Fish assemblages in the Bia river-lake system (south-eastern Ivory Coast): A self-organizing map approach

F K Konan*, E O Edia, Y K Bony*, A Ouattara and G Gourène

Laboratory of Aquatic Environment and Biology, UFR-Sciences and Environment Management,University of Nangui Abrogoua, 02 BP 801 Abidjan 02, Ivory Coast.
* University Jean Lorougnon Guédé, BP 150 Daloa, Ivory Coast.
konanfelix@yahoo.fr

Abstract

Fish assemblages in the Bia river-lake system in south-eastern Côte d’Ivoire were investigated using a non linear method, the Self-Organizing Map (SOM). Samples were collected monthly during 23 sampling surveys between August 1995 and September 1997 at seven sampling sites (one site in the upstream, four in the lacustrine area and two in the downstream). A total of 78 fish species belonging to 47 genera, 27 families and 9 orders were recorded.

The upstream, with 42 species, was less rich in species than the lake and downstream areas (respectively 47 and 55). Samples were classified into four clusters and the classification was mainly related to the spatial factor. The downstream area (near the Aby lagoon) was distinguished from the others by the presence of numerous marine/brackish species. The downstream presented low similarity in fish composition with the two others areas. This heterogeneity could be due to the presence of dams on this river. These barriers impede fish migration between downstream and lake areas. These results led us to conclude that SOM displayed reliable pattern fish assemblages in the Bia river-lake system and can be effectively used as a tool for patterning tropical aquatic communities.

Keywords: Artificial neural networks, Bia hydrosystem, Ichthyofauna, West Africa

Introduction

In rivers, numerous factors potentially affect biological communities: water quality, food, hydraulics, the zoogeographic history, biotic factors, fishing activities and dams (Orth 1987; Gorman 1988; Lobb and Orth 1991). Dams affect the longitudinal gradient by changes in basin geomorphology and hydrology and, consequently, in physicochemical and biological variables (Oliveira et al 2004). As for biological variables, the presence of those infrastructures limits for example the migrations of biological communities and along such gradient, fish assemblages may vary widely in composition (Matthews et al 1989; De Mérona 2005). In tropical areas, the hydroelectric reservoirs have been studied widely (Carvalho et al 1998; Gourène et al 1999; Da Costa et al 2000; De Mérona 2005).

Ecosystems are structured in both spatial and temporal, so one of the most relevant questions when studying an ecosystem and its component organisms is the strategy of its occupation with respect to space and time (Reyjol et al 2005). The study of ecological communities leads to obtaining a huge matrix, which is often a difficult structure(Giraudel and Lek 2001). Ordination methods, designed to summarize and simplify complex data sets, can reveal factors contributing to the structure of the community under consideration (Ludwig and Reynolds 1988). Multivariate approaches may offer an appropriate means for examining the spatial and temporal variation in an ecosystem (Resh and Rosenberg 1989). For example, to elucidate factors contributing to the structure of fish communities, multivariate approaches are widely used (Da Costa et al 2000; Koné et al 2003; Kouamélan et al 2003; Reyjol et al 2005; Konan et al 2006; Grenouillet et al 2011).

The fish community of Bia River-lake system comprises 63 fish species belonging to 41 genera, 24 families and 9 orders (Gourène et al 1999; Da Costa et al 2000). Previously, these biota and environmental factors were examined with use of PCA and cluster analysis (Gourène et al 1999; Da Costa et al 2000). These authors have shown that diversity of fish community in Bia river displays very high spatial variation. To study the spatial and temporal organization of fish assemblages, we used an unsupervised artificial neural network, the self-organizing map (SOM), which is a clustering technique capable of displaying patterns in complex data sets (Kohonen 2001). This method has proved to be effective in characterizing distribution patterns in community ecology analysis (e.g. Park et al 2003; Tison et al 2004) with the advantage of representing non-linear relationships (Lek et al 2000). Moreover, SOM handles species with low frequency of occurrence (i.e., rare species) contained in many ecological data sets (Brosse et al 2001; Ibarra et al 2005), unlike conventional methods such as multivariate analysis (e.g. PCA). It shares with conventional ordination methods the basic goal of displaying a multidimensional data set in a lower, usually two-dimensional, space.

