Characterization of fish assemblages and population ...

Environ Monit Assess (2016) 188:364 DOI 10.1007/s10661-016-5364-6

Characterization of fish assemblages and population structure of freshwater fish in two Tunisian reservoirs: implications for fishery management Sami Mili & Rym Ennouri & Amel Dhib & Houcine Laouar & Hechmi Missaoui & Lotfi Aleya

Received: 14 October 2015 / Accepted: 16 May 2016 # Springer International Publishing Switzerland 2016

Abstract To monitor and assess the state of Tunisian freshwater fisheries, two surveys were undertaken at Ghezala and Lahjar reservoirs. Samples were taken in April and May 2013, a period when the fish catchability is high. The selected reservoirs have different surface areas and bathymetries. Using multi-mesh gill nets (EN 14575 amended) designed for sampling fish in lakes, standard fishing methods were applied to estimate species composition, abundance, biomass, and size distribution. Four species were caught in the two reservoirs: barbel, mullet, pike-perch, and roach. Fish abundance showed significant change according to sampling sites, depth strata, and the different mesh sizes used. From the reservoir to the tributary, it was concluded that fish biomass distribution was governed by depth and was

S. Mili Unité de recherche: Exploitation des Milieux Aquatiques, Institut Supérieur de Pêche et d’Aquaculture de Bizerte, Errimel, B.P.15, 7080 Bizerte, Tunisia R. Ennouri : A. Dhib : H. Missaoui Laboratoire Milieu Marin, Centre la Goulette, Institut National des Sciences et Technologies de la Mer (INSTM), La Goulette, Tunisia H. Laouar Centre Technique d’Aquaculture, 5, rue du Sahel Montfleury, 1009 Tunis, Tunisia L. Aleya (*) Laboratoire de Chrono-Environnement, UMR CNRS 6249, Université de Bourgogne Franche-Comté, Besançon, France e-mail: [email protected]

most abundant in the upper water layers. Species size distribution differed significantly between the two reservoirs, exceeding the length at first maturity. Species composition and abundance were greater in Lahjar reservoir than in Ghezala. Both reservoirs require support actions to improve fish productivity. Keywords Fish assemblage . Multi-mesh gill nets . Fish metrics . Tunisian reservoirs

Introduction Freshwater fish farming is a recent activity in developing Tunisia that began with the experimental stocking of reservoirs with mullet. The majority of reservoirs are located in the northern part of the country (DGPA 2013), with 450 fishermen and 232 boats currently involved. The most commonly caught species are as follows: carp (Cyprinus carpio), pike-perch (Stizostedion lucioperca), mullet (Mugil cephalus and Liza ramada), eel (Anguilla anguilla), catfish (Silurus glanis), roach (Rutilus rutilus), barbel (Barbus setivimensis), and tilapia (Oreochromis niloticus). Production increased from 843.5 tons in 2000 to over 969 tons in 2012, with mullet as the most heavily fished species, comprising 35 % of total production and totalling an annual average of about 345 tons (DGPA 2013). The principal means adopted to ensure continuity and development of fish farming in Tunisia is to stock the reservoirs with Mugilidae. This practice

364

Environ Monit Assess (2016) 188:364

Page 2 of 11

should thus be encouraged in order to develop the economies of inland areas, especially in the country’s disadvantaged regions (Loss et al. 1991). However, despite efforts undertaken by institutions and the promising growth rate observed in the freshwater fish population (Fischer and Quist 2014), production levels of these species remain low as fisheries encounter many problems, especially the lack of reliable production statistics and overall planning for the further development of the activity (CTA 2011). In addition, despite numerous cited studies of the freshwater fish biology in Tunisian reservoirs (Djemali 2005; Mili et al. 2015b; Tlili et al. 2010; Toujani et al. 2000), little information concerning their population status, distribution, and abundance is available. While fish population monitoring by acoustic methods has previously been conducted in three of the country’s reservoirs, gaps in our general knowledge of the fish spatial distribution in these ecosystems persist (Djemali et al. 2009) due to the fact that this technique cannot provide complete information. Additionally, acoustic sampling methods are limited by their inability to differentiate species or to detect fish near boundaries, in particular in surface waters or on lake bottoms (Simmonds and MacLennan 2005). We thus hypothesized that by adopting the fish sampling technique using multi-mesh gill nets inspired by the European norm CEN (European Committee for Standardization 2005) for lake fish would (1) provide good information on trends in fish density based on the catch per unit effort (CPUE), (2) enable us to attempt a comparison of CPUE with real population density (Achleitner et al. 2012), and, ultimately, (3) help in sustaining the management of these resources. The objective of this study was to determine the spatial distribution and abundance of freshwater fish in two reservoirs—Ghezala and Lahjar—located in different areas in Tunisia, by means of multi-mesh gill nets (CEN 2005 amended). This method has recently become standard throughout Europe (Deceliere-Vergès and Guillard 2008; Argillier et al. 2013) in assessing fish assemblages in lakes and reservoirs, but to date, unfortunately, has never been applied in Tunisian reservoirs. The present study will be useful for fishery managers by giving details on the status of the freshwater fish resources in reservoirs so as to help improve productivity.

