Lake Victoria is the largest of the African Great Lakes. It supports a riparian community of about 45 million people who depend on the lake directly and indirectly for fish protein, economic livelihoods, transport, climate regulation, cultural values, and other ecosystem services (Sun et al., 2014). The Lake Victoria ecosystem is also a unique biodiversity hot spot whose endemic cichlid fish species flock is a global inspiration to students of macroevolution (Kaufman, 1992; Seehausen et al., 1997; Witte et al., 1992, 2007). Most fishes and some invertebrates are endemic to Lake Victoria and its surrounding satellite lakes. Other members of the biota are more widespread either within the Nile headwaters or as members of the trans-Saharan assemblage. The lake has however undergone changes largely attributable to the introduction of invasive species, uncontrolled resource exploitation, climate change, and resource use conflicts (Hooper et al., 2005; Nyamweya et al., 2020). Lake Victoria's ecosystem has progressively deteriorated since the early 1960s both in water quality and fish species diversity (Aloo et al., 2017; Hecky et al., 1994; Ochumba & Kibaara, 1989; Sitoki et al., 2010; Stager et al., 2009; Verschuren et al., 2002; Witte et al., 1992). A few introduced species have come to dominate the flora and fauna of the lake (Acere, 1988; Kaufman, 1992; Ogutu-Ohwayo, 1990). The key drivers of this mass extinction are anthropogenic, chief among them cultural eutrophication, climate change, and exotic species introductions (Nyamweya et al., 2020; Witte et al., 2012). Catchment degradation coupled with point source and non-point source pollution has contributed to nutrient enrichment leading to water quality deterioration especially in Winam Gulf, Kenya (Lung'ayia et al., 2001; Sitoki et al., 2012; Gikuma-Njuru et al., 2013). The invasive water hyacinth (Eichhornia crassipes) has thrived, especially in the eutrophic shallow inner bays of the lake (Gichuki et al., 2012; Mwamburi et al., 2020).
The lake's changing ecology coupled with overfishing has led to reductions in the wild capture fishery, as well as decline and extinction within the endemic species flock (Kaufman, 1992; Nyamweya et al., 2020; Ogutu-Ohwayo, 1990; Taabu-Munyaho et al., 2016). Up to the late 1970s, the indigenous tilapiines Oreochromis esculentus and O. variabilis abounded, while the endemic haplochromine cichlids contributed up to 80% of the total catch landings in the fishery (Ogutu-Ohwayo, 1990). Then the fishery became dominated by three species, two of which were introduced to the lake: Nile perch (Lates niloticus) and Nile tilapia (O. niloticus). The third is a native shoaling pelagic minnow, “omena” (Rastrineobola argentea). Total annual fish catches lakewide amount to one million tons, with 230,000 tons of landed Nile perch fueling an export fishery that generates substantial foreign revenue (LVFO, 2016). Omena dominate in weight (400,000–600,000 tons), while catches of Nile tilapia declined by over 70% and were estimated at just 20,000 tons in 2015 (LVFO, 2016).
With declining catches from capture fisheries, cage aquaculture of Nile tilapia was introduced to augment fish production and meet local market demands and food security needs. Cage aquaculture across the world is diverse in the species farmed, farming intensity, and methods employed (FAO, 2006). High-income and some emerging economies of Southeast Asia have practiced cage culture for some time, leading to increased food security and per capita fish consumption (Beveridge & Muir, 1997; Naylor et al., 2021). However, it is a relatively new industry across the wider African Great Lakes. Cage aquaculture in Kenya consist of floating platforms (mean area 14.26 m2, ranging from 3.13 m2 to 105 m2) moored together to form large farms (Hamilton et al., 2020). The number of cages in Kenyan waters is growing rapidly yet the estimated 4357 cages occupy just 0.0621 km2 of the lake's surface area (Aura et al., 2018; Hamilton et al., 2020; Njiru et al., 2019). Cage fish farming as a livelihood has begun to offer a viable alternative to wild capture fisheries (Njiru et al., 2019).
