Ohwayo, 1990, 1993, 1995). In contrast, in several small satellite lakes of Lake Victoria and Lake Kyoga, where Nile perch was absent, haplochromines remained abundant (Ogutu-Ohwayo, 1993; Namulemo and Mbabazi, 2000; Aloo, 2003, Mbabazi et al., 2004). However, it has to be mentioned as a confounding factor, that in some of these lakes the water was also clear (Kaufman et al., 1997; G. Namulemo, FRRI, Jinja, Uganda, pers. comm.). There are a few lakes where Nile perch and haplochomines seem to coexist. Aloo (2003) found both haplochromines and Nile perch in the murky Lake Sare (transparency 0.25 m), but did not record when Nile perch entered this lake and how many cichlid species used to live there before Nile perch introduction. Nile perch and haplochromines also coexist in Lake Saka in Uganda (O. Seehausen, University of Bern, Switzerland, pers. obs.). In Lake Nabugabo, haplochromines apparently found refugia in the hypoxic and highly structured shoreline wetlands (Chapman et al., 1996, 2002, 2003). The same may hold for a few wetland species of Lake Victoria, but not for the sub-littoral and deepwater species, or for those of sandy shores of Lake Victoria, because many of them were strongly restricted to these habitats (e.g. Witte, 1984) that are often at great distances from wetlands.
In contrast to (trawl)fishery, both eutrophication and the Nile perch boom were lake-wide phenomena (Arunga, 1981; Okemwa, 1981, 1984; Acere, 1988; Ochumba and Kibaara, 1989; Kaufman, 1992; Mugidde, 1993; Ogutu-Ohwayo, 1990; Hecky et al., 1994; Witte et al., 1995; Seehausen et al., 1997a; Wanink et al., 2001). Consequently, they provide more powerful explanations for the lake-wide decline of the haplochromines than local over-fishing. It has so far been impossible to establish the causal relationship between the Nile perch boom and eutrophication, and the relative impact of each of these phenomena separately. There are a number of reasons for this. First, both the Nile perch upsurge in Lake Victoria and the increase of eutrophication occurred between the late 1960s and early 1980s. Furthermore, systematic data on haplochromine abundance and diversity were not collected until 1969/70 and 1978 respectively (Kudhongania and Cordone, 1974a,b; Witte, 1981). Comparing data of satellite lakes with different degrees of eutrophication and different stages of Nile perch colonization may help solving the questions about the causes of the decline of the haplochromines.
The hypothesis that Nile perch had a large impact on haplochromine biomass is supported by the observations of a partial recovery of haplochromines in Lake Victoria, Lake Nabugabo and Lake Kyoga, following declines in Nile perch due to heavy fishing pressure (Ogutu-Ohwayo, 1995; Witte et al., 2000; Chapman et al., 2003; Getabu et al., 2003; Mbabazi et al., 2004). On the other hand, the incomplete recovery of mainly one trophic group in Lake Victoria suggests that Nile perch may not be the only factor (see below).
There are several potential causes for the differences in decline among the various haplochromine trophic groups and species. Initially, the fast decline of the piscivores, molluscivores and insectivores may be explained by selectivity of fishing gear for large species. A possible preference of Nile perch for larger prey may have played a role at a later stage. However, differential sensitivity to eutrophication cannot be excluded. The small detritivores and zooplanktivores were the last to disappear from the Mwanza Gulf. Despite their lower densities and equal or slightly larger size, the zooplanktivores persisted even longer in the catches than the detritivores. This may be explained by different habitat requirements of the two groups. Demersal detritivores would be affected more strongly than semi-pelagic zooplanktivores by bottom trawling, decreased light transmission and oxygen depletion, and they have larger habitat overlap with Nile perch (Goldschmidt et al., 1990; Goldschmidt et al., 1993; Goudswaard et al., 2004).
The size-selective decline within the piscivores suggests an important effect of fishing gear and Nile perch. The observation that large piscivorous haplochromines disappeared faster than small ones, from the catches with the small trawler, is corroborated by a comparison of catches with large trawlers in the same area in 1978 and 1985. In 1978 we caught 3 small, 19 intermediate and 5 large bottom-dwelling piscivorous species, and 2 pelagic species of intermediate size. In 1985, just before they eventually vanished, the only three piscivorous species that were caught were a small bottom-dweller and the two pelagic species (Witte et al., 1992a). In Lake Nabugabo, the large piscivore Haplochromis (Prognathochromis) venator (maximum size 17.8 cm SL; Greenwood, 1965) was the only haplochromine species that had completely vanished in the early 1990s (Ogutu-Ohwayo, 1993; Chapman et al., 2003). However, it was still present in the adjacent lakes Kayanja, Kayugi and Manywa where Nile perch had not been introduced (Ogutu-Ohwayo, 1993). On the other hand, according to Mbabazi et al. (2004), several piscivores, including H. (P.) argenteus (maximum size 20.2 cm SL; Greenwood, 1967), survived the Nile perch introduction in Lake Kyoga. During a more recent sampling programme (2000–2004), however, no piscivorous haplochromines were encountered in this lake (G. Namulemo, pers. comm.).