This study aimed to analyze the patterns of fish assemblage diversity in the Bia river-lake system with the Kohonen self-organizing map (SOM).

Materials and methods

Study area and sampling sites

Located in the South-East of Côte d’Ivoire, the Bia River (Figure 1) belongs to the Western Guinean ichthy oregion, sector Eburneo-Ghanaian. This river encompasses an area of 9300 km². With 300 km length, the Bia River (05°30’ – 05°50’ N and 03° – 03°15’ W) has a mean annual flow of 83 m³.s⁻¹. Two hydroelectric dams were built on this river (Ayamé I in 1959 and Ayamé II in 1963). There is no particular scheme for the transit of fish in these two dams.

Sampling sites were retained in the upstream (Bianouan), the lake (Kétésso, Ebikro, Bakro, Ayamé) and the downstream (Aboisso, Krindjabo) areas of this stream (Figure 1).

Figure 1. Location of Bia River and the sampling stations.

Fish sampling

At all sampling sites retained, 102 samples were collected monthly between August 1995 and September 1997. The sampling sites covered a river section of approximately 1.5 km in length (i.e. reach scale). This river and lake section length was selected to cover a fair degree of habitat heterogeneity. At each sampling site, fishes were collected with two sets of 8 gill nets (mesh sizes 10, 12, 15, 20, 25, 30, 40, 45 and 50 mm), allowing the capture of almost all the fish longer than 90 mm total length. These gill nets were 25 m long and 2 m high. At each sampling occasion, fishing was done overnight (17.00 to 7.00) and during the day (7.00 to 13.00). All fish specimens were identified according to the identification keys of Lévêque et al (1990; 1992).

Species diversity analysis

First, we established a list of species sampled in the upstream, lake and downstream areas of Bia basin. Then, the beta diversity index was applied to quantify the similarity between fish composition along longitudinal gradient. Whittaker’s index (β_w) (Whittaker 1972) was used: β_w = (S_R/α_mean)-1. With S_R = total richness in each zone, and α_mean = mean richness of zones within study area.

Self-Organizing Map (SOM)

The distribution pattern was displayed using the Self-Organizing Map (SOM) by means of the toolbox developed by Alhoniemi et al (2003) for Matlab^®. First, a species occurrence data set was arranged as a matrix of 102 rows (i.e., the samples) and 78 columns (i.e., the species). Each of the 102 samples of the data set can be considered as a vector of 78 dimensions. Species occurrence was preferred to fish abundance because of the selectivity of gill net sampling techniques, which can bias the calculation of species abundance (Hugueny et al 1996). Then, the species occurrence data set was patterned by training the SOM. The architecture of the SOM consisted of two layers of neurons (or nodes): (i) the input layer that was composed of 78 neurons connected to each vector of the data set and (ii) the two-dimensional output layer was composed of 54 neurons (i.e. a rectangular grid with 9 by 6 neurons laid out on a hexagonal lattice). We chose a 54-neuron grid because this configuration presented minimum values of both quantization and topographic errors, which are used to assess classification quality (e.g. Kiviluoto 1996; Kohonen 2001; Park et al 2003). The SOM algorithm calculates the connection intensities (i.e. vector weights) between input and output layers using an unsupervised competitive learning procedure (Kohonen 2001), which iteratively classifies samples in each node according to their similarity in species composition. Thus, the SOM preserves the neighborhood so that samples with close species occurrences are grouped together on the map, whereas samples with very different species occurrences are far from each other.