There has long been environmental concerns about cage farms. Waste in the form of feed, feces, and chemicals can increase local nutrient loading and lead to eutrophication, increased turbidity, anoxia, and even fish kills (Dauda et al., 2019; Guo et al., 2009; Neofitou & Klaoudatos, 2008; Western, 2001). Cages can also influence fish communities in positive ways by providing structural refugia and increased trophic resources (FAO, 2006; Oakes & Pondella II, 2009). In Lake Victoria, areas with cage farms have reduced illegal, unreported, or unregulated (IUU) fishing due to the employment of 24-hour guards by farm owners to protect against poaching and encroachment of IUU fishers. Cage farms are positioned in nearshore areas popular for IUU fishing but poor for legal commercial fishing. This security confers a level of protection similar to that of a no-take marine reserve to cage farm areas. Fishing activities in the surrounding waters remain unrestricted; illegal monofilament and small-size mesh nets were consistently found in control areas during this survey. In Lake Victoria, there is a prospect that well-managed cage farms could act as jurisdictional and structural refugia from fishing pressure and natural predation, while also stimulating a periphyton-based benthic food web within a wider eutrophic, phytoplankton-based pelagic system. This has the potential to help restore and sustain some portion of Lake Victoria's biodiversity and wild capture fishery value while supporting regional food security. In Lake Tanganyika and Australia's Murray-Darling Basin, protected areas were found to have higher species' abundances and diversity than unprotected areas (Chessman, 2013; Sweke et al., 2016). Related studies that consider cage aquaculture farms as protected areas are scarce, and none exist for Lake Victoria.
This study examined how water quality and fish biomass and diversity varied between areas with and without cage farms and assessed the implications of these results for biodiversity conservation and aquaculture management. It was hypothesized that a higher abundance and diversity of fish species would be found at the cage areas than control areas, a higher nutrient load in Winam Gulf than in the open waters of Lake Victoria, and an elevated nutrient load and higher turbidity at cage stations. The results inform the global debate concerning cages as a source of nutrient enrichment, and as potential refugia in conserving fisheries resources in freshwater ecosystems.
METHODS Study areaThe study was conducted in Kenyan waters of Lake Victoria from November 2018 to July 2019 to account for seasonal changes in precipitation (short rains October to December and long rains March to May). Sampling sites were located within Winam Gulf and the open lake (Figure 1). The gulf is shallow with a mean depth of 10 m, and it is 70 km long and 15 to 30 km wide. It connects to the main lake by the 5 km wide Rusinga Channel, the primary channel of water exchange with the main lake (Calamari et al., 1995). The opening of Mbita channel in 2017 further increased exchange with the open lake (Simiyu et al., 2021). The gulf has undergone dramatic physicochemical and biological changes associated with anthropogenic activity in the past four decades, leading to a strong gradient in eutrophication and deteriorating water quality within the gulf (Lowe-McConnell, 2009; Mwamburi et al., 2020; Nyamweya et al., 2020; Ochumba & Kibaara, 1989; Sitoki et al., 2010; Taabu-Munyaho et al., 2016). In contrast, the water quality in the open lake has remained relatively oligotrophic (Sitoki et al., 2010).
Sampling strategy
Sampling was conducted at paired cage and control stations at four sampling sites: the inner gulf (Dunga and Ndere), mid gulf (Naya) and open lake (Mfangano) (Figure 1). Cage and control stations at each site were at minimum 1 km apart. Similar studies found measurable impacts on water quality did not extend past 100 m from cage farms (Nyanti et al., 2012; Price et al., 2015). Extending this distance tenfold ensured control stations were sufficiently independent of cage influence. The four sites chosen had cage farms of comparable area and are all considered low-density aquaculture regions (Hamilton et al., 2020). Paired control stations had flat lake bottoms with similar substrate composition of mud, sand, and gravel. The primary structural difference between cage and control stations were the cages themselves.
