The effects of land use disturbance vary with trophic position in littoral cichlid fish communities from Lake Tanganyika

1. Impacts of anthropogenic disturbance are especially severe in freshwater ecosystems. In particular, land use disturbance can lead to increased levels of pollution, including elevated nutrient and sediment loads whose negative impacts range from the community to the individual level. However, few studies have investigated if these impacts are uniform across species represented by multiple trophic levels. To address this knowledge gap, we focused on Lake Tanganyika cichlid fishes, which comprise hundreds of species representing a wide range of feeding strategies. Cichlids are at their most diverse within the near‐shore environment; however, land use disturbance of this environment has led to decreasing diversity, particularly in herbivores. We therefore tested if there is a uniform effect of pollution across species and trophic groups within the hyper‐diverse rocky shore cichlid fish community. 
 2. We selected three sites with differing levels of human impact along the Tanzanian coastline and 10 cichlid species, comprising varying taxonomic and trophic groups, common to these sites. Nitrogen and carbon stable isotope values for 528 samples were generated and analysed using generalised linear mixed models. We also estimated stomach contents including sediment proportions. 
 3. Our study highlights that multiple sources of pollution are having differing effects across species within a diverse fish community. We found that nitrogen stable isotope values were significantly higher at the most disturbed (urbanised) site for benthic feeding species, whereas there was no difference in these isotopes between sites for the water column feeding trophic group. Stomach contents revealed that the elevated δ15N values were unlikely to have been caused by differences in diet between sites. However, at the most disturbed site, higher proportions of sediment were present in most herbivores, irrespective of foraging behaviour. 
 4. It is likely that anthropogenic nitrogen loading is the cause of higher nitrogen stable isotope values since there was no evidence of species shifting trophic levels between sites. Results support our previous study showing herbivore species to be most affected by human disturbance and make the link to pollution much more explicit. As lower diversity of consumers can negatively affect ecosystem processes such as stability, alleviating environmental impact through sewage treatment and afforestation programmes should continue to be a global priority for the conservation of aquatic ecosystems, as well as human health.


| INTRODUC TI ON
The impacts of anthropogenic disturbance are especially severe in freshwater ecosystems because they are subject to a variety of anthropogenic stressors (Søndergaard & Jeppesen, 2007), which, combined with their disproportionately high diversity, results in these ecosystems being some of the most endangered in the world (Dudgeon et al., 2006;Magurran, 2009). In particular, changes in land use can lead to elevated nutrient and sediment loads in some freshwater systems (Saunders, Meeuwig, & Vincent, 2002), with both stressors implicated as threats to freshwater species (Richter, Braun, Mendelson, & Master, 1997).
Anthropogenic impacts have led to higher levels of nitrogen deposition into aquatic ecosystems in the form of domestic, industrial, and agricultural waste products (Camargo & Alonso, 2006). Excess nitrogen is known to have many negative effects on the individual fitness of organisms, including ammonia, nitrite, and nitrate toxicity, in addition to habitat level effects such as eutrophication (Camargo & Alonso, 2006). Anthropogenic nitrogen in the form of sewage has been identified in organisms using nitrogen stable isotope analysis (Fry, 1999), with areas subjected to treated wastewater effluent, and untreated sewage contamination, having higher δ 15 N values (Anderson & Cabana, 2005;Cabana & Rasmussen, 1996;Kelly, Mtiti, McIntyre, & Vadeboncoeur, 2016;Schlacher, Mondon, & Connolly, 2007;Vermeulen, Sturaro, Gobert, Bouquegneau, & Lepoint, 2011). Consequently, δ 15 N values have been proposed as an indicator for detecting anthropogenic nitrogen in aquatic ecosystems (Costanzo, O'Donohue, Dennison, Loneragan, & Thomas, 2001;Lake et al., 2001;Vermeulen et al., 2011). Environmental stress can also increase individual variation in δ 15 N values in invertebrates in controlled conditions, possibly because of more variable growth rates (Gorokhova, 2018). One of the advantages of measuring δ 15 N in fish muscle is that it averages nutrient flux over a period of several months (Trueman, McGill, & Guyard, 2005), compared to measuring anthropogenic nitrogen levels in the water, which can be temporally more variable and therefore may not show any significant differences between polluted and non-polluted sites (Kelly et al., 2016).
Deforestation of near shore habitats and subsequent soil erosion can lead to increased rates of sedimentation and is one of the main threats to aquatic ecosystems (Dudgeon et al., 2006).
Increased sediment influx can negatively impact freshwater habitats by altering light, oxygen, and temperature in the water column (Donohue, Verheyen, & Irvine, 2003) as well as smothering substrata and reducing the nutritional value of periphyton (Graham, 1990).
Pollutants such as pesticides and trace metals, as well as being absorbed directly by primary producers and magnifying up the food chain (Gersberg, Elkins, Lyon, & Goldman, 1986), can also be concentrated in sediment (Donohue & Garcia Molinos, 2009), and assimilated through indirect sediment ingestion (Eggleton & Thomas, 2004). Aquatic organisms ingesting sediment have a higher likelihood of physiological and behavioural defects, as well as increased risk of mortality (Donohue & Garcia Molinos, 2009), and by covering breeding and feeding grounds sediment can reduce habitat heterogeneity (Passy & Blanchet, 2007), resulting in more homogenous assemblages (Balata, Piazzi, & Benedetti-Cecchi, 2007).

