Stream fish assemblage and habitat structure in a tropical African river basin (Nyagui River, Zimbabwe)




Few studies have been carried out on stream ecology in southern Africa although many species are endangered. This study investigated the stream fish assemblage and their habitat associations over a period of 3 months (October 2004 to January 2005), in view of the proposal to build a dam across the Nyagui River. Twenty-four fish species were collected and were separated into groups based on preferred microhabitats. The first group, dominated by Barbus paludinosus, comprised species collected from the upstream stations with slow flow, shallow depth (pools) and fine substrate type. Species associated with riffles, which included Chiloglanis neumanni, Labeobarbus marequensis and Opsaridium zambezense, comprised the second group on the downstream. The last group comprised species preferring pools with rock substrate and slow flow such as Pharyngochromis acuticeps and Pseudocranilabrus philander. The species were consistently associated with their habitat types throughout the sampling period. This relationship may be explained by the fish’s morphological adaptations. Species richness increased from nine in the upstream section to twenty in the downstream section and this was related to increasing habitat complexity downstream. The construction of the Kunzvi Dam across the Nyagui River is likely to lead to loss of rheophilic species while cichlids and introduced species may increase.


On a réalisé peu d’études sur l’écologie des eaux courantes en Afrique australe bien que de nombreuses espèces y soient en danger. Cette étude s’est penchée sur l’assemblage des poissons de rivière et les associations avec leurs habitats pendant quatre mois (octobre 2004–janvier 2005) en raison de la proposition de construction d’un barrage sur la rivière Nyagui. On a récolté un total de 24 espèces de poissons que l’on a classés en groupes distincts selon le micro-habitat préféré. Le premier groupe, dominé par Barbus paludinosus, comprenait des espèces récoltées des stations en amont où le courant est lent, peu profond (bassins) et le substrat de type fin. Les espèces associées aux ondulations, qui comprennent Chiloglanis neumanni, Labeobarbus marequensis et Opsaridium zambezense, composent le deuxième groupe en aval. Le dernier groupe comprenait des espèces préférant les bassins au substrat rocailleux et un courant lent: Pharyngochromis acuticeps et Pseudocranilabrus philander. Ces espèces sont restées constamment associées à leur type d’habitat pendant toute la période de prélèvement. Cette relation pourrait s’expliquer par les adaptations morphologiques des poissons. La richesse en espèce est passée de neuf dans la section située en amont à vingt dans celle de l’aval et ceci était liéà un habitat de plus en plus complexe. La construction du barrage Kunzvi sur la rivière Nyagui est susceptible d’entraîner la perte d’espèces rhéophiles alors que les cichlides et les espèces introduites pourraient proliférer.


Stream freshwater fishes are the most threatened among all classes of vertebrates worldwide (Collares-Pereira & Cowx, 2004), with about 24 species considered to be rare or endangered in southern Africa (Skelton, 1993). The central theme in stream fish community ecology is understanding the factors responsible for structuring fish assemblages in the lotic ecosystems (Godinho, Ferreira & Santos, 2000; Koel & Peterka, 2003). The structure of stream fish assemblages may be a result of stochastic factors, such as disturbance, and deterministic interactions, such as competition and predation (Lowe-Mcconnell, 1987) in addition to the physical and chemical environment (Vila-Gispert, Garcia-Berthon & Mereno-Amich, 2002). It has been argued that abiotic factors become increasingly important as determinants of stream fish community structure from downstream to upstream (Horwitz, 1978; Closs & Lake, 1996; Taylor, 2000). On the other hand, biotic factors tend to be more influential in downstream sections that are environmentally benign (Gilliam, Fraser & Alkins-Koo, 1993). A general gradient of increasing species diversity and richness from upstream to downstream has been noted in stream fishes (Bistoni & Hued, 2002; Ostrand & Wilde, 2002; Moyle & Cech, 2004).