Physicochemical parameters were determined in situ at 1 m depth using a YSI 650 MDS with YSI 6600 multiparameter sonde. The parameters were measured in triplicate and included: depth, pH, dissolved oxygen (DO), temperature, turbidity, conductivity, total dissolved solids (TDS), chlorophyll-α, alkalinity, and oxidation–reduction potential (ORP). Three replicate water samples for nutrient analysis were taken at 1 m depth and fixed using sulfuric acid (H2SO4) and kept at 4°C until laboratory analysis and final quantification. Concentrations of the following nutrients were measured: total nitrogen (TN), total phosphorous (TP), ammonium (NH4+), nitrates (NO3−), nitrites (NO2−), and silicon (SiO2), using standard methods (APHA, 2005; Wetzel & Likens, 1991).
Fish samples were collected using gillnets following the Lake Victoria Fishery Organization's (LVFO) Standard Operating Procedures (LVFO, 2005a) for all fish sampling. Guidelines on collecting and monitoring biological and ecological information of fishes also directed environmental characterization (LVFO, 2005b). Fishes were sampled at depths between 2 and 10 m using three replicates of monofilament gillnets (two nets set parallel and one net set perpendicular to the shoreline). Six panels, each with a width of 1.37 m and length of 30 m (before mounting), were joined end to end to form one fleet 180 m in length. The mesh sizes of the panels ranged from 0.5″ to 2.0″ (inches) (i.e., 12.7 to 50.8 mm), at an interval of 0.25″ (i.e., 0.5″, 0.75″, 1.0″. 1.25″, 1.75″, 2.0″). Nets of different mesh sizes were used to increase the variability of sizes of fish and minimize sampling errors. The nets were set and hauled after 1 h. After hauling, all the fish samples (except haplochromines) were sorted and identified to species level according to Trewavas (1983), Witte and Van Densen (1995), and Greenwood (1966). The haplochromine cichlids were recorded as a group (haplochromines) due to the high difficulty of species-level identification in the field and fluid taxonomic status. Total catch, species, and wet weights of fish were recorded at each station.
Data analysisA principal components analysis (PCA) reduced the multidimensional dataset into fewer significant principal components (PCs) that represent large-scale patterns of limnology driven by highly correlated physicochemical, nutrient, and diversity parameters. Analysis of variance (ANOVA) tests were used to determine differences in PC scores (proxy variables for correlated patterns of limnology) and individual ecological parameters between sites (Dunga, Ndere, Naya, Mfangano) and station (cage and control). Tukey's honestly significant difference (HSD) Test on ANOVA results revealed which comparisons were significant. All statistical analyses were performed in RStudio version 1.3.1093 (RStudio Team, 2020).
The Shannon Diversity Index (H) was calculated as a composite measure of species diversity for each site and station, following Shannon and Weaver (1963). Though the value of H gives an indicator of overall diversity, it can also be difficult to understand. It was supplemented with two other parameters: the simple number of species present (S), and “evenness” of the community (J) following Pielou (1969). Species evenness (J) ranges from zero to one, with zero signifying low diversity (no evenness) and one, indicating high diversity (complete evenness).
RESULTS Spatial variation in limnology and diversityPC1 and PC2 accounted for the majority of limnological variation (Figure 2). A two-way ANOVA on PC1 and PC2 scores revealed significant differences between sites (PC1, FSite = 139.426, p < .001; PC2, FSite = 10.675, p < .001) but not between cage and control stations at any site. Thus, cage and control stations were grouped together at each site in further analyses. Tukey's HSD Tests revealed significant differences in PC1 scores for all comparisons except for Ndere and Dunga, the two inner gulf sites (p = 0 for all comparisons). Comparisons of PC2 scores were significant between Dunga-Naya (p < .001) and Naya-Mfangano (p < .001). PC1 (30.4%) represented nutrient loading, species diversity, and water clarity, while PC2 (9.4%) accounted for pH, temperature, and DO (Appendix S1). Dunga and Ndere were characterized by low-water clarity, high-overall non-haplochromine species diversity, and a high-nutrient load, while the transitional Naya and open-lake Mfangano had progressively clearer water and lower nutrients. The PCA indicates a strong association of Dunga and Ndere with high conductivity, turbidity, TDS, alkalinity, hardness, SiO2, ORP, chlorophyll-α, TP, SRP, NH4+, and TN. Interestingly, Shannon diversity and evenness for non-haplochromine species were positively correlated with nutrients and turbidity (Appendix S1).