Materials and methods Study sites The two reservoirs under consideration, Ghezala (122 ha) (37° 03′ 12″ N/37° 04′ 54″ N and 09° 30′ 26″ E/09° 31′ 43″ E) and Lahjar (100 ha) (36° 50′ 18″ N/36° 52′ 51″ N and 11° 01′ 14″ E/11° 02′ 37″ E), are located in the north and northeast of Tunisia (Mili et al. 2015a), respectively, each with a maximum depth ranging from 13 to 14 m (Fig. 1). Ghezala reservoir provides irrigation for farmland situated at some distance, near Bizerte, whereas Lahjar is used exclusively for irrigation of a large surrounding area (Soudoud 2006). The annual rainfall observed at the two reservoirs was 740 mm for Ghezala and 540 mm for Lahjar (Soudoud 2006). Sampling Sampling methodology was after the European norm related to the study of water quality sampling of fish using multi-mesh gillnets (EN 14575 amended). Samples were taken between April 17 and May 5, 2013, a period during which most freshwater fish species in reservoirs do not spawn and when water temperature usually exceeds 15 °C (Djemali 2005). To cope with the irregular distribution of fish species, a stratified random sampling design was used. The sampled reservoir was divided into depth strata, with samples taken within each stratum. Benthic species were sampled with specially designed multi-mesh gill nets 20 m long by 1.5 m deep. Eight different mesh sizes were used, ranging from 18 to 80 mm, knot to knot, following a geometric series. Gill nets used for sampling pelagic fish were 20 m long by 6 m deep. Thread diameter varied from 0.23 to 0.28 mm. The hanging ratio was 0.5 for all horizontal mesh sizes and 0.71 for lateral ropes. A sampling effort of 24 gill nets per night (CEN 2005) was needed in order to detect a 50 % change in relative abundance. Vertical distribution of species was analyzed to assess the information obtained from surface to bottom. Four depth strata were defined for gillnet catches (0– 3 m, 3–6 m, 6–9 m, 9–12 m), with random samples taken within each stratum. To avoid calculating abundance relative to hours of net setting time, a standard fishing period of 12 h was adopted; nets


Page 3 of 11 364

Fig. 1 Study areas and sampling station locations in Ghezala (A) and Lahjar (B) reservoirs

were set between 6 and 8 p.m. and lifted between 6 and 8 a.m. The total surface cover per night in each reservoir for benthic gill nets was 720 m2; for pelagic gill nets, the total surface cover was 120 m2 for Lahjar and 240 m 2 for Ghezala. Each sampling station (gillnet sets) was composed of two benthic nets. For the pelagic zone, nets were placed separately. Catches, expressed as number per unit effort (NPUE), and biomass (weight per unit effort (WPUE)) were calculated for each reservoir. NPUE and WPUE for gill nets were expressed per 1000 m2 net area and for an exposure time of 12 h according to EN 14757 (CEN 2005). All fish caught during the gillnet fishing surveys were numbered, identified, counted, measured for total length (TL to nearest 1 mm), and weighed (to 1 g accuracy).

Data analysis For fish abundance, redundancy analysis (RDA) was performed to define the structuring effects of sampling sites and depth strata, along with the effects of the different nets used. Fish abundance values were normalized prior to analysis. A generalized additive model (GAM) was used to define the most abundant size class of each species (Wood 2006). GAMs may be considered as a nonparametric generalization of linear regressions and are increasingly used in marine ecology. Model validation was assessed according to Züur et al. (2010). Statistical analysis and graphic display were produced using R 2.15.0 (R Development Core Team 2012) with the R packages of BVegan^ (Oksanen et al. 2011), BPgirmess^ (Giraudoux 2012), and Bmgcv^ (Wood 2006).

364

Page 4 of 11

Results Species composition Using the sampling protocol with multi-mesh gill nets (CEN 2005), the capture of three fish species was recorded in Lahjar reservoir: mullet (L. ramada), roach (R. rutilus), and pike-perch (Sander lucioperca). Four species were caught in Ghezala reservoir: barbel (B. setivimensis), mullet, roach, and pike-perch. A total of 375 fish were captured during sampling in Lahjar reservoir, while only 51 were caught in Ghezala reservoir. Mugilidae (mullet) and cyprinid (roach) species were the most numerous, representing over 80 % of the total catch. Benthic and pelagic gillnet catches were heavy in the epilimnetic water layer with enhanced species diversity.