Very little work has been undertaken on stream fish ecology in southern Africa, yet some stream fishes are already on the IUCN Red List. Before the development of any conservation strategy, it is important to identify the fish assemblages and determine causal processes underlying the observed patterns. The proposal to build Kunzvi Dam across the Nyagui River motivated this study. It was considered prudent that pre-impoundment fish studies be undertaken to mitigate adverse effects of the proposed project.

Materials and methods

Study area

The Nyagui River rises on the central plateau of Zimbabwe at an altitude of 1600 m a.s.l. near the town of Marondera and flows north to join the Mazowe River at an altitude of 840 m a.s.l. near Shamva town (Fig. 1). The river basin can be divided into three sections. Its upper section, at altitudes exceeding 1400 m a.s.l.(sampling stations 1–5), flows through extensive area of grassland and dambo wetlands. The upper section and its tributaries tend to consist of pools with sandy bottoms and grassy verges with extensive macrophyte growth (commonly Phragmites and Nymphaea spp.). The middle section (sampling stations 6–12) lies at an altitude of 1000–1400 m a.s.l and the river flows through mostly degraded woodland and its bed is rocky with occasional sandbanks. The lower section (sampling stations 13 and 14) is below 1000 m in altitude and the river is rocky. In this section, some sand banks and alluvial terraces are covered with Phragmites. The mean annual rainfall is relatively high (850 mm at Marondera). It is highly seasonal and the rivers flow strongly during the rainy season (November–April), diminishing during the course of the dry season to reach their lowest levels in October–November. By this time, the upstream sections and most of the smaller tributaries have ceased to flow although the Nyagui itself is perennial.

Figure 1.

 The Nyagui River in Zimbabwe with its major tributaries and the location of sampling stations

Data collection

All stations were sampled in October 2004 (before the rains), in November 2004 (after the first rains) and in January 2005 when the river was flowing strongly. However, station 14 was sampled only in November and January. Habitat data were collected using methods described by Gorman & Karr (1978) and Schlosser (1982). Twenty transects were set up at each station at 5-m intervals perpendicular to the stream flow. Starting from the wetted edge, the water depth (m) and velocity m s−1 (measured with an FP201 electronic current meter; Global Water Inc., Sunnyvale, CA, U.S.A.) were determined at 50-cm intervals along each transect. The dominant substrate was visually characterized within a radius of 25 cm. The number of points assessed varied from 38 at the upper stations, where the streams were smaller, to 320 at the lower stations where the streams were larger. Each habitat was classified following the study by Schlosser (1982) which identifies four depth classes, five velocity classes and five substrate classes, and two other substrate classes (vegetation and debris). The proportion (%) of each class at each station was determined. The habitat category for each point was allocated to one of the potential 170 categories based on a combination of depth, velocity and substrate (Schlosser, 1982). The proportion of each category was determined and a habitat diversity index was calculated using the Shannon diversity index (Shannon & Weaver, 1949).

Temperature (°C), dissolved oxygen (mg l−1), turbidity (NTU), conductivity (μs cm−1) and pH were determined on site using a Horiba U23 multiprobe meter (Horiba Ltd, Osaka, Japan). Fish were captured using a Smith-Root VI-A electrofisher (Smith-Root Inc., Vancouver, WA, U.S.A.) powered by a Honda EZ4500 generator (Honda Wenzl Hruby, Honda Motor Europe, Hamburg, Germany). Each station was blocked with an 8-mm mesh net to prevent fish from escaping, and fished for 10 min by moving upstream from the lower block net. All fish caught were identified, counted, weighed (g) and measured to standard length (cm). The proportional abundance of each species was determined. The relative abundance of fish was expressed as number of fish caught per minute.

Environmental variables (temperature, conductivity, dissolved oxygen and turbidity), grouped by section and month, were compared using two-way analysis of variance (ANOVA). Conductivity and turbidity were log (x + 1) transformed to stabilize variance. Fisher’s least significant difference (LSD) mean separation pairwise tests were conducted on significant ANOVAs to determine the stations and months that were different. To summarize the relationship between species and habitat variables, a canonical correspondence analysis (CCA) was performed using CANOCO 4.0 (Ter Braak & Smilauer, 1998) based on the relative abundance of species and the proportion of each habitat variable at each station. Step-wise forward selection of environmental variables, available in CANOCO, was used to select the most important habitat variables. Species richness was regressed against habitat diversity.