Enlarge this image.FIGURE 2. PCA biplot of all measured variables. PC1 accounts for nutrient loading, overall species diversity, and water clarity with low PC1 scores having high loading, high diversity, and low clarity. PC2 accounts for temperature, DO, and pH, reflecting seasonal variability
Seasonal variability of physicochemical parameters was highest in the inner gulf and decreased towards the open lake (Figure 2). Variables associated with water transparency (turbidity, TDS, secchi depth) indicated a steady increase in water clarity from the inner gulf to the open lake. The highest values for turbidity and TDS were recorded at Dunga and lowest at Mfangano; secchi depth showed a reverse trend and ranged from 2.65 (SD 0.4) m at Mfangano to 0.38 (SD 0.09) m at Dunga. Site had a significant effect on secchi depth (F = 305.7, p < .001), and Tukey's HSD Test revealed this was significant between all sites except for Dunga and Ndere (p = 0 for all significant comparisons). Site also had a significant effect on conductivity (F = 80.82, p < .001), alkalinity (F = 40.48, p < .001), and hardness (F = 11.69, p < .001), which followed the same spatial gradient as water clarity. Dunga had a lower pH than all other sites (F = 12.22, p < .001) (Tukey's HSD; p < .001 for all comparisons); there was no variation among the other sites. Surface water temperature, DO, and ORP did not vary along the inner gulf-open lake gradient, and there were no significant differences between cage and control stations at each site for any physicochemical parameter.
Nutrients and chlorophyll-αNutrient concentrations fluctuated greatly at Dunga and Ndere; both seasonal variability and loading decreased towards the open lake (Figure 2). Site had a significant effect on NO3− (F = 11.47, p < .001), NO2− (F = 26.9, p < .001), SRP (F = 5.275, p < .01), SiO2 (F = 24.71, p < .001), NH4+ (F = 4.783, p < .01), and TP (F = 12.23, p < .001). All these parameters were highest at Dunga and Ndere, and Tukey's HSD Tests revealed significant differences between the open lake and inner gulf for all variables (p < .01 for all variables) except for SRP.
The measure of phytoplankton biomass, represented as chlorophyll-α concentration, differed significantly between sites (F = 3.168, p < .05), though only between Mfangano and Dunga (Tukey HSD, p < .05). At Ndere and Dunga, concentrations of chlorophyll-α peaked in March, while levels remained relatively constant throughout the year at Naya and Mfangano. Mean chlorophyll-α ranged from 59.74 (SD 96.53) μgL−1 at Dunga to 6.07 (SD 1.9) μgL−1 at Mfangano (Appendix S2).
Fish biomassOverall, a two-way ANOVA revealed cage stations had a significantly higher mean total monthly biomass than control stations, and there was no difference between sites (FStation = 6.517, p < .05; FSite = 0.499, p = .684) (Figure 3a). In total 17 species (considering Haplochromis spp. as one group) were recorded, belonging to eight families (Appendix S3).
Enlarge this image.FIGURE 3. Mean monthly biomass (in grams) of threatened fish species caught near cage and control stations in Lake Victoria, Kenya from November 2018 to July 2019
For L. niloticus, there was no significant difference in biomass between cage or control stations at any site, though the biomass at Ndere was significantly higher than at Mfangano (FSite = 2.753, p < .05; Tukey's HSD, p < .05). The biomass of Enteromius apleurogramma was significantly higher at the Dunga cages than at any other site or station. The highest Brycinus sadleri biomass was recorded at Naya cages, though this was not significantly different from any other site or station. Mean Brycinus jacksonii biomass was significantly higher at the Naya cages than all other sites and stations, including the Naya control (FSite = 21.98, p < .001; FStation = 13.92, p < .001) (Tukey's HSD, p < .001 for all comparisons) (Figure 3c). Synodontis victoriae biomass differed significantly between sites (F = 14.04, p < .001) and stations (F = 10.64, p < .01), with cage stations having a higher mean biomass (Figure 3d). There was no difference in S. afrofischeri biomass between any sites or stations. The highest Haplochromis spp. biomass was found at Mfangano (Figure 3b), which was significantly higher than all other sites (FSite = 76.347, p < .001) with no difference between cage and control stations (FStation = 0.012, p = .911). Other fish species recorded, though in low numbers or single specimens, included O. niloticus, O. variabilis (very rare), Schilbe intermedius, Pollimyrus nigricans, Marcusenius plagiostoma, M. victoriae, Mormyrus kannume, Coptodon zillii, and E. profundus.