Fish species dynamics Fish abundance changed significantly according to sampling site (RDA, F = 12.29, p < 0.001), depth strata (RDA, F = 2.57, p < 0.05), and the different nets used (RDA, F = 3.41, p < 0.05). These factors together explain 52.73 % of changes in fish abundances on the RDA axes 1 and 2, both supporting a significant effect on this variability (p < 0.05; Fig. 2). Concerning the sites, Ghezala reservoir shows RDA scores close to 0 and appears to involve the different

Fig. 2 RDATriPlot depicting association between fish abundance and sampling sites, depth, and nets. Eigenvalues of the first two axes are indicated by 1 and 2. BN benthic nets, PN pelagic nets


types of fish. Barbel has similar scores at Ghezala and appears to be exclusively present there. Lahjar reservoir clearly stands out with a positive RDA axis and appears to involve a greater number of roach. In fact, in the absence of barbel, roach comprise the greatest number of prey for pike-perch. These species were significantly correlated, suggesting similar behavior; their correlation with the sampling zone near the dam was also considerable. However, the number of specimens caught did not allow determination of fish size variability depending on dam and depth strata. As for depth strata, a clear association was observed between the (0–3 m) range and the majority of fish. Roach had a positive RDA axis 1 and a negative RDA axis 2 and appears to be more present in the 3–6-m and 6–9-m ranges. Where the nets are concerned, barbel has RDA scores close to those of the benthic nets and appears to be caught with the latter exclusively. The same result was observed for roach which may be more present in pelagic nets. The remaining fish appear to be caught equally well with both nets. Vertical distribution As shown in Fig. 2, an association between fish species and the 0–3-m stratum was observed. Catches of individual species in Lahjar reservoir were not uniformly distributed across the sampled depth layers (0–3 m, 3–6 m, 6–9 m, and 9–12 m) (RDA, F = 2.57, p < 0.05). Most catches (50.93 % of total) were made in the 0–3-m stratum, followed by the 3–6-m (33.33 %) and the 6–9-m (15.73 %) strata, respectively (Fig. 3). The most numerous species caught in the different strata of this reservoir was roach, more abundant in the 3–6-m and 6–9-m strata (Fig. 2). No specimen of pike-perch was caught at a depth exceeding 6 m though mullet was heavily abundant below this depth (Fig. 3). The fish inhabiting both the reservoirs under study are present in the 0–3-m, 3–6-m, and 6–9-m strata, but they are particularly abundant in the upper strata (RDA, F = 2.57, p < 0.05). Figure 3 above shows that no specimen was caught at a depth exceeding 9 m, in either reservoir. In Ghezala reservoir, most catches (76.47 %) were made in the 0–3-m stratum, followed by the 3–6-m (21.56 %) and the 6–9-m (1.96 %) strata, respectively. The most numerous species caught in the different strata


Page 5 of 11 364

Lahjar reservoir

Fig. 3 Distribution of catches in Lahjar reservoir

of this reservoir was roach, followed by mullet (Fig. 4). Catch rates were not uniformly distributed across the sampled depth strata (0–3 m, 3–6 m, 6–9 m, and 912 m). Catches of individual species differ significantly between strata (RDA, F = 2.57, p < 0.05) with abundance higher in the superficial stratum (0–3 m) than in the deeper ones.

The number of roach caught (326) was much higher than those of mullet (26) and pike-perch (23), an observation noted for both benthic and pelagic nets (RDA, F = 12.29, p < 0.001). NPUE of roach captured using benthic nets (432 fish/1000 m2) exceeds the pike-perch (32 fish/1000 m2) and mullet yields (24 fish/1000 m2). Numerical and weight percentages of roach caught with benthic nets were 88.86 and 71.33 %, respectively. Percentages for mullet were 4.86 and 20.89 % and for pike-perch 6.29 and 7.79 %. For pelagic nets, the numerical yield of roach (125 fish/1000 m 2 ) was higher than for mullet (75 fish/1000 m2) and pike-perch (8 fish/1000 m2), their NPUE percentages being 60, 36, and 4 %, respectively. The catch rates by weight for roach (16.175 kg/ 1000 m2) and mullet (14.633 kg/1000 m2) were similar; they were lowest for pike-perch (0.716 kg/1000 m2). Ghezala reservoir

General and specific abundance Yields in Lahjar reservoir (CPUE) were high in number (446 fish/1000 m2 net) but weak in terms of weight (26.38 kg/1000 m2 net). In Ghezala reservoir, they were low in both number and weight (53 fish/1000 m2 net and 5.43 kg/1000 m2 net) (Table 1). By including information derived from the pelagic nets, total NPUEs were (39.68 fish/1000 m2 net and 13.54 fish/1000 m2 net) lower than NPUE values based on the benthic nets in Lahjar and Ghezala reservoirs, respectively (RDA, F = 3.41, p < 0.05). The NPUE and the WPUE were different in the two studied reservoirs (RDA, F = 12.29, p < 0.001).