We checked whether relevant groups of samples characterized distinct fish assemblages by performing a hierarchical cluster analysis (Ward’s linkage and Euclidean distance). To do so, we used a new matrix (54 × 78, output neurons × species) of the connection intensity values estimated by the SOM.

Between-cluster differences in species richness were evaluated using the Kruskal-Wallis test, a non-parametric analysis of variance, followed by Mann-Whitney test to identify specific differences.

The relationships between the fish community associated with each cluster and the factors of the sampling plan (year, month and sampling zone), which were not used in the SOM learning process, were assessed. To do so, we calculated for each cluster the relative contribution of each modality (e.g., lake) of a certain factor (e.g., sampling zone) to the total number of modalities of the factor. Then, we tested for the H₀ that no significant differences were present among these relative values using a test for proportions based on the likelihood ratio c² statistics (Sachs 1997). Statistical calculations were performed using R (Ihaka and Gentleman 1996).

Results

Fish species composition

In the ichtyofauna of Bia river-lake system, a total of 78 fish species belonging to 47 genera, 27 families and 9 orders were captured (Table 1). Perciform (37% of families, 43% of genera and 37% of species) was the most abundant order, followed by Siluriform (18%, 15% and 21%), Characiform (15%, 13% and 15%) and Osteoglossiform (11%, 15% and 12%). Among the families sampled, Cichlidae (19% of species), Alestidae (10%), Mormyridae and Cyprinidae (9%) and Clariidae, (8%) were largely represented. Twenty five marine/brackish species (i.e. 32% of the species) were collected, as well as two introduced species (Oreochromis niloticus and Heterotis niloticus).

The upstream, with 42 species, was less rich in species than the lake and downstream areas (respectively 47 and 55). However, no difference (P> 0.05) was recorded between these areas (Figure 2). Moreover, 37 species are common to the upper and lake areas of Bia basin. The upstream and downstream areas gathered 27 species. Then, 27 species are common to the lower courses and the lake area. Last, the three areas gathered 24 species.

Figure 2. Boxplots comparing fish species richness in the three sampling zones. Box-plots were performed using the species richness of samples gathered in each sampling zones. The box corresponds to 50% of the values, the horizontal bar in the box to the median and vertical bars to the minimum/maximum values.

Among longitudinal gradient, beta diversity index recorded in the Bia River was 0.47 between the downstream and the lake, 0.44 between the downstream and the upstream. The lowest beta diversity value, which was observed between the lake and upper course (0.17), revealed a high similarity between those last both sampling sites.

Classification of samples using SOM

The samples were classified by the SOM according to their species composition in the 54 output nodes, so that each node included samples with similar species (Figure 3). The sample distribution pattern classification was mainly related to the spatial factor. Indeed, all samples from downstream courses of Bia River were located in the upper part of the SOM map, whereas samples from the upstream course were classified mostly in the lower-right part. The samples collected in the lake were mainly grouped in the middle and the lower-left areas of the map. The SOM units were classified into two main clusters based on the dendrogram of the cluster analysis with the Ward’s linkage method. The SOM map was further divided to four clusters at different levels of the Euclidean distance. Cluster I were composed solely of samples from the upstream area, whereas those from cluster IV were exclusively related to downstream area. Samples from cluster II and III were notably associated to the lake.

Figure 3. Classification of samples using presence-absence data through the learning process of the self-organizing map. Up: Bianoua (upstream of Bia River); L1: Ketesso (upstream of the Lake Ayamé I); L2: Ebikro (middle of the Lake Ayamé I); L3: Ebikro (middle of the Lake Ayamé I); L4: Ayamé (downstream of the Lake Ayamé I); D1: Aboisso (downstream of Bia River); D2: Krindjabo (downstream of Bia River); numbers 01 to 12 = months (January to December); numbers 95 to 97 = years (1995 to 1997).