Fish species diversityShannon diversity was greatest at the most eutrophic sites, with greater diversity of non-haplochromine species in the inner gulf than the open lake (FSite = 29.657, p < .001; Tukey's HSD, p = 0) and no difference between cage and control stations (Figure 4a). The same pattern was found for mean species number (FSite = 18.364, p < .001; Tukey HSD, p = 0) (Figure 4b) and evenness (FSite = 25.55, p < .001; Tukey's HSD, p = 0) (Figure 4c).
Enlarge this image.FIGURE 4. Mean diversity indices for sampled sites and stations in Lake Victoria, Kenya from November 2018 to July 2019
Results from this study confirm the existence of a strong limnological gradient in Winam Gulf, with the inner gulf holding highly eutrophic, turbid waters that progressively clear towards the open lake (Figure 2). Higher fish biomass (Figure 3a) and significantly greater diversity of non-haplochromine species were observed in the gulf than in the open lake (Figure 4a–c), and several species were found to be most abundant at cage stations or within the inner gulf (Figure 3b–d). These results have implications for the future of cage aquaculture, biodiversity conservation, and the marriage between the two in Lake Victoria.
Nutrient concentrations and physicochemical parameters could vary spatially between cage and control stations, between inner and outer gulf stations, and temporally across the months of sampling. Little evidence was found for differences between paired cage and control stations, but all sites were strongly affected by their position along the inner gulf to open lake gradient. Water transparency was positively correlated with increasing distance from the inner gulf, while nutrient loading showed a reverse trend (Figure 2, Appendix S1). This inner gulf-open lake limnological gradient is consistent with previous studies of Winam Gulf's limnology, including those conducted prior to the onset of cage aquaculture (Ochumba & Kibaara, 1989; Lung'ayia et al., 2001; Gikuma-Njuru & Hecky, 2005; Sitoki et al., 2010; Juma et al., 2014; Kundu et al., 2017; Guya, 2019; Mwamburi et al., 2020; Nyamweya et al., 2020). Thus, the gradient is primarily still driven by terrestrial and riverine inputs rather than cage aquaculture.
Spatial trends of physicochemical variables showed higher mean values of turbidity, conductivity, TDS, hardness, alkalinity, and silicates in the inner gulf (Appendix S2). Water transparency showed an inverse trend with high-mean light penetration of 2.65 (SD 0.4) m in the open lake at Mfangano cages and low-mean values of 0.38 (SD 0.9) m in the gulf at Dunga cages. This is much lower than earlier values of 1.1–1.6 m reported in Winam Gulf (Gikuma-Njuru & Hecky, 2005). The high turbidity and low-water clarity at Dunga and Ndere is not surprising; the gulf has high-suspended solids content following the short rains (October–December) and long rainy season of March–June in Kenya (Mwamburi et al., 2020). Much of the materials in the gulf are depositions from rivers draining into the lake from the catchment. The rivers transport materials from both anthropogenic activities and natural rock weathering and mineral dissolution processes in the watershed rocks (Nyamweya et al., 2020). Light limitation within the gulf (Gikuma-Njuru & Hecky, 2005) has been associated with limitation of primary production, and eventual dominance by planktonic cyanobacteria. Low turbidity in the open waters is attributed to increased settling of deposited particulate matter and the dilution effect of the deep offshore waters (Mwamburi et al., 2020). Conductivity decreased from a high of 174.89 (SD 33.75) μS cm−1 at Dunga cages to 99.02 (SD 6.33) μS cm−1 at Mfangano cages; the higher mean values in the gulf are driven by terrestrial runoff and discharges from rivers. There was no spatial variation in DO concentration. However, there were temporal differences with low DO concentrations at all sites recorded in June. This coincides with a period of deep water column mixing experienced from June to July during which there is stratification breaking and hypoxic water from the lake bottom comes to surface (Hecky et al., 1994). Surface water temperature did not vary among the four sites, with a mean of 26.58 (SD 0.44) °C for the entire gulf.