Although a bioenergetics model predicted the recovery of haplochromines with increasing exploitation of Nile perch (Kitchell et al., 1997), it could not predict which trophic groups or species were likely to revive. Compared to the 1970s, recovering species in the 1990s had to cope with: (1) decreased DO levels; (2) decreased water transparency; (3) changes in food composition; (4) a relatively high predation pressure, because, in spite of the decline of Nile perch, predator densities were still much higher than in the past. At the end of the 1990s piscivorous fish constituted more than 50% of the total fish mass, whereas in the 1970s they made up less than 10% (Witte et al., 2000).
Knowing this, how could it be explained that zooplanktivores recovered faster than detritivores? In the 1990s, fishing with bottom trawls had been stopped, gill net fishery used mesh sizes that were too large for haplochromines, and light fishery for the zooplanktivorous cyprinid Rastrineobola argentea mainly caught zooplanktivorous haplochromines as a by-catch (Witte et al., 2000). Therefore, fishery probably had a higher impact on zooplanktivores than on detritivores, and consequently cannot explain the slower recovery of the latter group. In contrast detritivorous species, which live closer to the bottom than zooplanktivores, could be more affected by low light transmission, low DO levels, or by the bottom dwelling Nile perch. Apart from these explanations, environmental changes that occurred concomitantly with the decline, or even more recently, may also have an impact on detritivores. Due to a shift in phytoplankton composition from a dominance of diatoms to a dominance of cyano-bacteria in the mid-1980s (Ochumba and Kibaara, 1989; Mugidde, 1993; Verschuren et al., 2002), digestibility and nutrient quality of detritus may have decreased. This might have an impact on the condition of detritivorous cichlids. In addition, elevated levels of heavy metals in bottom sediments and fish have been reported for various locations in Lake Victoria, including the Mwanza Gulf (Kishe and Machiwa, 2003; Campbell et al., 2003). Heavy metals may have a stronger impact on detritivores than on zooplanktivores, as bioaccumulation into the latter group is thought to be prevented by molting of zooplankton (A. Schäffer and H.T. Ratte, unpublished). Moreover, negative synergy between hypoxia and heavy metal contamination would produce stronger contamination effects in the deeper living detritivores (Yediler and Jacobs, 1995; Vosylienà and Kazlauskienà, 1999).
Differential recovery was also observed within the zooplanktivores. In the 1990s, Haplochromis (Yssichromis) pyrrhocephalus Witte and Witte-Maas, became the most common cichlid in the area, whereas its sister species H. (Y.) heusinkveldi Witte and Witte-Maas had disappeared (Witte et al., 2000). Before the ecological changes, both species had nearly identical distribution patterns (Goldschmidt et al., 1990) and were equally abundant. The retinas of H. pyrrhocephalus were likely to be more light-sensitive than those of H. heusinkveldi, whereas the retinas of the latter species probably allowed for better resolution at the cost of their photopic threshold (van der Meer and Bowmaker, 1995; van der Meer et al., 1995). This has been interpreted as an adaptation to a different feeding strategy in combination with the detection of small food items such as phytoplankton (Goldschmidt et al., 1990; van der Meer et al., 1995). Spawning in H. heusinkveldi mainly occurred in months with high water transparency, but H. pyrrhocephalus spawned all-year-round (Goldschmidt and Witte, 1990). Witte et al., (2000) suggested that the adaptation to feeding on phytoplankton may have constrained H. heusinkveldi's ability to spawn under turbid water conditions. The decreasing water transparency may have led to genetic introgression of H. heusinkveldi into H. pyrrhocephalus or other species, and would have prevented a recovery of H. heusinkveldi when Nile perch predation declined. Similarly, H. (Y.) “plumbus,” another closely-related species, survived the peak of the Nile perch abundance and was still found in 1991, but then failed to recover and ultimately disappeared.
The gill surface area of the recovering H. pyrrhocephalus is considerably larger than that of the pre-Nile perch population (Witte et al., in press). This is likely to be a response to the declining DO levels in the environment. In H. tanaos, another recovering zooplanktivore (van Oijen and Witte, 1996; Seehausen et al., 1997b), the retina shows indications of adaptation to decreased water transparency (L. Wagenaar, H. J. van der Meer, J. H. Wanink and F. Witte, Institute of Biology, Leiden, the Netherlands, unpublished). Ecological adaptations in the recovering zooplanktivores comprised, among other things, increased fecundity, habitat extension, and shifts to a more generalized diet, including larger and less evasive prey that can be obtained without good visibility (van Oijen and Witte, 1996; Wanink and Witte, 2000; Katunzi et al., 2003).
In summary, differential recovery (and decline) of the haplochromines may have been caused by interspecific differences in morphological and ecological features that already existed before the recent ecological changes in the lake; by differences in adaptability between species to the current environmental conditions; or they may simply be due to chance.