The Kruskal-Wallis test showed highly significant difference in species richness between clusters (P< 0.001) (Figure 4). There were no differences (Mann-Whitney test, P > 0.05) between clusters I and III. These two clusters were different (P < 0.001) from clusters II and IV. Cluster II comprised fewer species than the others (P< 0.001).

Figure 4. Boxplots comparing fish species richness in the four clusters defined by the self-organizing map. Box-plots were performed using the species richness of samples gathered in the clusters identified in Figure 3. The box corresponds to 50% of the values, the horizontal bar in the box to the median and vertical bars to the minimum/maximum values. The different alphabets (a, b, c) indicate significant differences between the clusters based on the Mann–Whitney comparison test (p < 0.01).

The fish assemblage pattern in the SOM map is presented (Table 2). Cluster IV was mainly associated with marine/brackish fish species (22 species). Then, cluster II was characterized by families represented by only one species, excepted Alestidae and Cichlidae. Samples gathered in cluster III had high richness of Cichlidae(10 species), Alestidae (6 species) and Mormyridae (5 species). Last, the ichtyofauna of cluster I samples was dominated by the family of Cyprinidae (6 species).

Discussion

There were no significant differences between species richness of sampling zones during the sampling period. Along the longitudinal gradient, the upstream and lake areas presented lower species composition turnover (i.e., higher similarity). This homogeneity in fish composition among these two sampling zones results from higher connectivity between upstream and lake, since they are contiguous and the fish swim more easily between them. In contrast, the lake and downstream in the Bia River presented the highest beta-diversity value (i.e., lower similarity). The highest species composition heterogeneity within lake and downstream suggests that both zones present partially distinct ichtyofauna. The upper course and the lake were restricted to the downstream part after the construction of the two dams Ayamé I and II. These barriers impede fish migration between downstream and lake areas.

Fish assemblages, in this study, were patterned through an adaptive learning algorithm, the self-organizing map (SOM), according to the distribution similarities of each species. The suitability of this tool is known to provide more relevant classifications and ordinations than conventional multivariate analysis due to the ability of SOM to consider rare species without over fitting bias (Giraudel and Lek 2001). The clearest patterns observed were those associated to the spatial gradient. Overall the SOM showed four clusters of fish assemblages, related to the longitudinal river gradient. Cluster IV exhibited the highest species richness and was significantly different from the three others. Cluster II comprised fewer species than cluster I, whereas there were no significant differences between clusters I and III. Moreover, cluster I and IV were exclusively composed of samples from the upper and the lower course respectively, while samples from cluster II and III were significantlyassociated to the lake area. Cluster IV was distinguished from the others by the presence of numerous marine/brackish species. This cluster includes all the samples of the downstream area of the Bia River which is closest to the Aby lagoon (a brackish water). According to Pouyaud (1994), Lévêque and Paugy (1999) and De Mérona (2005), a number of species, from the brackish and/or marine water species, can go upstream in the rivers. Overall, more than 33% of the species collected during this study are estuarine/marine species and characterize the lower course of the river. Numerous authors such as Da Costa et al (2000), Koné et al (2003), Kouamélan et al (2003), Konan et al (2006), already reported the presence of estuarine/marine species in Western African river basins. According to Albaret (1994), some taxa such as Liza falcipinnis which is perfectly euryhaline can migrate in downstream areas. Bruslé (1981) and Albaret and Legendre (1985) explained this distribution directly by its hyper euryhalinity and its broad food spectrum. As Liza falcipinnis, the diverse species with both estuarine and marine affinities, adapted to these various hydrosystems, would be better suited for using additional food resources and hence optimizing the energy costs of the reproduction, which can occur in brackish water (Albaret and Legendre 1985), or in freshwater (Bruslé 1981), and sometimes in the ocean.