Ghezala reservoir is poor in terms of species diversity and abundance compared to Lahjar (RDA, F = 12.29, p < 0.001). The total number of roach caught (31) in Ghezala reservoir is much higher than for mullet (11) barbel (8) and pike-perch (1), an observation similar to that found for Lahjar. For benthic nets, the number of roach (28) caught is the highest. Catch rates for barbel (2.103 kg) is close to that of mullet (1.930 kg). Numerical and weight percentages in benthic nets were, respectively, 58.33 and 18.67 % for roach, 22.92 and 37.92 % for mullet, and 16.67 and 41.32 % for barbel. The lowest numerical and weight percentages were observed for pike-perch (2.08 and 2.08 %, respectively). For pelagic nets, only three specimens of roach were caught, with a total weight of 0.127 kg. WPUE for barbel and mullet were similar (2.190 kg/1000 m2 and 2.010 kg/1000 m2, respectively), while lower for roach and pike-perch (1.121 kg/1000 m2 and 0.110 kg/1000 m2, respectively). Population structure

Fig. 4 Distribution of catches in Ghezala reservoir

The size of the roach in the catches ranged from 14.5 to 17 cm and from 17 to 27 cm in the Ghezala and Lahjar reservoirs, respectively (Fig. 5). A significant nonlinear relationship was observed between roach abundance and the different species sizes (GAM, F = 23.2;

364


Page 6 of 11

Table 1 Catch and biomass of freshwater fish expressed, respectively, as number per unit effort (NPUE) and weight per unit effort (WPUE) in Ghezala and Lahjar reservoirs Gillnets

Benthic

Reservoir

Lahjar

Ghezala

Species

NPUE (fish/1000 m2)

WPUE (g/1000 m2)

Roach

432

20,177.78

39

1319.44

Mullet

24

5908.33

15

2680.56

11

2920.83

31

2202.78

1

147.22

Total

486

28,288.89

67

7068.06

Roach

125

16,175.00

13

529.17

Mullet

75

14,633.33

Barbel Pike-perch Pelagic

Pike-perch Total

NPUE (fish/1000 m2)

WPUE (g/1000 m2)

8

716.67

Total

208

31,525.00

13

529.17

Roach

388

19,605.95

32

1121.88

Mullet

31

7154.76

11

2010.42

8

2190.63

Barbel Pike-perch Total

27

1990.48

1

110.42

446

26,238.10

53

5433.33

p < 0.001, Fig. 3). The covariate (size) accounted for 80 % of the deviance in variability. The plot (Fig. 5) accentuates high roach abundance in the 18–22-cm size class. The large number of small meshes used in the sampling campaign induced a high numerical abundance but a fairly modest biomass. All fish captured were adults, exceeding the size of sexual maturity which, in this species, is reached at the age of 1 year, with size ranging from 7.8 to 8.5 cm (Djemali 2005). Two age groups are present in the collected samples. The juvenile stages are not present in catches in view of the fact that the minimum mesh size used is 18 mm. Introduced in the Ghezala and Lahjar reservoirs as a forage fish for pike-perch, roach appear to have found acceptable conditions for the completion of their life cycle. The very low size of this species recorded in Ghezala reservoir may be explained by its recent introduction (January 2012). In the present study, we can confirm that roach is the most abundant species of the Cyprinidae families in Tunisian reservoirs. Fish sizes in the mullet catches ranged from 16.5 to 27 cm and 22 to 30 cm in Ghezala and Lahjar reservoirs, respectively (Fig. 6). In our study, the sampled population is composed of two age groups in each reservoir. A significant nonlinear relationship was observed between mullet abundance and the different species sizes (GAM, F = 3.91; p = 0.04). The covariate (size) accounted for

33 % of deviance in variability. The plot (Fig. 6) accentuates the most abundant size class of mullet which is 25–30 cm. The trophic conditions and physicochemical characteristics of water appear more favorable for mullet growth in Lahjar reservoir than that in Ghezala. Young fry are observed in the shallow shoreline zones which have not been prospected with nets.

Fig. 5 Generalized additive model depicting relationship between roach abundance and size. The solid line is the predicted value of the dependent variable as a function of the x-axis. The dotted lines are ± two standard errors


Page 7 of 11 364

classes of pike-perch were caught in Lahjar. The majority of specimens were adults given that this fish first reaches sexual maturity at the age of 1 year, corresponding to a size of 24 cm and a weight of 90 g (Toujani et al. 2000). The record for the lowest density of all the species in Ghezala reservoir, observed in pike-perch, may have been due to its recent introduction. In fact, the study conducted in Ghezala reservoir in 2010 recorded no presence of pike-perch (CTA 2011). Fish sizes in the barbel catches ranged from 27.5 to 33.5 cm in Ghezala reservoir whereas no specimen was caught in Lahjar (Fig. 8). Additionally, no significant effect of size class on the barbel abundance was detected.