Potential effects of changed trophic dominance
The recovery of zooplanktivorous haplochromines is a lake-wide phenomenon. H. (Y.) pyrrhocephalus and H. (Y.) laparogramma Greenwood and Gee are the dominant haplochromines in the catches in the Mwanza area (Witte et al., 2000). The latter species and H. (Y.) fusiformis Greenwood and Gee (a species that was never found in the southern part of the lake) are currently common in the Ugandan and Kenyan waters (Ogutu-Ohwayo, 1999; R. Tumwebaze, Fisheries Resources Research Institute, Jinja, Uganda, unpublished; A. Getabu, I. Cowx and O. Seehausen, Kenya Marine and Fisheries Research Institute, Kisumu, Kenya, unpublished). In contrast to the 1970s, zooplanktivorous haplochromines are now far more common than detritivores (Figure 3; Table 2, 3). The effects on the ecosystem of the recovery of predominantly zooplanktivores are still unknown. Through competition the zooplanktivorous haplochromines may cause a decline in the population of the cyprinid Rastrineobola argentea, which had increased substantially with the decline of the haplochromine cichlids in the 1980s (Ogutu-Ohwayo, 1990; Wanink, 1991, 1998; Witte et al., 1999). The reversed dominance of detritivores and zooplanktivores and the changes in species diversity during the past decades may also have effects at the ecosystem level. The current dominance of the zooplanktivores may change the species composition and decrease the densities of zooplankton. Subsequently, such changes may cause an increase of phytoplankton in the lake (top down effect; Carpenter and Kitchell, 1993), which could be further enhanced by the decline of the detritivores. However, in the recovering species H. pyrrhocephalus and H. tanaos a change in diet has been observed resulting in a decreased contribution of zooplankton and an increase of larger prey such as insect larvae, shrimps, fish and even molluscs (van Oijen and Witte, 1996; Katunzi et al., 2003). Further, it is possible that the current dominance of zooplanktivores is only a temporary stage in the process of recovery of cichlid stocks.
Haplochromine diversity and fishery sustainability
The ecosystem effects of loss of functional diversity have been a controversial issue during the past 50 years (May, 1972; Rozdilsky and Stone, 2001). Some ecological studies during the past decade showed that greater diversity may lead to greater ecosystem predictability, temporal stability and in the long term to greater ecosystem productivity (Naeem and Li, 1997; McGrady-Steed et al., 1997; Norberg et al., 2001; Rozdilsky and Stone, 2001; Bolam et al., 2002). Adaptive foragers may play an important role in the relationship between ecosystem complexity and stability (Kondoh, 2003). Despite their reduction in species diversity, their high biomass after recovery suggests that haplochromine cichlids do still play a key role in Lake Victoria's ecosystem. Therefore, their species composition and ecology has to be carefully monitored in the coming years.
Balirwa et al. (2003) suggested that conservation of biodiversity and fishery sustainability may not have to be antitheses in the management of Lake Victoria. A modelling study suggested that Nile perch prefer and grow fastest on a haplochromine prey base (Kaufman and Schwarz, 2002). If the model is realistic, it would suggest that it is worth thinking of management strategies that allow enough fishing on Nile perch to ensure an abundance of their haplochromine prey, but not so much pressure as to threaten the Nile perch stock itself (Balirwa et al., 2003). However, to allow maintenance and restoration of haplochromine diversity, the most urgent measures must include serious attempts to reverse the eutrophication of Lake Victoria.
The rate at which haplochromine trophic groups in the study area declined in the 1980s differed; the relatively large piscivores, insectivores and molluscivores were the first to disappear. The small detritivores and zooplanktivores, which made up respectively 60-80% and 10-30% of the numbers of haplochromines, declined at lower rates, especially the zooplanktivores.
From the beginning of the 1990s a resurgence of some zooplanktivorous and detritivorous species was observed, but the relative abundance of the two groups reversed. By 2001, the number of zooplanktivorous individuals had reached their previous level and made up more than 80% of the haplochromines, whereas the former dominant detritivores constituted only about 15%.
The patterns of decline and recovery indicate that, though fishery played a role locally, predation by Nile perch and eutrophication were the main factors determining the decline and resurgence of the haplochromines. It has so far been impossible to establish the causal relationship between the Nile perch boom and eutrophication, and the relative impact of each of these phenomena separately. Further studies on satellite lakes with different degrees of eutrophication and different stages of Nile perch colonization may help solving the questions about the causes of the decline of the haplochromines.
We thank our colleagues from the Haplochromis Ecology Survey Team (HEST), the Tanzania Fisheries Research Institute (TAFIRI) and the Freshwater Fisheries Training Institute at Nyegezi for support and co-operation during the fieldwork. We wish to thank the crews of the trawlers for their skilful labour. We are indebted to Kees Barel, Gertrude Namulemo and Mike Richardson for their comments on earlier drafts of the paper and to Martin Brittijn for making of the figures. The research of HEST was financially supported by The Netherlands Foundation for the Advancement of Tropical Research (WOTRO; grants W87-129, W87-161, W87-189, W84-282, W84-488), the Section for Research and Technology of The Netherlands Minister of Development Co-operation, the Schure Beijerinck-Popping Fonds, the van Tienhoven Stichting and by Yellow Springs Instruments.