© L.A. Rasoamihaingo et al., Published by EDP Sciences 2023

1 Introduction

In the biodiversity hotspot of Madagascar, freshwater ecosystems receive little attention compared to forest and marine systems (Bamford et al., 2017) despite containing similar proportions of endangered species (Máiz-Tomé et al., 2018) and being more threatened (Kull, 2012, Bamford et al., 2017). Stressors faced by wetlands in Madagascar are typical of tropical countries, including habitat clearance for agriculture, invasive species, deforestation, and pollution (Junk et al., 2013, Bamford et al., 2017, Máiz-Tomé et al., 2018, Williams-Subiza and Epele, 2021). Also in common with other tropical regions, freshwater wetlands are under-represented in protected area networks (Bamford et al., 2017, Reis et al., 2017). The overwhelming majority of lakes in Madagascar have turbid water, few benthic invertebrates and few submerged or floating macrophytes (Bamford et al., 2017). While there is no data on algae or plankton, the vast majority of lakes appear to be eutrophic. High rates of sedimentation due to deforestation are generally blamed for turbid, eutrophic lakes (e.g., Máiz-Tomé et al., 2018), as increased sedimentation is well known to have severe consequences for lakes (Donohue and Molinos, 2009). However, there is there is little data on the extent or impacts of wetland stressors, so it is not clear if deforestation is the sole cause of turbidity in Malagasy lakes, or if turbidity is the sole cause of the lack of submerged macrophytes and benthic invertebrates.

Nearly all lakes in the country contain several species of introduced fish (Sparks and Stiassny, 2003, Bamford et al., 2017). Competition from and predation by introduced species is often blamed for declining native fish populations (Sparks and Stiassny, 2003, Canonico et al., 2005), but their role in causing other changes to the lake ecosystems has not been studied in detail (Canonico et al., 2005, Vejříková et al., 2018). For example, common carp Cyprinus carpio can have a substantial role in increasing turbidity and reducing populations of benthic invertebrates and submerged macrophytes (e.g. Zambrano et al., 2001, Williams et al., 2002, Miller and Crowl, 2006). Carp are widespread in Madagascar and their effect on lake systems is often overlooked. However, the most widespread introduced fish in Madagascar are various tilapia, and the ecological effects of these species are poorly studied anywhere in their introduced range. Two species in the genus Oreochromis, Nile and Mozambique tilapia (hereafter referred to as tilapia), are most commonly introduced and they live feral in every country where they have been cultured (Canonico et al., 2005). In Madagascar, they are deliberately introduced to lakes to provide protein and livelihoods in poor communities. Their ability to thrive in a wide range of environmental conditions makes them well suited to this role (Canonico et al., 2005) and consequently tilapia dominate the catch in many lakes in Madagascar (e.g. Lammers et al., 2020). The role of tilapia in outcompeting native fish species is well studied, but there is less information on effects they have on other aspects of the lake ecosystems.

Here, we aim to determine the effects of two wetland stressors, introduced tilapia and turbidity, on macrophytes and benthic invertebrates. An additional aim was to determine if management interventions can quickly increase the abundance of either macrophytes or invertebrates. We utilise in situ enclosure experiments to (1) quantify the effects of tilapia and turbidity on macrophyte survival and invertebrate abundance and diversity, (2) determine if tilapia play a role in increasing turbidity, (3) determine if tilapia effects vary with stocking density, and (4) test management interventions including re-introducing macrophytes and adding woody debris to the sediment.

2 Methods

2.1 Study site

Lake Sofia, in the northern highlands of Madagascar (Bealanana District, Sofia Region), is in relatively good condition. The lake retains a large, healthy marsh, dominated by Cyperus papyrus, and is a dry season refuge to large numbers of ducks, including the Endangered Meller's duck Anas melleri. Lake Sofia is representative of stressors faced by wetlands in Madagascar, but most are present at low levels (Bamford et al., 2017). For example, there is marsh clearance for agriculture, but natural vegetation surrounds the lake. Sedimentation is lower than at similar sized lakes. Crucially for our aims, tilapia are the only introduced non-native species of fish, and fishing pressure is low. Despite all this, Lake Sofia is a eutrophic lake with few benthic invertebrates, limited floating macrophytes and no submerged macrophytes. This state may represent a recent change, as photographs show abundant floating vegetation covering the lake as recently as 1960. Residents report that this vegetation consisted of water lilies, and that fish used to be more abundant in the lake. No information is available on the native fish community, with most residents stating there were no fish in the lake before the introduction of tilapia. Small lakes in the catchment have abundant submerged vegetation, mostly Charophyte algae. It is reasonable to suppose that Lake Sofia contained similar algae, as is typical for a papyrus-dominated lake (Pacini et al., 2018).