Fig. 6 Generalized additive model depicting relationship between mullet abundance and size. The solid line is the predicted value of the dependent variable as a function of the x-axis. The dotted lines are ± two standard errors

Fish size in the pike-perch catches ranged from 22 to 52 cm in Lahjar reservoir, while a single specimen, 28.5 cm in length, was caught in Ghezala (Fig. 7). A significant nonlinear relationship was observed between Pike-perch abundance and the different species sizes (GAM, F = 4.72; p = 0.001,). The covariate (size) accounted for 57 % of deviance in variability. The plot (Fig. 7) accentuates high pike-perch abundance in the 25–30-cm and 52-cm size classes. Altogether three age

Fig. 7 Generalized additive model depicting the relationship between pike-perch abundance and size. The solid line is the predicted value of the dependent variable as a function of the x-axis. The dotted lines are ± two standard errors

Discussion Fish biomass distribution in different size-classes appears to be influenced directly by local factors such as water quality and reservoir morphometry (area and depth). Precision in abundances and biomass depends on fish density and thus on the trophic status of the sampled reservoir (Deceliere-Vergès et al. 2009). Lake morphology and geographical position were important determinants of fish assemblages (Mehner et al. 2007). A similar study from Lake Balaton (Hungary) shows that most variance in the gill net catch-per-unit-effort data was essentially associated with water transparency which can influence gill netting efficiency in at least two ways. Firstly, in turbid water, the probability that a fish might see the net before being entangled is lower than in clear water. Secondly, the activity of most fish is influenced by light intensity, being highest in low light though diminished in complete darkness (György et al. 2012). Holmgren and Appelberg (2000) demonstrated that the

Fig. 8 Length-frequency distribution of barbel in Ghezala and Lahjar reservoirs; note that no specimen was caught in Lahjar

364

Page 8 of 11

size structure of benthic freshwater fish communities was in relation to environmental gradients. Also, fish community composition has been shown to be determined mainly by maximum and mean depth, chlorophyll a content, and reservoir volume (Mehner et al. 2005). The fish community has preferences and optima along environmental gradients in terms of density and biomass (Prchalová et al. 2008). As observed in this study, fish are rarely distributed homogenously or randomly within their environments; distribution is usually determined by the physical (habitat), along with chemical and historical constraints of environmental conditions (Benson and Magnuson 1992; Borcard et al. 1992). Furthermore, biotic interactions such as competition in a waterbody can greatly modify the spatial distribution of fish (Diehl and Eklov 1995; Mehner et al. 2005). Carol et al. (2006) observed clear effects of water quality on the fish species composition in Spanish reservoirs, but no significant effect of water quality on overall fish richness or diversity. Other authors have shown that species composition and other fish community attributes respond better to environmental degradation and that a response of fish richness and diversity to variation in water quality should not be taken for granted in freshwater ecosystems (Carol et al. 2006). Fish richness was low in Ghezala and Lahjar reservoirs because, as in most Tunisian reservoirs, most of the species were artificially introduced (Mili et al. 2015a). Our results are similar to those of Vašek et al. (2009), indicating that the majority of fish were captured in the surface layer (0–1.5 m). According to the study of Djemali et al. (2010), the distribution of the freshwater fish biomass is governed especially by depth. Studies related to analyses of variation in gillnet catches and vertical distribution of fish in European reservoirs show that during daytime, roach are most abundant in the layer just above the thermocline (3–4.5 m), whereas at night they occupy the upper part of the epilimnion (0–3 m) (Vašek et al. 2009), which is also the case in our reservoirs. In addition, pike-perch were concentrated in the lower epilimnion and the upper metalimnion (1.5–6 m). However, these authors indicate that individuals of the two species made diel migrations between pelagic and littoral habitats, which are inconsistent with our results. According to a biomass assessment study in Tunisian reservoirs (Djemali et al. 2009), downstream fish density was higher than that in the middle and upstream strata, while the biomass was lower, consistent with big fish occupying deep water and small fish occupying shallower upstream areas.