| Pollution of a global aquatic hotspot
One of the world's most diverse freshwater ecosystems is East Africa's Lake Tanganyika (LT) with c. 1,470 animal species, which is renowned for its impressive adaptive radiations of cichlids fishes (e.g. Day, Cotton, & Barraclough, 2008). Although the least disturbed of the great lakes (Dobiesz et al., 2010), there have been major increases in anthropogenic stressors around its shores, and the lake is especially vulnerable to pollution due to the slow rate of water renewal in its effectively closed system (Coulter & Mubamba, 1993). This is particularly problematic in near shore urban areas where incorrect treatment and disposal of domestic, agricultural and industrial waste is prevalent (Chale, 2003;Kelly et al., 2016). Increased erosion and sediment transport is also a major threat to the LT ecosystem, primarily caused by deforestation of the lake shore's miombo woodland (Alin et al., 1999;Cohen, Bills, Cocquyt, & Caljon, 1993;Coulter & Mubamba, 1993), particularly in areas with steep rift basins such as around the Kigoma region (Alin et al., 2002;Cohen, Kaufman, & Ogutu-Ohwayo, 1996). Increased sediment loads have been shown to affect the community dynamics of LT benthic invertebrate and fish communities by decreasing species richness and abundance (Donohue et al., 2003); however, the uptake of sediment by cichlids in this region has to our knowledge not been investigated.
A recent study showed the multi-faceted effect of human disturbance on the cichlids with a clear decline in α-diversity with increasing human disturbance, especially among herbivorous species within the Tropheini be most affected by human disturbance and make the link to pollution much more explicit. As lower diversity of consumers can negatively affect ecosystem processes such as stability, alleviating environmental impact through sewage treatment and afforestation programmes should continue to be a global priority for the conservation of aquatic ecosystems, as well as human health.

K E Y W O R D S
cichlids, nitrogen loading, sedimentation, stable isotopes, stomach contents tribe. By contrast, there was a positive relationship between (β-diversity) nestedness across surveys and human disturbance implying rare or specialist species were being lost, and the functional form of ζ-diversity (the expected number of species common to n-surveys) was found to be qualitatively different between disturbed and protected sites indicating stochastic processes dominate in the former, but niche processes dominate in the latter (Britton et al., 2017). These results hint that species are being differentially affected by human disturbance, perhaps based upon their feeding behaviour, but how this is occurring, and whether some species are escaping disturbance effects due to diet changes remains an open question.
Previous LT studies have investigated the effect of anthropogenic loading on the δ 15 N values of sediment, and across a variety of taxonomic groups, with mixed results. Alin et al. (2002) reported higher sediment rates, and higher δ 15 N values of sedimentary organic matter from a disturbed site compared to a nearby National Park (Gombe) in the Kigoma region, probably due to increasing inputs of terrestrial organic matter from shoreline erosion through deforestation. Elevated δ 15 N values in gastropods from village shorelines from this region were attributed to anthropogenic nitrogen loading from human waste (Kelly et al., 2016), although these authors found no difference in nutrient concentrations at these sites compared to reference sites. It is likely that nutrients in LT are quickly sequestered by phytoplankton and periphyton (Corman et al., 2010;McIntyre, Michel, & Olsgard, 2006); however, due to the open nature of the littoral zone, phytoplankton are rapidly washed away. In contrast, diversity and δ 13 C and δ 15 N values of crab species from Kigoma were similar at sediment impacted and reference sites (Marijnissen, Michel, Cleary, & McIntyre, 2009) with dietary breadth potentially contributing to their resilience to sedimentation, based on the wide range of δ 13 C values from both sites. This study suggested that some species may adapt their feeding behaviour to negate the effects of environmental changes, or that they may be immune to the effects of pollution because of a broad diet. However, it is hard to know whether these mixed results, associated with increased sedimentation, are due to differences in the taxonomic groups or differences in the feeding behaviour. Here, we focus on the cichlid fish community, which is a useful study group to test out these possibilities as they contain a wide range of feeding strategies across multiple trophic levels.