Lake Sofia is 1100 metres above sea level and covers 2 km², reaching a maximum depth of 4 m in the rainy season or 3 m in the dry season. There is 3.5 km² of papyrus-dominated marsh, with a fringe of floating mats of grasses and ferns and few water lilies (Nymphaeaceae species). Beyond the marsh are 17 km² of rain-fed rice fields. The lake catchment is almost entirely deforested, but historically would have been part of Madagascar's eastern rainforest belt. The lake has simple hydrology, being river fed with two main inflows plus numerous small seasonal channels. There is a single outflow. Rainfall averages 1500 mm annually, mostly from January to March. There is little flow modification. Catchment soils are mainly ferrosols, red clay based soils with little organic content (Ramifehiarivo et al., 2016). Water quality measurements suggest the lake is eutrophic, with high phosphate levels (30 μg/l), low nitrate levels (160 μg/l), and high turbidity (16 NTU, or secchi depth of 0.6 m). The water is well oxygenated, with no difference between the surface (7.5 mg/l) and 2 metres depth (7.0 mg/l), although there is a drop in DO levels below 2 m. Benthic Chironomidae abundance is low (mean ± S.E. = 100±11 individuals m⁻²). See Pruvot et al. (2020) for additional detail.

2.2 Exclosure experiments

We used a split-plot experimental design to manipulate the presence of tilapia and the substrate type. It was not possible to manipulate the turbidity level in this in situ experiment. As a proxy, we instead varied depth, with treatments placed close to the surface and at 1.5 m depth. Water quality analysis (see Study Site above) show no difference in nutrient levels between the surface and 1.5 m depth, so only light levels and pressure differ. Substrates tested were introduced water lilies and Charophyte algae, plus coarse woody debris and lake sediment transplanted as a control. The experiment was run twice with different fish stocking densities. Six cages were constructed as fish exclosures (Fig. 1). Each cage measured 1.5 m(W) ×  1.5 m(D) × 2.0 m(H), made from a wooden frame with 1.5 mm wire mesh sides and bottoms. The mesh allows small invertebrates to pass. It may also allow small fry fish through, but this is not likely to cause a problem over the length of these experiments. Four treatments were made, in plastic trays (measuring 35 cm × 35 cm × 20 cm deep):

• Lake sediment as a control (2 litres of wet sediment).

• Coarse woody detritus (1.5 kg dry weight).

• Charophytes (Charophyceae species, 500 g wet weight) growing in sediment (1 litre wet sediment).

• Water lilies (Nymphaeaceae species, whole plants) growing in sediment.

Lake sediment was transplanted directly from the area of the experiments. Detritus was taken from native forests near the lakeshore. Charophytes were transplanted from another lake (3 m deep) within the Lake Sofia catchment and rinsed clean. Few Charophytes remain in the study area so we were limited to the single species that could be sourced locally. Water lilies were transplanted from the marsh surrounding the lake. Detritus for the treatments was measured by dry weight, but Charophytes and sediments had to be measured by wet weight or volume. Water lilies were transplanted as whole plants, so no standardisation in size was possible. Two trays of each treatment were placed into each exclosure, one suspended close to the surface and suspended at 1.5 m depth (deep vs. shallow, as a proxy for turbid and non-turbid conditions). This does not apply to the water lilies, which were placed near the surface only.

Finally, tilapia from the lake were then introduced into three of the cages. The experiment was run twice, with different fish stocking densities. In Experiment I we used four tilapia per cage, and in Experiment II we used two tilapia per cage. Tilapia were caught from the lake and fish of roughly equal size (11–13 cm total length) were selected for the experiment. Experiment I ran from March 2020 until May (7 weeks) and Experiment II ran from September 2020 until November (6 weeks). The experiments were stopped when the wire mesh used to construct the cages started to degrade before any fish could escape or enter. The cages were located in the lake within 10 metres of the emergent fringing vegetation.