Studies by Djemali et al. (2009), conducted in Tunisian reservoirs, show that densities vary considerably between surface and deeper layers because of the spatial and temporal mobility of fish populations with no clear behavioral tendencies. Jeppesen et al. (2006) revealed distributional patterns similar to those in the present study. Our results are in concordance with other studies (Djemali et al. 2009, 2010) indicating that fish biomass and densities were higher downstream than upstream because the greater depths were located near the dam. This observation is available only for shallow reservoirs such as Ghezala and Lahjar. A similar pattern is seen in several previous studies of lakes and reservoirs (Fischer and Eckmann 1997; Romare et al. 2003; Lewin et al. 2004; Prchalová et al. 2006; Lauridsen et al. 2008). Absence of small mesh in our nets resulted in the deficiency of small fish in our catch rates. These results are in accordance with the study by Dekar and Magoulick (2007), in which fish density was negatively related to maximum depth, and with the study by Power (1987) who asserted that increasing maximum depth may result in greater predation risk from larger aquatic predators. The same observation was made in a study of Danish lakes where fish are concentrated in the littoral/benthic part of the upper two depth strata (0–6-m depth) (Lauridsen et al. 2008). In the Danish study, increasing abundances of, for example, roach and pike-perch occurred in the littoral zones (Lauridsen et al. 2008), a distribution pattern which was also in accordance with the general pattern observed in the Tunisian reservoirs. Additionally, Achleitner et al. (2012) indicate that benthic and pelagic gillnet catches were similar in the epilimnetic water layer. Pelagic biomass estimates peaked near the thermocline, which was detected at a lower depth using hydroacoustics. Referring to the study of Winfield et al. (2009), analysis of size distribution data shows that relatively more middle-sized fish were caught with gill nets than with other fishing methods such as electrofishing. The biggest detected fish in our reservoirs were found near the dam (i.e., downstream) while the smallest were close to the tributary (i.e., upstream). Small fish may benefit from the better trophic conditions, but the tributary area may also be preferred because some species populating these two reservoirs (pike-perch and barbel) may use the upstream area for spawning (Hladık and Kubecka 2003; Vašek et al. 2006). Our results are in accordance with the study by Carol et al. (2006) showing that nutrient concentrations and productivity of freshwater ecosystems


increase along the river continuum and that downstream reservoirs should be expected to be more eutrophic (Frossard et al. 2014). For our reservoirs, water quality data are not yet available, but it has been suggested that eutrophication probably increases with the age of the reservoir, affecting fish assemblages that are dominated by barbel in older reservoirs (for Ghezala and Lahjar reservoirs: 1984 and 1996, respectively). A clear difference in fish abundance between the littoral and the pelagic zones has been found particularly in more nutrient-rich lakes (Lauridsen et al. 2008). Jespersen et al. (2000) confirmed that changes occurred in the fish community structure along with the nutrient gradient. For example, trout CPUE by numbers and weight decreased with increasing nutrient concentration. In the literature, the percentage of piscivores is suggested as a useful indicator of water quality (Søndergaard et al. 2005). Consequently, the potential piscivore biomass, including all size classes of pike-perch, measured as a percentage (6.1 and 1.9 %, in Lahjar and Ghezala reservoir, respectively) of the total fish biomass indicates an imbalance in the food chain. The fish communities were composed mostly of cyprinids, predominantly roach (R. rutilus), as is the case in most lakes in France (Deceliere-Vergès and Guillard 2008). These authors indicate that the presence of most species populations in the pelagic zone can be best assessed by hydroacoustic surveys, due to the selectivity bias that generally occurs when using passive gear for younger fish. Benthic gillnet fishing has detected, on average, 92.9 % of species and provided the best representation of fish species in reservoirs. Certain species such as the European eel, A. anguilla (L.), or the European catfish, S. glanis L., are not easy to catch using gill nets because of their morphology, behavior, or their preferred habitat in lakes (Jeppesen et al. 2006). Although other studies suggest that pelagic gill nets do not yield any additional information on species composition compared to benthic gill nets (Deceliere-Vergès and Guillard 2008), this was not the case in the present study. Gill nets represent a passive sampling gear that is highly selective and only captures actively moving fish. Thus, estimates derived from gillnet fishing are biased because of differences in fish size, behavior, and activity (Achleitner et al. 2012). Generally, gillnet surveys provide good information on trends in fish density based on the catch per unit effort (CPUE), but gill nets cannot provide stock estimations on the entire lake, and only a few comparisons of CPUE to real population density can be attempted (Hubert

Page 9 of 11 364

1996). Lauridsen et al. (2008) suggest that the CEN standard will not give a true picture of the proportion of different fish species in lakes. This passive sampling method gives a performance largely influenced by environmental and methodological factors. Since most fish species display a diurnal activity rhythm and many of them group to shoals, the timing and duration of gillnet sampling may influence gillnetting results (Olin et al. 2009; Specziar et al. 2009; Vašek et al. 2009). This study cautions us to use gill nets carefully when analyzing trends in fish abundance. Fish sampling with multimesh gill nets has some limitations which must be considered when interpreting data. In addition, the use of multiple sampling techniques may afford additional understanding of population characteristics (e.g., recruitment variability) by providing estimates of abundance in different life history stages of particular species (Fischer and Quist 2014; Jesse and Michael 2014). These authors show that additional information using multiple gear may be crucial to determining when an insufficient number of fish may limit inferences for targeted populations (e.g., monitoring rare species).