Physical and chemical variables

Stations in the upper section, except station 4, were dominated by silt and sand, with shallow and slow flow throughout the sampling period (Table 1). Aquatic macrophytes were relatively dense at these stations, covering at least 30% of the total area at three of the five stations sampled. Downstream stations (6–14) had a combination of riffles, with gravel and pebble substrates, and pools, with sand and rock substrate (Table 1). Flow generally increased in the downstream section from October 2004 to January 2005.

Table 1.   The habitat composition (%) at the fourteen sampling stations in the Nyagui River, October 2004 to January 2005
Altitude (m)156215581552152914021328122212731192115611911152844822
 Riparian cover7507105107585102015
 Very shallow5743472053131727201323101315
 Very slow 7033535338582383401355
 Fast 015743333020404516234335
 Torrent 0200018221118192102230

The water temperature increased with decreasing altitude (Fig. 2a) and temperature in the upstream section was significantly lower than the downstream (ANOVA, < 0.01). Temperature also varied significantly (ANOVA, < 0.01) for the 3 months, the upstream section having the highest temperature in January while the downstream sections had the highest temperature in November. While dissolved oxygen did not vary in each section for the 3 months (ANOVA, > 0.01), it was significantly lower in the upper section compared with the lower sections (2 and 3) (Fig. 2a). Conductivity increased significantly from upstream to downstream (Fig. 2b) and from October to January (ANOVA, < 0.01). Although there was no discernible pattern among the three sections for turbidity (Fig. 2b), it increased significantly from October to January (ANOVA, < 0.01). The water pH was generally neutral to slightly alkaline on the lower sections and slightly acidic to neutral on the upper section except in October (Fig. 2c).

Figure 2.

 Water quality characteristics of the Nyagui River, data are mean ± standard deviation

Fish assemblage

In all, 24 species in eight families were collected during the whole sampling period (Table 2). The most frequently occurring species (collected at ≥50% of all stations) were Chiloglanis neumanni, Clarias gariepinus, Labeo cylindricus, Labeobarbus marequensis, Micropterus salmoides, Pseudocranilabrus philander and Tilapia rendalli (Table 2). The most numerous species (constituting >10% of the total) were C. neumanni and L. marequensis. Four exotic species were collected, namely Mesobola brevianalis, M. salmoides, Oreochromis niloticus and Serranochromis robustus. Species richness was relatively low in the upstream section where nine species were collected (Table 2). The upstream section was numerically dominated by B. paludinosus and C. gariepinus, each contributing >20% of the total. Species richness increased at the stations in the middle section, with a total of 17 species being collected. The numerically dominant species in this section were C. neumanni (28% of the total number), L. marequensis (19.4%) and Barbus trimaculatus (14.7%) (Table 2). The downstream section (station 13 and 14) had the highest species richness, where twenty species were collected (Table 2). The most numerous species were L. cylindricus (16.6% of the total), C. neumanni (15.7%), T. rendalli (9.5%) and M. brevianalis (6.3%).

Table 2.   The composition (% by numbers) of the fish captured during the sampling period in the Nyagui River
>1400 m
1000–1400 m
<1000 m
  1. aIntroduced species.

MormyridaeCyphomyrus discorhynchus   0.40.1
Mormyrus longirostris  0.20.0
CyprinidaeMesobola brevianalisa  6.31.4
Opsaridium zambezense
Barbus lineomaculatus11.10.5 0.8
Barbus radiatus  2.50.7
Barbus trimaculatus
Barbus paludinosus21.60.3 2.9
Barbus unitaeniatus
Labeobarbus marequensis 19.424.017.1
Labeo cylindricus 4.416.67.7
AlestiidaeBrycinus imberi  0.80.2
Micralestes acutidens  4.60.6
AmphiliidaeZaireichthys rotundiceps
ClariidaeClarias gariepinus22.
MochokidaeChiloglanis neumanni 28.015.728.6
CentrachidaeMicropterus salmoidesa11.
CichlidaePseudocranilabrus philander6.
Pharyngochromis acuticeps
Tilapia sparmanii7.70.3 0.7
Tilapia rendalli15.
Oreochromis mossambicus1.
Oreochromis niloticusa
Serranochromis robustusa2.9  0.5