| Aims and expectations
To answer whether changes in water quality have a uniform effect on the LT cichlid fish community, we characterised the stable isotope signatures of rocky shore littoral cichlids from sites with differing levels of anthropogenic disturbance. Firstly, we asked whether nitrogen stable isotope values differ between urban and non-urban areas, and which taxonomic and trophic groups are most affected. Because high stable nitrogen signatures are used as indicators of anthropogenic impact (Vermeulen et al., 2011), we predicted elevated δ 15 N values and variance in species at urban sites compared to non-urban sites. We expected all cichlids, but particularly benthic herbivores to have higher δ 15 N values, in part because their diversity has recently been shown to be more affected by human disturbance within this region than the other feeding groups (Britton et al., 2017). Secondly, we investigated possible biological mechanisms causing changes in nitrogen stable isotope values using stomach content analysis. We expected to rule out dietary changes as the cause of higher δ 15 N values, and instead predicted anthropogenic nitrogen input (Vermeulen et al., 2011) to be the principal driver. Based on the higher sedimentation at urban sites, we also predicted higher sediment content in the stomachs of fishes at the urban site compared to the non-urban site.

| Study system
Lake Tanganyika's littoral and sublittoral zones contain highly diverse communities of endemic fishes, molluscs and crustaceans (Coulter, 1991;Day et al., 2008). Of these, cichlid fishes are the most diverse animal group (200 species) and dominate the rocky littoral zone (c. 65% of all cichlid species) (Konings, 2015). Upwards of 60 species have been recorded at some rocky-shore locations (Britton et al., 2017) with communities comprising a vast range of feeding strategies and taxonomic groups (Konings, 2015). Based on these attributes, and that rocky-shore cichlids have previously been shown to be affected by human impacts (e.g. Alin et al., 2002;Britton et al., 2017) they were therefore selected as our focal system.