Conclusion Reservoirs are essential to the management of Tunisia’s water resources and also in providing an important contribution to fish production in areas far from the coasts. In our study at Ghezala and Lahjar, multi-mesh gill nets inspired by the CEN (European Committee for Standardization 2005) were used for the first time in Tunisia to provide information about commercial fishing in the country’s reservoirs. Only four species were present in Ghezala and Lahjar reservoirs: barbel, mullet, pike-perch, and roach. Barbel is an autochthonous species; pike-perch and roach, once introduced in a reservoir, reproduce naturally. Mullet, however, is regularly planted near the mouths of major Tunisian rivers by the Technical Center of Aquaculture, and mullet fry is caught at various stations nearby. Both general and specific abundances obtained in Lahjar reservoir were higher than those in Ghezala. Most of the sampled fish were found in the upper water layers (0–6 m). No specimens were caught in the deep layer (9–12 m). All Tunisian freshwater fish species are abundant in the superficial layer of the deep water column near the dam. In the entire water column of Tunisian reservoirs, from the dam to the tributary, fish distribution

364

Page 10 of 11

is governed by depth. In future investigations, small and large fish distributions must be studied separately. The roach populations in Lahjar reservoir do not require support from fishery managers, whereas pike-perch must be assisted. New planting of pike-perch and roach in Ghezala reservoir must be monitored carefully. Single-method surveys such as sampling protocols, including the CEN 2005 protocol used here, are useful for obtaining comparisons of reservoir fish data, though this method does not produce the full range of fish sizes in reservoirs. Hydroacoustics can easily detect large fish which are difficult to catch with gill nets, but smaller specimens will go under-sampled. For future studies, we suggest the use of standardized fishing methods (gill nets) concomitantly with both electrofishing and hydroacoustics. Acknowledgments This work is part of a collaborative project involving the Technical Centre of Aquaculture and the Higher Institute of Fisheries and Aquaculture, Bizerte, Tunisia, and the French National Centre for Scientific Research (CNRS, ChronoEnvironnement 6249), Besançon. We would like to thank Becher Ammami and Ahmed Kassab who contributed to the success of the fishing surveys. We are also very grateful to Messeher Akkari and Ghaith Zneigi for their help which considerably improved the manuscript. We thank the anonymous reviewers whose comments have helped to greatly improve the manuscript.

References Achleitner, D., Gassner, H., & Luger, M. (2012). Comparison of three standardised fish sampling methods in 14 alpine lakes in Austria. Fisheries Management and Ecology, 19, 352–361. Argillier, C., Caussé, S., Gevrey, M., Pédron, S., De Bortoli, J., Brucet, S., et al. (2013). Development of a fish-based index to assess the eutrophication status of European lakes. Hydrobiologia, 704, 193–211. Benson, B. J., & Magnuson, J. J. (1992). Spatial heterogeneity of littoral fish assemblages in lakes, in relation to species diversity and habitat structure. Canadian Journal of Fisheries and Aquatic Sciences, 49, 1493–1500. Borcard, D., Legendre, P., & Brapeau, P. (1992). Partialling out the spatial component of ecological variation. Ecology, 73, 1045–1055. C.E.N. (2005). Water quality-sampling of fish with multi-mesh gillnets. EN 14757. C.T.A. (2011). Rapport technique des activités de l’unité de développement de la pisciculture continentale en Tunisie. Tunis: Centre Technique d’aquaculture. Carol, J., Benejam, L., Alcaraz, C., Vila-Gispert, A., Zamora, L., Navarro, E., et al. (2006). The effects of limnological features