The cages were checked daily during the experiments, but to avoid disturbance the individual treatments were not checked. Turbidity and dissolved oxygen were measured weekly at the surface and bottom of cages using a Micro 600 DO meter and Photometer 7100, both made by Palintest (Gateshead, UK). At the end of the experiment, we retrieved the treatments by gently pulling them up to the surface and immediately searching for invertebrates. Specimens were stored in ethanol for later identification. The remaining sediment and plant material was dried and dry weight recorded. Water lilies were photographed at the start and end of the experiments, and their relative size at the end (compared to the start) was scored based on the number of leaves and the size of each leaf.

Fig. 1 (A) The experimental set up, a split-plot design with fish introduced into 3 of the 6 cages, and three treatments in each cage near the surface (S) and the same three treatments at 1.5 m depth (B), plus a Nymphaeaceae plant transplanted into each cage. Treatments were assigned positions at random for each cage. The three treatments were: lake sediment, woody detritus, and transplanted Charophyte plants. The experiment was run twice with different fish stocking densities. (B) A side on view of one cage to show the differing depths. (C) A photograph of the cages in situ, February 2020.

2.3 Statistical analysis

Analyses were carried out in R v4.0.5 (R Core Team, 2021). Invertebrate abundance was measured as counts of individuals, diversity as Gini-Simpson index calculated on insect genera. We related response variates to explanatory variables in GLMs, assuming a Gaussian distribution for weight and size variates (plants and substrate), a quasi-binomial distribution for diversity indices, and a quasi-poisson distribution for invertebrate counts. Quasi-poisson distributions were used to account for model dispersion being divergent from 1. Experiments I and II were analysed separately, and then pooled data from the two experiments was analysed with experiment number included as a two-level random effect. Significance levels were checked using an ANOVA on the full model. Model performance and assumptions were checked using the package performance.

3 Results

3.1 Effects on macrophyte survival and substrate suspension

Charophyte survival was affected by both depth and fish presence in both experiments and in the overall data (Tab. 1). The mass of Charophytes was zero in all samples when fish were present, and greater mass of Charophytes remained in the shallow treatments (Fig. 2). Results were the same in Experiments I and II, so there was no seasonal effect and no effect of stocking densities.

Water lily survival was also affected by fish presence (F_1,34 = 6.3, P = 0.01). Fish presence was associated with a decrease in plant size or slower growth (Fig. 2). This effect was not significant in each experiment analysed individually, but was significant when data from both experiments were pooled.

Substrate weight was affected by fish presence (Linear regression on log weights, F_1,79 = 4.7, P = 0.03). Dry weight of sediment remaining in the tray at the end of the experiment was less when fish were present (Fig. 3). Dry weight of detritus at the end of the experiment was also less with fish presence.

Fig. 2 Effects of treatments on macrophytes in the two experiments. (A) Changes in size of water lilies over the course of the experiments. (B) Charophytes remaining at the end of the experiments. Results from the two experiments are pooled, as the outcome was the same in both cases. (Note that in the presence of fish, remaining Charophyte dry weight was always 0). n.s. = Not Significant, * = P < 0.05, ** = P < 0.01.

Fig. 3 Sediment remaining at the end of the experiments. n.s. = Not Significant, * = P < 0.05, ** = P < 0.01.

3.2 Invertebrates

Chironomidae were by far the most abundant invertebrates recorded (Tab. S1), and most individuals belonged to species in the genus Chironomus. Other insect taxa were recorded too infrequently for analysis on their own, but we could analyse abundance and diversity of all non-Dipteran insects pooled (a mixture of Ephemeroptera, Trichoptera, Odonata and Hemiptera). Odonata accounted for 61% of this abundance, Hemiptera 20%, Ephemeroptera 16% and Trichoptera 3%. Abundance of non-Dipteran insects was greater in experiment two, while Chironomid abundance was roughly equal between the two experiments.

Chironomidae abundance was unaffected by either tilapia or depth (Tab. 1, Fig. 4a), while diversity was negatively affected by depth but not tilapia. This reduction in diversity was caused by an increased dominance of the genus Chironomus, which accounted for 95% of specimens in the turbid treatments, versus 77% in the surface treatments. Both abundance and diversity of non-Dipteran insects were affected by depth but neither were affected by tilapia (Tab. 1, Fig. 4b). All groups were less common in the deep treatments so that often only one specimen was found in each treatment, hence the lower diversity score. Substrate type had an effect on abundance of both groups: Chironomidae were more abundant on detritus and Charophytes than on control sediment (Tab. 1, Fig. 4a), and non-Dipteran insects were more abundant on Charophytes than on detritus and control sediment (Tab. 1, Fig. 4b). The effects of fish presence, depth and substrate type and significance levels were similar in experiments one and two, suggesting that fish stocking density does not influence their effect on invertebrates.