| Study sites
We focused on three sites, TAFIRI Bay, Kigoma Deforested, and Kalilani Island, in the Kigoma region of Tanzania ( Figure 1). These sites were situated within localities surveyed by Britton et al. (2017), and in the absence of baseline stable isotope data before anthropogenic impact (Rowell, Dettman, & Dietz, 2010), were selected due to their differing levels of human disturbance ranking (Britton et al., 2017). An urban site, TAFIRI Bay, was classified as the most disturbed site, followed by the nearby uninhabited but non-urban Kigoma Deforested site, with the more distant largely forested site of Kalilani Island being the least disturbed (Table 1). TAFIRI Bay is located in the south of Kigoma Bay, on the shores of Kigoma Town, a large urban area with a human population of 215,458 (Tanzania National Bureau of Statistics, 2013), and a population density of over 32 people per 100 m 2 (Linard, Gilbert, Snow, Noor, & Tatem, 2012). TAFIRI Bay is responsible for the town's water supply, but is also a repository for untreated domestic and industrial waste (Chale, 2003). The area surrounding TAFIRI Bay has been developed considerably resulting in a population density of 56 people per 100 m 2 (Linard et al., 2012), and a reduction in tree cover to <10% canopy density (Hansen et al., 2013). Two sampling points within TAFIRI Bay were selected, (a) 4°87.879′S, 29°62.169′E in the northern bay; (b) 4°88.652′S, 29°61.566′E, located 900 m south west, in the southern bay.
The Kigoma Deforested site (4°90.216′S, 29°59.472′E) is located south of TAFIRI Bay, separated by a 2-km stretch of deforested and uninhabited shoreline. Kigoma Deforested has a tree canopy density of approximately 10% (Hansen et al., 2013), and is uninhabited. The Jakobsen's Beach reserve is 100 m south of this site, encompassing 1 km of shoreline with 16% canopy density (Britton et al., 2017;Hansen et al., 2013).
Kalilani Island (6°02.023′S, 29°74.243′E) is a small 1,000 m 2 uninhabited island with c. 46% tree canopy density (Hansen et al., 2013), located 125 km south of Kigoma Town. It is situated approximately 300 m north of the border of Mahale National Park, a pristine area that protects 1,613 km 2 of lake shore forest (Sweke, Assam, Matsuishi, & Chande, 2013) and 96 km 2 of the lakes aquatic littoral habitat (West, 2001). It is also situated 200 m west of Kalilani village, a small fishing village covering 2 km of shoreline. Kalilani Village has a tree cover of approximately 25% canopy density (Britton et al., 2017;Hansen et al., 2013), and a population of fewer than three people per 100 m 2 (Britton et al., 2017;Linard et al., 2012). To ensure a range of taxonomic and trophic groups were represented, species encompassing benthic herbivores and invertivores, and water column planktivores and piscivores were targeted. Samples were collected underwater by SCUBA and snorkel at depths of 0-15 m in the rocky littoral zone. Cichlids were caught using a seine net with each species targeted separately. Sampling was random, with several collections made all within a radius of c. 100 m of the GPS coordinates of the anchored boat. Only adult fish were collected, although nitrogen isotopes have been shown to be independent of age in fish and mollusc species (Hobson & Welch, 1995;Kiriluk, Servos, Whittle, Cabana, & Rasmussen, 1995;Minagawa & Wada, 1984). Similar sized individuals were targeted since total individual length has been positively linked to δ 13 C, and to a lesser extent  (Genner, Hawkins, & Turner, 2003). We were only able to determine sex for one cichlid species-Ophthalmotilapia ventralis -in the field, and therefore our sampling was indiscriminate. However, sex has not been found to influence stable isotope values in haplochromine cichlids (Genner, Turner, Barker, & Hawkins, 1999).

| Sampling
Upon collection, fish were immediately euthanised with an overdose of clove oil (Neiffer & Stamper, 2009) and preserved in 80% ethanol. Baseline invertebrate species were collected by hand from rocks and benthic substrate, and algae were scraped from multiple (>3) rocks at each site resulting in one multispecies algae sample per site. As comprehensive sample processing could not be conducted on site due to logistical constraints, samples were preserved in ethanol for 2 months before being processed in the laboratory. Therefore, 40 cichlid individuals (including at least one individual of each of the 10 species) across all sites had an additional sample sun dried with the aid of a desiccant (silica gel), and used as controls to allow correction for the effect of ethanol preservation (Correa, 2012).

| Stable isotope analysis
White muscle tissue from the left dorsal flank of each fish was oven dried at 55°C for 48 hr, along with muscle tissue of mollusc baseline samples and the multispecies algae samples. A subset of the algae samples at each site were acid washed after drying, because of the presence of inorganic carbon in the sediment, and were used to obtain the algae δ 13 C values (Schlacher & Connolly, 2014). Samples were homogenised to a powder using a pestle and mortar for consistency.

| Corrections for lipid content and tissue preservation
Due to carbon isotope fractionation during lipid synthesis, a lipid normalisation was applied to the δ 13 C values (Kiljunen et al., 2006). A revised model (δ 13 C′ = δ 13 C + D -[I + 3.9/1 + 287/lipid proportion]) modified from McConnaughey and McRoy (1979) with updated parameters applicable to freshwater fish was used for the correction (Kiljunen et al., 2006). The parameters, have been used in one other study of cichlid stable isotope values (Ford et al., 2016). A lipid extraction was not performed because it can change the nitrogen isotopes in an unpredictable way (Kiljunen et al., 2006), and δ 15 N values were not modified because there is very little nitrogen in lipids. The baseline species were also not lipid corrected.
Ethanol preservation can affect δ 13 C and δ 15 N values (Correa, 2012). When the δ 13 C values of sun-dried control samples were pooled between the four sites, there was a significant difference with ethanol preserved specimen δ 13 C values, for both raw and lipid-normalised values (Supporting Information Table S3). The pooled, lipid-corrected, ethanol-preserved δ 13 C values and lipid-corrected, air-dried δ 13 C values were plotted and a linear correction was applied to all cichlid ethanol-preserved δ 13 C values using the data fitted equation: δ 13 C corrected = 1.0387 × δ 13 C ethanol + 0.3758 (Bicknell et al., 2011;Bugoni, McGill, & Furness, 2008;Kelly, Dempson, & Power, 2006).
However, it has been reported that δ 13 C values decrease by over 1‰ in fish muscle tissue preserved in ethanol for 6 weeks, and longer preservation might increase variation (Arrington & Winemiller, 2002).
Therefore, the corrections applied to δ 13 C mean fine scale differences in stable isotope values (<2‰) will not be used to make ecological inferences in this study, although δ 13 C values can be used to differentiate between larger scale differences in habitats (e.g. Piola, Moore, & Suthers, 2006 (Woodcock et al., 2012), and the large variation in some baseline species δ 15 N values between sites prevented a reliable normalisation (Supporting Information Table S4).