There was a spatial gradient in nutrient loading with significantly higher concentrations in the gulf than the open lake (Figure 2). However, little evidence was found that suggests cage farms are driving this gradient. The temporal fluctuations in nutrient levels were more pronounced in the inner gulf than in the open waters. This is explained by the shape and bathymetry of the inner gulf, which consists of embayments and vast shallow littoral areas near river mouth discharges. These shallow sites within the gulf, such as Ndere and Dunga, are likely more influenced by riverine inputs, urban activities, and other land-based runoff inputs than by effluent from cage farms. The higher chlorophyll-α concentrations in the inner gulf are a consequence of the proximity to terrigenous and riverine nutrient inputs. Many rivers entering the inner gulf bring in silt, wastes, and other effluents from the catchment. These wastes increase turbidity in the inner gulf and decrease dissolved oxygen in the water column due to their biodegradation (Nyamweya et al., 2020; Sitoki et al., 2010). Water quality affects aquatic species abundance, composition, stability, productivity and physiological condition (APHA, 2005). In the inner gulf (Ndere and Dunga), the water quality of cage stations was no different than the ambient conditions of the surrounding waters. This is likely a signal that terrestrial anthropogenic inputs, which are a major source of high nutrient concentrations, are driving the limnology of the inner gulf (Guya, 2019; Sitoki et al., 2012). The eutrophic state of the gulf is a concern; the occurrence of large algal blooms in eutrophic areas is a threat to aquatic ecosystems services and native fauna. Lake circulation patterns drive major in-lake processes that cause distribution of elements, nutrients, and organisms. Several studies have described lake characteristics as being dependent on the influences of rainfall and evaporation processes, leading to spatially homogenous conditions (MacIntyre et al., 2014). The net effects of the complex hydrodynamics in lakes determine the dominant processes and extent of nutrient re-mobilization, especially in the shallow and intermediate depths (5–10 m) of the gulf.
Overall, cage stations had a significantly higher mean total monthly fish biomass than controls, and there was no overall difference between sites. Despite the gulf being highly eutrophic, it held a high biomass of non-haplochromine fishes. The high biomass of L. niloticus in the gulf would suggest either the fish use the inner gulf as a foraging or breeding ground. The abundance of squeaker catfish (S. victoriae and S. afrofischeri) was remarkably higher near the cages in the inner gulf, where they likely feed on periphyton that establishes on the cage nets and structures. The submerged cage structures and nets provide substrate for the establishment of periphyton, which creates food-rich microhabitats for fish to forage as they take shelter/refuge beneath the cages. B. sadleri and B. jacksonii were most abundant at the Naya cages. The co-occurrence of these two species is surprising from what is previously known; B. sadleri was found to occur in pelagic areas far from river mouths, whereas B. jacksonii were confined to the river mouths only (Ojuok, 2008). This would suggest the cage environment at Naya has river mouth-like conditions, and attraction to the cages may be enhanced by abundant periphyton and shelter. Cages typically extend from 2 to 6 m below the surface, providing a significant structural increase in otherwise featureless areas. E. apleurogramma and L. victorianus, two cyprinids, were most abundant at the Dunga cage and control, respectively. These benthic feeders scrape material from submerged structure and substrate, likely benefiting from the periphyton and structure of the cages and eutrophic waters. Haplochromine biomass was significantly highest in the open lake and thus also related to high water transparency. In the early 1970s, haplochromine cichlids and other indigenous fishes comprised the major fishery in Winam Gulf (Wanjala & Marten, 1974). This is no longer the case. The low haplochromine biomass in the inner gulf as reported in this study should be a wake-up call on how rapidly changing ecology and degrading habitats can lead to species extirpation from a given area.