Fig. 4 Insect abundance at the end of the experiments. (A) Chironomidae abundance (all genera) with shallow and deep samples combined, showing results from experiment 1 (high fish stocking density) and experiment 2 (low fish stocking density). (B) Non-Dipteran insect abundance (all genera), showing only the surface treatments in experiment 2.

4 Discussion

Quantitative assessments of the effects of local stressors on ecological responses are rare in tropical systems (Jackson et al., 2016, Williams-Subiza and Epele, 2021). While the experiments reported here were short term and small scale, the results of this type of experiment can help us understand the factors causing changes to the ecosystem and limiting factors on restoration (Ratajczak et al., 2018). Our results identified strong effects of turbidity and introduced tilapia on a lake system in Madagascar.

As water chemistry is broadly similar between the shallow and deep treatments, and the transplanted Charophytes were taken from deeper water than the deep treatments so depth is not directly having an adverse effect, we conclude that the depth effect in our experiments is due to reduced light levels due to the turbidity level in the lake. Both turbidity and tilapia had an impact on macrophytes in the lake. Turbidity affected invertebrates, lowering abundance and diversity of non-Dipteran insects and reducing diversity of Chironomidae, but we recorded no effect of tilapia on invertebrates. The strongest effect on invertebrate abundance was from substrate type, with the lowest abundance on the control sediment, so our results suggest that this sediment is poor invertebrate habitat. The mechanism by which tilapia affect macrophytes appears to be direct, herbivory, and leads to an indirect effect of fish on invertebrates, by consuming macrophytes that are beneficial habitat. This may be an oversimplification, as tilapia diet may vary with age and resource availability (Rao et al., 2015), so direct effects on invertebrates may occur with different age tilapia.

Our results suggest that tilapia are suspending sediment and so potentially increasing turbidity in the lake, and thus indirectly affecting invertebrates (and Charophytes) in another way. This is surprising for a species not noted as being benthivorous, but this effect of tilapia on sediment has also been reported by Attayde et al. (2007) and Zhang et al. (2017). Overall, our results indicate complex effects of a single introduced fish species on a lake ecosystem. Many lakes in Madagascar have both tilapia and common carp, plus several other species too (Sparks and Stiassny, 2003, Bamford et al., 2017, Lammers et al., 2020), and the effects resulting from these interactions may be difficult to predict.

There are reasons to be cautious interpreting our results on invertebrates. First, it is not ideal grouping all non-Dipteran insects, as different groups may respond differently − for example in small lakes in Europe, fish presence is negatively associated with Coleoptera, and positively associated with Odonata, Hemiptera and Trichoptera (Hinden et al., 2005, Hassall et al., 2011). In contrast, our results suggest no effect on Odonata and a negative effect on Hemiptera. Second, the lake may lack invertebrates that are sensitive to the stressors present in the lake (including tilapia presence), so they could not colonise our experiments. Third, swimming insects may transiently visit our experimental set up, unlike more sedentary Chironomidae, so our experimental design may underestimate numbers of these groups. These factors mean that tilapia effects on insects may be underestimated in our results.

Turbidity in Malagasy lakes is assumed to be increasing due to increased sediment run-off caused by deforestation (e.g. Máiz-Tomé et al., 2018). This may be an oversimplification, as we have already noted a role for introduced fish and burning of grasslands may contribute to turbidity (Brosens et al., 2022). The extent to which increased sediment loading and eutrophication are related is unclear at Lake Sofia. Catchment soil contains low levels of organic material, so sediment run-off may not contribute to nutrient levels in the lake. However, increased sediment loading alone can affect lakes, including declines in submerged macrophytes and benthic invertebrates, a reduction in plankton abundance and reduced primary productivity (Donohue and Mollinos, 2009). In theory, the surrounding papyrus marsh should filter inflowing water and remove suspended solids. Turbidity is lower inside the papyrus than outside at Lake Sofia, indicating that the marsh is preventing some allochthonous inputs. The width of the marsh is critical to filtration, however (Boar, 2006), and the marsh at Lake Sofia is thin on numerous small inflows. A critical need for lake management, therefore, is to understand where suspended solids are coming from, and to determine if there is a link between sediment input and eutrophication.