Moreover, Awaous lateristriga absent from Bia River (Teugels et al 1988; Gourène et al 1999; Da Costa et al 2000) was recorded in neighboring rivers such as Soumié and Eholié Rivers (Konan et al 2006) as well as Tano River (Teugels et al 1988). This species was also sampled in Agnéby River (Da Costa et al 2000). Lévêque and Paugy (1999) and De Mérona (2005) reported that dams prevent the migration of estuarine and/or marine fishes to upper course during floods. This situation disturbs their life cycle, especially reproduction in favorable habitats. Moreover Lévêque and Paugy (1999) were noted that dams limit the distributional range of these species. On the Soumié River, a tributary of the Bia River in its downstream part, there were no dams that could impede fish migration and these species were sampled in both upstream and downstream sites (Konan et al 2006).

In the lake area, the presence of aquatic macrophytes and dead woody maintain the presence of many fauna. Indeed, this environment supports the colonization for a various organisms (e.g. crustaceans and insects) (Oliveira et al 2004). Moreover, for Thomaz and Bini (1999), aquatic plants are important habitats for fish fauna because they increase spatial heterogeneity and feeding resource availability. In addition, the lake which is directly connected to the upstream, offer a wide range of exploitable microhabitats not only by species typical of lacustrine environments but also by those of riverine ones from upstream. This can explain the highest homogeneity, in terms of the icthyofauna composition, between the lotic (upper course) and lentic zones of Bia basin.

Decreased of species richness in the reservoir towards the upper course could be correlated to: flow regulation; reduction of aquatic vegetation, mainly floating macrophytes and to limnological conditions of this gradient.

Acknowledgments

We are grateful to D F E Thys, Van Den Audenaerde and G G Teugels, co-promoters of project VLIR (Vlaamse Interuniversitaire Raad) entitled “Evolution of freshwater fish biodiversity after a dam construction: case of Bia River in Côte d’Ivoire” financed by the Belgian Directorate General for International Cooperation.

References

Albaret J J and Legendre M 1985 Biologie et écologie des Mugilidae en lagune Ebrié (Côte d’Ivoire): Intérêt potentiel pour l’aquaculture lagunaire. Revued’Hydrobiologie Tropicale 18: 281-303.http://journalseek.net/cgi-bin/journalseek/journalsearch.cgi?field=issn&query=0240-8783

Albaret J J 1994 Les poissons : biologie et peuplements. In Durand JR, Dufour P, Zabi, SGF (eds) Environnement et ressources aquatiques de Côte d’ivoire : les milieux lagunaires, Tome II. ORSTOM, Abidjan: 239-279.

Alhoniemi E, Himberg J, Parankangas J and Vesanto J 2003 SOM Toolbox 2.0. Laboratory of Computer and Information Science. Neural Networks Research Center, Helsinki, Finland. http://www.cis.hut.fi/projects/somtoolbox/

Brosse S, Giraudel J L and Lek S 2001 Utilisation of non-supervised neural networks and principal component analysis to study fish assemblages. Ecological Modelling 146: 159-166. http://www.journals.elsevier.com/ecological-modelling/

Bruslé J 1981 Sexuality and biology of reproduction in grey mullets. In Oren OH (ed) Aquaculture of grey mullets. University Press, New York, Cambridge: 99-154.

Carvalho E D, Silva V F B, Fujihara C Y, Henry R and Foresti F 1998 Diversity of fish species in the River Paranapanema-Jurumirim Reservoir transition region (São Paulo, Brazil). Italian Journal of Zoology 65 : 325-330. http://www.uzionlus.it /Attivit/IJZ.aspx

Da Costa K S, Gourène G, Tito De Morais L and Thys Van Den Audenaerde D F E 2000 Caractérisation des peuplements ichtyologiques de deux fleuves côtiers ouest-africains soumis à des aménagements hydroagricoles et hydroélectriques. Vie &Milieu 50: 65-77. http://www.obs-banyuls.fr/Viemilieu/

De Mérona B 2005 Le fleuve, le barrage et les poissons: le Sinnamary et le barrage de Petit-Saut en Guyane française. Editions IRD, Paris 135p.