The Shannon diversity index indicated that fish species diversity for non-haplochromines was significantly higher in the inner gulf than in the open waters. Shannon diversity and evenness for non-haplochromine fishes were positively correlated with nutrients and turbidity. Though highly eutrophic, the inner Winam Gulf is still a critical habitat to these diverse species. Most of these fishes favor riverine and wetland habitats that serve as spawning and nursery areas (Nyamweya et al., 2020). Given the anthropogenic inputs from the catchment into the rivers, plus wetland shrinkage due to encroachment, the inner gulf could be the only habitat available for fluvial fishes. This underscores the importance of the inner gulf as a critical habitat for species of commercial, ecological, and social importance. The most abundant species recorded in the inner gulf at Dunga and Ndere were L. niloticus (Nile perch), E. apleurogramma (a barb), B. sadleri (a small tetra), L. victorianus (an endemic, potamodromous cyprinid), S. afrofischeri and S. victoriae (catfish).
Haplochromines were recorded separately as a group and were dominant in open water at Mfangano. The haplochromine cichlids are extremely diverse ecologically; however, their modest overall morphological diversity, prevalence of intraspecific variation, and the existence of many undescribed and previously unknown taxonomic entities make a rapid assignment of individuals to known taxa in the field a difficult and incomplete exercise (Witte et al., 2007). Therefore, the haplochromine cichlid flock was, for the purposes of this study, considered as a single unit. Viewed in this way, fish species diversity and richness were higher in the gulf than the open lake with no significant differences between cages and controls. However, the haplochromines found at Mfangano comprise multiple genera and trophic groups, which the diversity indices do not account for. The clear water of the open lake promotes trophic differentiation, species coexistence, sexual selection, and year-round spawning, indicating that the functional diversity of Mfangano is much greater than what is reported in this study (Seehausen & van Alpen, 1997, Witte et al., 2012).
It is worth noting that very few tilapiines (both introduced and indigenous) were recorded during this study. No single indigenous O. esculentus was recorded, and only two individuals of O. variabilis were found, at Mfangano. Few individuals of the introduced Nile tilapia (O. niloticus) were caught; its plummeting catches is a major reason for its prevalence in cage aquaculture in the lake.
Fishing pressure in Lake Victoria is currently very high (Van der Knaap, 2013), whereas all forms of exploitation including fishing are prohibited around the cage areas, and cage owners actively restrict any encroachment by fishermen. This study found that these cage areas have a higher total fish biomass than controls. B. jacksonii biomass was highest at the Naya cages, and significantly higher than at the control (Tukey HSD, p < .001) as well as every other station at each site. This suggests that in areas with high-fishing pressure like Naya, cages could act as a refugium and foraging ground for at least some fishes. Other studies have documented that protected areas harbor higher fish diversity as compared to unprotected areas (Sahyoun et al., 2013). Poor enforcement of fisheries regulations by local and national authorities can help to explain the low-fish biomass and diversity in a station like Naya control–an area not protected from exploitation by cage farmers. The use of prohibited fishing gears, like beach seine nets and monofilaments, was encountered in all sampled sites during the study, and in some places distanced from cage farms, gill net density can be so high as to impede free passage of a boat. Many nets are lost, and the use of illegal, unreported, and unregulated fishing (IUU) and ghost gear are serious threats to sustainable fisheries, conservation, and food security.