Canonical correspondence analysis showed a significant association between species relative abundance and the environmental variables (Monte Carlo permutations, < 0.05). Of the sixteen environmental variables analysed, the forward selection procedure retained six variables for the October CCA, eight variables for the November CCA and ten variables for the January CCA (Table 3). The first two axes were used in the interpretation of results for the respective ordinations. The ordination plots of species scores indicate their relationship with reduced number of environmental variables (Fig. 3a–c). The main gradient for the October ordination showed a separation between shallow and slow upstream section and the riffle-dominated downstream sections. This gradient separated the species that dominated the upstream section, especially B. paludinosus, and those dominating the downstream section. The ordination showed the relative importance of riffles on the downstream section. The species associated with the riffles included B. trimaculatus, Opsaridium zambezense, C. neumanni, L. cylindricus and L. marequensis (Fig. 3a), all collected from station 6–12. In broad terms a similar habitat-species association was obtained in November (Fig. 3b). In addition, another gradient was shown on the second CCA axis. Rock substrate, deep categories (not shown on the ordination diagram) and slow flow were the variables correlated to this axis. Species associated with this gradient included P. philander and Pharyngochromis acuticeps. Species were consistently associated with their habitat types during the period of high flow (January). The most important gradient separated the downstream sections with increasing flow from the upstream section that was generally characterized by slow flow and shallow habitats (Fig. 3c).

Table 3.   Canonical correspondence analysis (CCA) summary statistics for the 3 months of sampling on the Nyagui River
CCA axis 1CCA axis 2CCA axis 1CCA axis 2CCA axis 1CCA axis 2
Cumulative %30.450.823.541.129.647.3
Sand    0.665−0.189
Rock    −0.5870.009
Very shallow (<5 cm)−0.8000.1020.6660.4400.7270.193
Moderate (20–50 cm)      
Deep (>50 cm)    −0.629−0.331
Very slow (<0.05 m s−1)  0.6900.0490.823−0.132
Slow (0.05–0.2 m s−1)  0.6930.559  
Moderate (0.2–0.4 m s−1)−0.718−0.095−0.592−0.047−0.565−0.137
Fast (0.4–1 m s−1)−0.547−0.353−0.6960.198−0.7490.043
Torrent (>1 m s−1)  −0.6960.333−0.8080.214
Figure 3.

 (a) Canonical correspondence analysis (CCA) showing the relationship between significant (P < 0.05) environmental variables and species relative abundance in October 2004. (b) CCA showing the relationship between significant (P < 0.05) environmental variables and species relative abundance in November 2004. (c) CCA showing the relationship between significant (P < 0.05) environmental variables and species relative abundance in January 2005

There was a significant positive correlation between habitat diversity and species richness (r2 0.74, < 0.01 and r2 0.68, < 0.01 respectively). This relationship was, nevertheless, not significant in January.


The occurrence of fish species at specific locations in a river can be influenced by microhabitat factors such as substrate, depth and water velocity (Vadas & Orth, 2000; Bond & Lake, 2003). Taylor (2000) suggested that these microhabitat variables, which are influenced by elevation, stream gradient and stream size, ultimately influence fish communities in both pools and riffles. CCA showed a separation between the shallow and fine-substrate dominated upstream section and fast flowing riffle-dominated dominated downstream section of the river.