The major role of herbivory in shaping aquatic ecosystems has only recently become apparent and it remains understudied in tropical regions (Bakker et al., 2016, Wood et al., 2017, Vejříková et al., 2018). Our results suggest that Oreochromis fish are capable of supressing both Charophytes and water lilies. Anecdotal evidence from a nearby lake to which tilapia were introduced in 2018 and which rapidly lost all submerged Charophytes (but not yet all water lilies) suggests that tilapia are easily capable of clearing a lake of macrophytes. Water lilies were abundant at Lake Sofia in historical photographs, and herbivory by tilapia is a plausible cause of their removal and an impediment to their re-establishment. This hypothesis tallies with local reports, which say that fish were introduced around 30 years ago and consumed the lilies. Fish in particular can exert a strong effect on macrophytes (Wood et al., 2017), due to their inability to move to new sites as resources are depleted. These results have implications for lake management. A coalition of conservations NGOs and local associations aims to make Lake Sofia more suitable for the critically endangered Madagascar pochard Aythya innotata, which would involve re-establishing macrophyte populations and increasing invertebrate abundance. However, even with reduced turbidity, reintroduction of submerged and floating macrophytes will be impeded by tilapia in the lake (see Liu et al., 2018). Our results confirm results from temperate systems that invertebrates can be boosted by addition of coarse debris, but this is often only viable in the short term (Thompson et al., 2018).

Oreochromis species are widely introduced in the tropics (Canonico et al., 2005). These fish are valued for their ability to thrive in poor quality water and thus provide protein in impoverished regions. However, little consideration is given to the ecological effects of these introductions (Cucherousset and Olden, 2011) even though there are well-known negative effects (Canonico et al., 2005), in particular their role in outcompeting native species. However, other negative effects of introducing tilapia have not been studied in-depth. This includes their role in increasing turbidity and their role as herbivores (but see McCrary et al., 2001; Doupé et al., 2010), both demonstrated here.

Supplementary Material

Table S1. Supplementary Table 1. Counts of all macroinvertebrate taxa recorded during the experiments.

Access here

Acknowledgements

Fieldwork in such a difficult to reach location takes the help of a large number of people. Particular thanks must go to Jocelyn Rafaly, Randrianarimangason Floriot and all the other WWT and Durrell staff at Lake Sofia. Kevin Wood helped with information on aquatic herbivory and Lucy Smith helped tracked down literature on tilapia. Geoff Hilton and Viv Jones have been supportive of this work. We thank an anonymous reviewer whose comments substantially improved the manuscript. This research was funded by a British Ecological Society research grant awarded to AJB (ref: SR21\100726) and we thank the BES for their flexible approach to conducting research during the Covid-19 pandemic.

References

All Tables

All Figures

Fig. 1 (A) The experimental set up, a split-plot design with fish introduced into 3 of the 6 cages, and three treatments in each cage near the surface (S) and the same three treatments at 1.5 m depth (B), plus a Nymphaeaceae plant transplanted into each cage. Treatments were assigned positions at random for each cage. The three treatments were: lake sediment, woody detritus, and transplanted Charophyte plants. The experiment was run twice with different fish stocking densities. (B) A side on view of one cage to show the differing depths. (C) A photograph of the cages in situ, February 2020.

In the text

Fig. 2 Effects of treatments on macrophytes in the two experiments. (A) Changes in size of water lilies over the course of the experiments. (B) Charophytes remaining at the end of the experiments. Results from the two experiments are pooled, as the outcome was the same in both cases. (Note that in the presence of fish, remaining Charophyte dry weight was always 0). n.s. = Not Significant, * = P < 0.05, ** = P < 0.01.

In the text

	Fig. 3 Sediment remaining at the end of the experiments. n.s. = Not Significant, * = P < 0.05, ** = P < 0.01.
In the text

	Fig. 4 Insect abundance at the end of the experiments. (A) Chironomidae abundance (all genera) with shallow and deep samples combined, showing results from experiment 1 (high fish stocking density) and experiment 2 (low fish stocking density). (B) Non-Dipteran insect abundance (all genera), showing only the surface treatments in experiment 2.
In the text