| Stomach content analysis
To identify areas with higher rates of sedimentation and diet shifts, stomach content analysis was used to allow quantification of diet (e.g. Davis, Blanchette, Pusey, Jardine, & Pearson, 2012;Malins et al., 1985Malins et al., , 1987, although we note that it provides only a temporal snapshot of each individual's intake (Polito et al., 2011;Wagner, McIntyre, Buels, Gilbert, & Michel, 2009 upon removal meant that only stomachs were analysed. As contents were too small to weigh, a modified version of the points method of (Hynes, 1950) and (Hyslop, 1980) was used (Genner et al., 1999). Items were sorted and split into the broad categories: (1) sediment; (2) algae; (3) fish scales; (4) insects; (5) crustaceans; (6) gastropods; (7) fish; (8) plankton. Fish scales were included as a separate category from fish as they are reported within the stomachs of non-piscivorous cichlids species, and inferred to be ingested through aggressive territorial behaviour (Kohda, 1995). The categories were allocated points based on their proportional value.
The category with the largest volume was given 16 points and, if other categories were present, they were sequentially awarded 8, 4, 2, 1 or 0 points, in descending order of volume relative to the most abundant category (Genner et al., 1999). Total points were counted and the volume contribution of each category was calculated as a percentage, and averaged for the species at each site.
Pairwise comparison of dietary composition between each species at both study sites were performed using Schoener's dietary overlap index (Schoener, 1970)

| Nitrogen stable isotopes
All but one benthic feeding cichlid species, as well as the water column feeding cichlid piscivore Lepidiolamprologus elongatus, display significantly different mean δ 15 N values between the urban site of TAFIRI Bay and the less disturbed sites (i.e. the non-urban sites of Kigoma Deforested, and the forested Kalilani Island; Figure 2a and Table 2). In contrast, the water column feeding planktivore  (Table 3).
The difference in δ 15 N at urban and non-urban sites reveals a similar pattern for the baseline species, with benthic feeders more affected than filter feeders (Supporting Information Table S4).  Additionally, the differences between baseline and cichlid δ 15 N values ranged between 4‰ and 9‰ within trophic groups per site.

| Carbon stable isotopes
There are clear differences between the mean δ 13 C values of benthic and water column species (Figure 2b). All trophic group δ 13 C values are significantly different at TAFIRI Bay in 2015 compared to less disturbed non-urban sites, whereas there is no difference in trophic group δ 13 C values at TAFIRI Bay in 2016 compared to the non-urban sites (Table 3). In terms of the species level analysis, there are no consistent differences in mean δ 13 C values between sites ( Table 2), suggesting that, unlike nitrogen isotopes, differences in carbon isotopes are not linked to human disturbance.
Algae δ 13 C values range between approximately −10‰ and −15‰ δ 13 C, which is reflected in similar values for algivorous molluscs, the gastropods Lavigeria grandis and Lavigeria nassa, and herbivorous cichlid Tropheus brichardi (Figure 3a, c). In detritus-feeding mollusc Neothauma tanganyicense (Viviparidae), filter-feeding mollusc Pleiodon spekii (Iridinidae), and water column-feeding cichlid N. brichardi, δ 13 C values range between approximately −20 and −25‰ (Figure 3b, d). The δ 13 C values differ slightly within sites for the benthic cichlid and baseline species but as the differences are not consistently > 2‰ they were not considered large enough to make ecological inferences (Arrington & Winemiller, 2002).  Table S8). All cichlid species showed considerable dietary overlap between sites, apart from the benthic-feeding Neolamprologus toae and water column-feeding L. elongatus and N. brichardi. However, as the latter two species had the smallest differences in δ 15 N values, we can assume that diet did not play a role in the increased δ 15 N values at TAFIRI Bay. When these differences are visualised, multidimensional scaling shows clustering of herbivores and clustering of non-herbivores (Supporting Information Figure S1), however within the herbivores the species cluster by site.