Given the eutrophic status of Winam Gulf, it would be advisable to have low fish stocking densities at inner gulf cages and higher stocking densities in the open lake cages. Sites with good water exchange can have higher stocking densities, while those with poor water exchange should use lower stocking densities (Beveridge & Muir, 1997). Cage aquaculture has associated environmental and ecological problems, and these must be considered critically at all stages from the initial planning to site selection, and finally as part of the regular monitoring for daily operation. In Lake Victoria, these concerns have been recently raised by cage farmers, conservationists, and fishermen, particularly regarding the potential negative effect on the critical export fishery and endemic biodiversity. Most importantly, the findings of this study suggest that the current state of cage aquaculture is not the major cause of eutrophication in Winam Gulf. Rather, terrestrial, and other sources are the major contributing factors; cage farms have no significantly higher nutrient enrichment or turbidity than the ambient conditions of non-cage areas. Even at Mfangano, where deep water and fast dispersal due to wind and currents maintain oligotrophic conditions, the cage farms did not have a significant environmental signature. The capacity of the environment to assimilate nutrients varies greatly according to local conditions of depth, hydrography, and water exchange and sediment type (FAO, 2006). This gives promise to the expansion of cage aquaculture in Lake Victoria. Well-managed cage farms positioned in open, deep waters may contribute little to cultural eutrophication. However, the fact that eutrophication is already a serious problem in the lake, particularly in Winam Gulf, is not a reason to permit unregulated cage aquaculture. Cages should not be installed in areas that are known fish spawning grounds to avoid tampering with recruitment, or in areas that are prominent fishing grounds for capture fisheries. Doing so might redirect fishing pressure to limited areas thus posing management challenges. However, if cages are well managed, they can act as a refugia attracting diverse fish assemblages, which can enhance biodiversity conservation and, potentially, the fishery. In Indian inland waters, submerged bamboo structures assisted the growth of periphyton on the submerged parts. This served as a refuge for feeding either by grazing on periphyton, micro-invertebrates and insects or by predation on smaller fishes that take shelter under these refuge (Suresh, 2000). Such provisions have also been reported from Malawi (Banda et al., 2005). Taken together, these results suggest that cage farms positioned in outer gulf and open lake areas (Naya and Mfangano) can act as a refugia from fishing, while supporting a periphyton-based food web without discernably contributing to eutrophication. In addition, the growth of cage aquaculture may take pressure off the wild capture fishery.
As cage aquaculture continues to expand in the region, there is a cumulative and eventually wider effect on the environment. Therefore, continued coupled human and natural systems (CHANS) studies (Liu et al., 2007; Walker et al., 2004) should be prioritized to provide regular monitoring, management guidelines for best cage practices, and the capability to recognize when limits are reached. Future studies should give particular attention to how the riparian community can be engaged in sustainable cage aquaculture, wild capture fisheries, and biodiversity conservation in the region to safeguard food security and livelihoods.
AUTHOR CONTRIBUTIONSConceptualization: John Kengere Okechi, Leslie Kaufman; Methodology: John Kengere Okechi, Leslie Kaufman; Formal analysis: John Kengere Okechi, Nick Andrew Peoples, Chrisphine Sangara Nyamweya; Investigation: John Kengere Okechi, Chrisphine Sangara Nyamweya; Resources: Leslie Kaufman, Sarah Glaser; Data curation: John Kengere Okechi; Writing –original draft: John Kengere Okechi; Writing –review & editing: John Kengere Okechi, Nick Andrew Peoples, Leslie Kaufman, Chrisphine Sangara Nyamweya, Sarah Glaser; Supervision: Leslie Kaufman; Project administration: Leslie Kaufman, Sarah Glaser; Funding acquisition: Leslie Kaufman, Sarah Glaser.
ACKNOWLEDGMENTSThis study was funded by the National Science Foundation (NSF) Award #1518532 and supported by Boston University and Kenyan Marine and Fisheries Research Institute (KMFRI). The authors thank the KMFRI staff for logistics and support during sampling and lab processing of samples, and the cage farm owners for allowing access to their sites.
CONFLICT OF INTERESTThe authors declare no potential conflict of interest.
DATA AVAILABILITY STATEMENTThe data used for these analyses is freely available at:
© 2022. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Peoples, Nick 2
; Nyamweya, Chrisphine Sangara 3
; Glaser, Sarah 4 ; Kaufman, Leslie 5
1 Department of Biology, Boston University, Boston, Massachusetts, USA; Department of Freshwater Systems|Aquaculture, Kenya Marine and Fisheries Research Institute (KMFRI), Kisumu Research Centre, Kisumu, Kenya
2 Department of Biology, Boston University, Boston, Massachusetts, USA; Department of Evolution and Ecology, University of California, Davis, California, USA
3 Department of Freshwater Systems|Aquaculture, Kenya Marine and Fisheries Research Institute (KMFRI), Kisumu Research Centre, Kisumu, Kenya
4 Secure Fisheries program, One Earth Future Foundation, Broomfield, Colorado, USA
5 Department of Biology, Boston University, Boston, Massachusetts, USA