Giraudel J L and Lek S 2001 A comparison of self-organizing map algorithm and some conventional statistical methods for ecological community ordination. Ecological Modelling 146: 329-339. http://www.journals.elsevier.com/ecological-modelling/

Gorman O T 1988 The dynamics of habitat use in a guild of Ozark minnows. Ecological Monographs 58: 1-18. http://esapubs.org/esapubs/journals/monographs.htm

Gourène G, Teugels G G, Hugueny B and Thys Van Den Audenaerde D F E 1999 Evaluation de la diversité ichtyologique d’un bassin ouest africain après la construction d’un barrage. Cybium 23: 147-160. http://www.mnhn.fr/sfi/cybium/index.html

Grenouillet G, Buisson L, Casajus N and Lek S 2011 Ensemble modelling of species distribution: the effects of geographical and environmental ranges. Ecography 34: 9-17.  http://eu.wiley.com/WileyCDA/WileyTitle/productCd-ECOG.html

Hugueny B, Camara S, Samoura B and Magassouba M 1996 Applying an index of biotic integrity based on fish assemblage in West African river. Hydrobiologia 331: 71-78. http://www.springer.com/life+sciences/ecology/journal/10750

Ibarra A A, Park Y S, Brosse S, Reyjol Y, Lim P and Lek S 2005 Nested patterns of spatial diversity revealed for fish assemblages in the west European river. Ecology of Freshwater Fish 14: 233-242. http://www.blackwellpublishing.com/journal.asp?ref=0906-6691

Ihaka R and Gentleman R 1996 R a language for data analysis and graphics. Journal of Computational and Graphical Statistics 5: 299-314. http://www.amstat.org/publications/jcgs.cfm

Kiviluoto K 1996 Topology preservation in self-organizing maps. Proceedings of ICNN’96, IEE International Conference on Neural Networks. IEEE Service Center, Piscataway: 294-299.

Kohonen T 2001 Self - Organizing Maps, 3^rd edition. Springer-Verlag, Berlin, 501p.

Konan K F, Leprieur F, Ouattara A, Brosse S, Grenouillet G, Gourène G, Winterton P and Lek S 2006 Spatio-temporal patterns of fish assemblages in coastal West African rivers: a Self-Organizing Map approach. Aquatic Living Resources 19: 361-370. http://journals.cambridge.org/action/displayJournal?jid=ALR

Koné T, Teugels G G, N’Douba V, Gooré Bi G and Kouamelan E P 2003 Premières données sur l’inventaire et la distribution de l’ichtyofaune d’un petit bassin côtier ouest africain: rivière Gô (Côte d’Ivoire). Cybium 27: 101-106.http://www.mnhn.fr/sfi/cybium/index.html

Kouamélan E P, Teugels G G, N’Douba V, Gooré Bi G and Koné T 2003 Fish diversity and its relationships with environmental variables in a West African basin. Hydrobiologia 505: 139-146.http://www.springer.com/life+sciences/ecology/journal/10750

Lek S, Giraudel J L and Guégan J F 2000 Neuronal networks: algorithms and architectures for ecologists and evolutionary ecologists. In: Lek S, Guégan JF (eds) Artificial Neuronal Networks: Application to Ecology and Evolution. Springer-Verlag, Berlin :3-27.

Lévêque C and Paugy D 1999 Caractéristiques générales de la faune ichtyologique. In: Lévêque C, Paugy D (eds) Les poissons des eaux continentales africaines. Diversité, Ecologie, Utilisation par l’homme. Edition IRD, Paris : 43-54.

Lévêque C, Paugy D and Teugels G G 1990 Faune des poissons d’eaux douces et saumâtres de l’Afrique de l’ouest. Tome I. Faune tropicale, XXVIII. ORSTOM/MRAC, Paris/Tervuren, 384p.