In terms of individual dietary components (Supporting Information
Tables S6 and S7) a significantly higher proportion of sediment was found in the stomachs of four cichlid species at TAFIRI Bay compared to Kalilani Island in 2015 (Supporting Information Table S8). These species included three of the four herbivores, encompassing multiple behavioural adaptations for harvesting algae: Eretmodus cyanostictus (scraper), Petrochromis famula (grazer), T. brichardi (browser) as well as the planktonic column feeding N. brichardi (Supporting Information   Table S8). Pseudosimochromis babaulti (browser) had a higher proportion of sediment in its stomach at Kalilani Island than the other herbivores, probably because it often feeds in sediment-rich areas of the rocky shore (Koblmüller, Egger, Sturmbauer, & Sefc, 2010). We suggest that it is the extra stomach sediment in the urban site that leads to the herbivores clustering out according to site (Supporting Information Figure S1). We also found significant differences in stomach contents between the species (ANOSIM, R = 0.547, p = 0.01), but not between sites (ANOSIM, R = 0.001, p = 0.064) when the non-organic sediment category was removed from the ANOSIM.

| D ISCUSS I ON
Nitrogen stable isotopes have been shown to be a highly sensitive tool for monitoring anthropogenic allochthonous sources of nitrogen in freshwater and marine ecosystems (Anderson & Cabana, 2005;Kelly et al., 2016;Vermeulen et al., 2011), as well as an indicator of stress (Gorokhova, 2018). We utilised this method to investigate the effect of land-use disturbance in a species rich lacustrine tropical fish community and revealed significantly higher and more variable δ 15 N values in individuals sampled from an urban area than those at non-urban areas (Figures 2 and 3, and Tables 2 and 3). Notably, this finding is not uniform across species examined from the urbanised site as we showed that benthic feeding species, particularly herbivores, were more affected than species feeding in the water column (Figures 2 and 3, and Tables 2   and 3). Nitrogen isotope values for benthic feeders at the urban site are far higher than those for L. elongatus, which as a carnivore would hold a higher trophic position in a typical food web of this species rich community. The increase in nitrogen stable isotopes appears to be robust because it is large, consistent across years, and reflects long-term assimilation. These results support a previous study of freshwater habitats in North America where more variable δ 15 N values were reported in benthic feeding fish species than pelagic species (Lake et al., 2001).
Cichlid fishes have been shown to change their food source in response to ecological pressures, as reported in Lake Victoria cichlids responding to increased predation (Katunzi, Zoutendijk, Goldschmidt, Wanink, & Witte, 2003), and changing resource availability (Njiru, Okeyo-Owuor, Muchiri, & Cowx, 2004) Urban areas with higher human populations are subjected to inputs of anthropogenic waste nitrogen (Camargo & Alonso, 2006), and primary producers in these environments incorporate human sewage with elevated δ 15 N (Vermeulen et al., 2011). The higher δ 15 N values that we identify in the benthic food web at our urban site (TAFIRI Bay), and absence of a trophic level dietary shift in benthic feeding cichlid species, indicate that algae is the likely source of the elevated nitrogen stable isotopes. As reported in other lacustrine systems benthic algae absorb increased anthropogenic nitrogen input, and δ 15 N is subsequently biomagnified up the food chain (Cabana & Rasmussen, 1996). We suggest that the elevated δ 15 N is from anthropogenic nitrogen loading, and probably reflects the high δ 15 N of human sewage, as reported in other studies focused on a variety of aquatic systems (Cabana & Rasmussen, 1996;Schlacher, Liddell, Gaston, & Schlacher-Hoenlinger, 2005;Vermeulen et al., 2011). While subsistence agriculture is practiced along the shores of LT close to villages (Kelly et al., 2016), only our urbanised site identified elevated δ 15 N values, and by sampling during non-wet periods we also accounted for substantial nutrient runoff. Other pathways, such as fishery inputs, which are common practice in the focal region (i.e. fish processing on beaches) could also affect δ 15 N in aquatic systems; however, we did not encounter this activity at our urban site (Britton and Doble, personal observation We also showed that herbivorous cichlids, in an area of high human disturbance and with reported higher sedimentation rates (Marijnissen et al., 2009;McIntyre et al., 2005), have higher proportions of sediment in their stomachs, irrespective of foraging behaviour, than at a low human disturbance site, demonstrating that species at this trophic level are also particularly sensitive to high sediment pollution. Previous studies have shown that cichlid diversity decreases with increasing human disturbance (Britton et al., 2017;Cohen et al., 1993;Sweke et al., 2013).