The upstream section had Barbus paludinosus and Barbus lineomaculatus as the most frequently occurring species. These small sized fish species are adapted to the conditions of the headwaters that inundate during the rain season and gradually become fragmented during the dry season (Bell-Cross & Minshull, 1988). This observation corresponds to the view that headwater streams, due to fluctuating flow, depth and dissolved oxygen, are harsh environments inhabited by few species that can either withstand the extreme conditions or recolonize after a disturbance (Schlosser, 1982; Magoulick, 2000). The headwater streams of this study appear to be maintained by ground water seepage from adjacent wetlands (dambos) during the dry season.

The downstream comprised species favouring fast flowing water and rocky habitats, and of particular interest was O. zambezense. The genus Opsaridium has two species in Zimbabwe, O. zambezense and Opsaridium peringueyi whose status is a cause for concern with the latter probably now extinct (Marshall & Gratwicke, 1999). In the Nyagui River, this species is abundant. The availability of riffles on the downstream section of the river appears to influence positively the abundance of this species and other riffle-specialized species. Another species associated with fast flowing water is L. marequensis. It has disappeared in the upper Manyame catchment, yet it is abundant in the downstream section of the Nyagui River. Although it was abundant in the 1970s in the upper Manyame, L. marequensis has not been able to adapt to dam building, pollution and the presence of exotic predators (Gratwicke, Marshall & Nhiwatiwa, 2003).

No comparative study on fish species and microhabitat association has been carried out in Zimbabwe but several studies elsewhere suggest that these associations may be a result of factors such as predation (Gilliam et al., 1993), condition specific competition (Taniguchi & Nakamo, 2000) and morphological adaptations of the fish to the physical conditions of the river (Leveque, 1997). Gelwick (1990) indicated that large streams have distinct pool and riffle dwelling species, and pools only act as refugia during drought for riffle species, while riffles act as supplementary habitats for juveniles of pool dwelling species. The important predator in the system was M. salmoides but it was unlikely to have an impact as there was no correlation between its relative abundance and both species richness and relative abundance of other species. Species differences between pools and riffles were probably a result of morphological differences. In addition to O. zambezense, the riffles had species adapted to fast and turbulent waters such as C. neumanni, L. cylindricus and L. marequensis. Chiloglanis neumanni is a benthic species that prefers fast flowing gravel and pebble substrates, and possess a buccal sucker that enables it to adapt to these conditions (Skelton, 1993). Another species that possesses a buccal sucker is L. cylindricus that prefers pebble and rocky substrates. The mouthparts of O. zambezense and L. marequensis suggest that they are water column feeders. Among the cichlids, the river haplochromines, P. acuticeps and P. philander, were associated with deep pools with rock and pebble substrates while the tilapias were mainly associated with sand substrates and aquatic vegetation.

The study also illustrated the upstream and in downstream differences in species richness with the former having low species richness (nine) compared to the latter (twenty). The increase in number of species from upstream to downstream has been noted in many streams and explained in relation to increasing habitat complexity (Schlosser, 1982), increasing habitat area and volume (Angermeier & Schlosser, 1989) and increasing stability in flow from upstream to downstream (Horwitz, 1978). These are all factors associated with increasing the space for feeding, refugia and reproduction, which enable coexistence among stream dwelling fish.

Observations from this study suggest that the fish species and habitat associations have important implications for the management and conservation of stream fishes. An alteration in the flow, either natural or human induced, can lead to changes in the community structure. The construction of the Kunzvi Dam, for instance, would change the fish populations in this system, both in the area flooded and downstream. As in other reservoirs in Zimbabwe, the impoundment would lead to a decline or disappearance of rheophilic species (Kenmuir, 1984; Karenge & Kolding, 1995). The most important family to benefit from the construction of the dam is the cichlids and it is likely that O. niloticus would become the most important species, as it has in other impoundments. Species that will decline include most cyprinids and the rock and sand catlets (Chiloglanis and Zaireichthys spp.). Change in flow on the downstream would allow species like the cichlids to become more abundant than they are at present.


The Government of Belgium, through the University of Zimbabwe/Flemish Universities link, funded this study. Professor Brian Marshall helped with many aspects of this study. Albert Chakona, Taurai Bere, Tendai Samukange and Cleopas Chinheya assisted with fieldwork.