Lévêque C, Paguy D and Teugels G G 1992 Faune des poissons d’eaux douces et saumâtres de l’Afrique de l’ouest. Tome II. Faune tropicale, XXVIII. ORSTOM/MRAC, Paris/Tervuren, 526p.

Lobb M D and Orth D J 1991 Habitat use by an assemblage of fish in a large warmwater stream. Transactions of the American Fisheries Society 120: 65-78. http://www.tandfonline.com/toc/utaf20/current

Ludwig J A and Reynolds J F 1988 Statistical ecology. A primer on methods and computing. John Wiley and Sons, Chichester, 337p.

Matthews W J, Hill L G, Edds D R and Gelwick F P 1989 Influence of water quality and season on habitat use by striped bass in a large southwestern reservoir. Transactions of the American Fisheries Society 118: 243-250. http://www.tandfonline.com/toc/utaf20/current

Oliveira E F, Goulart E and Minte-Vera C V 2004 Fish diversity along spatial gradients in the Itaipu reservoir, Paraná, Brazil. Brazilian Journal of Biology 64: 447-458. http://www.scielo.br/scielo.php?script=sci_serial&pid=1519-6984&lng=en&nrm=iso

Orth D J 1987 Ecological consideration in the development and application of instream flow-habitat models. Regulated Rivers-research & Management 1: 171-181.  http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)1099-1646/issues?activeYear=2001

Park Y S, Chang J,  Lek S, Cao W and Brosse S 2003 Conservation strategies for endemic fish threatened by the three Gorges Dam. Conservation Biology17: 1748-1758. http://eu.wiley.com/WileyCDA/WileyTitle/productCd-COBI.html

Pouyaud L 1994 Génétique des populations de tilapias d’intérêt aquacole en Afrique de l’Ouest. Relations phylogénétiques et structurations populationnelles. Thesis, University of Montpellier, France, 229p.

Resh V H and Rosenberg D M 1989 Spatial-temporal variability and the study of aquatic insects. Canadian Entomologist 121: 941-963. http://www.bioone.org/loi/cent

Reyjol Y, Fischer P, Lek S, Rösch R and Eckmann R 2005 Studying the spatiotemporal variation of the littoral fish community in a large prealpine lake, using self-organizing map. Canadian Journal of Fisheries and Aquatic Sciences 62: 2294-2302. http://www.nrcresearchpress.com/journal/cjfas

Sachs L 1997 Angewandte Statistik, 8^th edition. Springer, Berlin, Heidelberg, New York. XXXIV, 884p.

Teugels G G, Lévêque C, Paugy D and Traoré K 1988 Etat des connaissances sur la faune ichtyologique des bassins côtiers de la Côte d’Ivoire et de l’ouest du Ghana. Revued’Hydrobiologie Tropicale 21: 221-237. http://journalseek.net/cgi-bin/journalseek/journalsearch.cgi?field=issn&query=0240-8783

Thomaz S M and Bini L M 1999 A expansão das macrófitas aquáticas e implicações para o manejo de reservatórios: um estudo na represa de Itaipu. In: Henry R. (ed) Ecologia de reservatórios: estrutura, função e aspectos sociais. FAPESP: FUNDIBIO, Botucatu: 597-625.

Tison J, Park Y S, Coste M, Delmas F and Giraudel J L 2004 Use of unsupervised neural networks for eco-regional zonation of hydrosystems through diatom communities: case study of Adour-Garonne watershed (France). Archiv für Hydrobiologie 159: 409-422. http://journalseek.net/cgi-bin/journalseek/journalsearch.cgi?field=issn&query=0003-9136

Whittaker R H 1972 Evolution and measurement of species diversity. Taxon 21: 213-251. http://www.botanik.univie.ac.at/iapt/taxon/

Received 16 August 2012, Accepted 13 December 2012; Published 4 January 2013

Go to top