In particular, the results presented here are in broad agreement with Britton et al. (2017) who identified that α-diversity and abundance of benthic species, particularly herbivores that are members of the Tropheini, were more greatly affected than other trophic guilds or tribes.
This highlights a potential causal link between the extent of human disturbance and the change in community diversity of the cichlids, but clearly more work is required to establish which aspects of life-history (survival, reproduction, growth) are being most affected. Sediment pollution suspended in the water column can cause negative health impacts in fish such as gill clogging (Bruton, 1985), in which resultant gill hypertrophy has been linked to decreased growth rate, possibly from respiratory impairment (Sutherland & Meyer, 2007). Environmental stress can place a limit on the energy available for growth (Smolders, Bervoets, De Coen, & Blust, 2004), and we found some evidence for this as two herbivorous species (T. brichardi and P. famula) were smaller at the urban disturbed site than the non-urban sites (unpublished results), although a detailed study is needed to test this.
Results from our study also raise questions regarding the accu-  Muschick et al., 2012;Wagner et al., 2009), this study demonstrates that care should be taken when associating isotopic niche to ecological niche (Jackson, Inger, Parnell, & Bearhop, 2011). Unfortunately, little is known about the effects of spatial variation in nitrogen loading since the influence of a local source of nitrogen on δ 15 N values will be the result of both physical (water movement) and biological (movement of individuals) factors. Therefore, systematic spatial sampling, sufficient intra-specific sampling (c. 15), and adequate baseline sampling are recommended as good practice to prevent biasing results.
Our findings, combined with the considerably lower diversity of herbivores previously identified at Kigoma Bay (Britton et al., 2017), suggest that habitat degradation through deforestation causing sedimentation, and water pollution including nitrogen deposition due to human disturbance, are possible causes contributing to negative changes in community composition and diversity of cichlids in this region (Britton et al., 2017). While consumer effects on prey are well known, the role of consumer diversity in affecting community structure or ecosystems is not particularly well understood, although Burkepile and Hay (2008) demonstrated that herbivorous fish species richness is critical for preserving coral reefs. Herbivores also form an important component of communities in the African Great Lakes (Hata & Ochi, 2016), and their decline may have serious implications for these systems. For example, the decrease in diversity of Lake Victoria herbivorous haplochromine cichlids after the introduction of the Nile perch (Lates niloticus) could have led to trophic cascades in this ecosystem (Goldschmidt, Witte, & Wanink, 1993).

| CON CLUS IONS
It is likely that the various forms of pollution identified are affecting the health of lake cichlids, and other fish groups, but several questions remain regarding how pollution is affecting individual fish and community structure. Elevated δ 15 N has been shown to reflect a host of negative health impacts in fish species, including a range of pathological tissue changes such as abnormalities in most major organs (e.g. Schlacher et al., 2007). As well as investigating histopathology, future studies of Great Lake cichlids could consider investigating transcriptome level changes related to increased human driven environmental stress to provide a better understanding of genes and biochemical pathways affected. Our study supports previous work on other aquatic ecosystems that anthropogenic nitrogen loading and sedimentation are major threats to aquatic biodiversity (Dudgeon et al., 2006;Gangloff, Edgar, & Wilson, 2016;Islam & Tanaka, 2004). As such, alleviating pollution through afforestation programmes (Deng, Shangguan, & Li, 2012) and the effective treatment and disposal of waste (Eggen, Hollender, Joss, Schärer, & Stamm, 2014) should continue to be a global priority for the conservation of aquatic ecosystems, as well as human health. We thank Ben Ngatunga and Semvua Mzighani (TAFIRI) and Mwita Chacha (University of Dar es Salaam) for their support of research and export permits, and George Kazumbe for fieldwork assistance.

ACK N OWLED G M ENTS
We also thank two anonymous reviewers and Martin Genner for their helpful comments on improving this manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.