Environ Monit Assess (2016) 188:364 on fish assemblages in fourteen Spanish reservoirs. Ecology of Freshwater Fish, 15, 66–77. D.G.P.A. (2013). Annuaire des statistiques de pêche de la Direction Générale de la pêche et de l’Aquaculture. Tunis: Direction Générale de la pêche et de l’Aquaculture. Deceliere-Vergès, C., & Guillard, J. (2008). Assessment of the pelagic fish populations using CEN multi-mesh gill nets: consequences for the characterization of the fish communities. Knowledge and Management of Aquatic Ecosystems, 389, 1–16. Deceliere-Vergès, C., Argillier, C., Lanoiselée, C., De Bortoli, J., & Guillard, J. (2009). Stability and precision of the fish metrics obtained using CEN multi-mesh gillnets in natural and artificial lakes in France. Fisheries Research, 99, 17–25. Dekar, M. P., & Magoulick, D. D. (2007). Factors affecting fish assemblage structure during seasonal stream drying. Ecology of Freshwater Fish, 16, 335–342. Diehl, S., & Eklov, P. (1995). Effects of piscivore-mediated habitat use on resources, diet, and growth of perch. Ecology, 76, 1712–1726. Djemali, I. (2005). Evaluation de la biomasse piscicole dans les plans d’eau douce tunisiens: approche analytique et acoustique. Institut National Agronomique de Tunisie: Tunisie. Thèse de Doctorat en Sciences Agronomiques, 250 p. Djemali, I., Toujani, R., & Guillard, J. (2009). Hydroacoustic fish biomass assessment in man-made lakes in Tunisia: horizontal beaming importance and diel effect. Aquatic Ecology, 43, 1121–1131. Djemali, I., Laouar, H., & Toujani, R. (2010). Distribution patterns of fish biomass by acoustic survey in three Tunisian manmade lakes. Journal of Applied Ichthyology, 26, 390–396. Fischer, P., & Eckmann, R. (1997). Spatial distribution of littoral fish species in a large European lake, Lake Constance, Germany. Archiv für Hydrobiologie, 140, 91–116. Fischer, J. R., & Quist, M. C. (2014). Characterizing lentic freshwater fish assemblages using multiple sampling methods. Environmental Monitoring and Assessment, 186, 4461– 4474. Frossard, V., Versanne-Janodet, S., & Aleya, L. (2014). Factors supporting harmful macroalgal blooms in flowing waters: a 2-year study in the lower Ain River, France. Harmful Algae, 33, 19–28. Giraudoux, P. (2012). Pgirmess: data analysis in ecology. http:// perso.orange.fr/giraudoux. György, Á. I., Tátrai, I., & Specziár, A. (2012). Relationship between horizontal hydroacoustic stock estimates and gillnet catches of surface-oriented fish in shallow Lake Balaton (Hungary). Knowledge and Management of Aquatic Ecosystems, 06, 1–17. Hladık, M., & Kubecka, J. (2003). Fish migration between a temperate reservoir and its main tributary. Hydrobiologia, 504, 251–266. Holmgren, K., & Appelberg, M. (2000). Size structure of benthic freshwater fish communities in relation to environmental gradients. Journal of Fish Biology, 57, 1312–1330. Hubert, W. A. (1996). Passive capture techniques. In B. R. Murphy & D. W. Willis (Eds.), Fisheries techniques. USA: American Fisheries Society.. Jeppesen, E., Pekcan-Hekim, Z., Lauridsen, T. L., Sondergaard, M., & Jensen, J. P. (2006). Habitat distribution of fish in late

Environ Monit Assess (2016) 188:364 summer: changes along a nutrient gradient in Danish lakes. Ecology of Freshwater Fish, 15, 180–190. Jespersen, E., Lauridsen, T. L., Mitchell, S. F., Christoffersen, K., & Burns, C. W. (2000). Trophic structure in the pelagial of 25 shallow New Zealand lakes: changes along nutrient and fish gradients. Journal of Plankton Research, 22, 951–968. Jesse, R. F., & Michael, C. Q. (2014). Characterizing lentic freshwater fish assemblages using multiple sampling methods. Environmental Monitoring and Assessment, 186, 4461– 4474. Lauridsen, T. L., Landkildehusa, F., Jeppesena, E., Jørgensena, T. B., & Søndergaarda, M. (2008). A comparison of methods for calculating catch per unit effort (CPUE) of gill net catches in lakes. Fisheries Research, 93, 204–211. Lewin, W. C., Okun, N., & Mehner, T. (2004). Determinants of the distribution of juvenile fish in the littoral area of a shallow lake. Freshwater Biology, 49, 410–424. Loss, G. F., Nau, W., & Winter, M. (1991). Le développement de la pêche en eau douce dans le nord de la Tunisie. Etude effectuée dans le cadre de la coopération technique Tunisoallemande. GTZ Gmbh: Tunisie. Mehner, T., Diekmann, M., Brämick, U., & Lemcke, R. (2005). Composition of fish communities in German lakes as related to lake morphology, trophic state, shore structure and humanuse intensity. Freshwater Biology, 50, 70–85. Mehner, T., Holmgren, K., Lauridsen, T. L., Jeppesen, E., & Diekmann, M. (2007). Lake depth and geographical position modify lake fish assemblages of the European ‘central plains’ ecoregion. Freshwater Biology, 52, 2285–2297. Mili, S., Ennouri, R., Laouar, H., & Missaoui, H. (2015a). Fisheries in the Tunisians dams: diagnosis of the current situation and development opportunities. FAO Fisheries and Aquaculture Proceedings, 39, 95–106. Mili, S., Ennouri, R., Laouar, H., & Missaoui, H. (2015b). Etude de l’âge et de la croissance chez deux espèces de Mugilidae (Mugil cephalus et Liza ramada) dans trois retenues de barrages en Tunisie. Bulletin de la Société zoologique de France, 140(3), 181–197. Oksanen, J., Blanchet, G., Kindt, R., Legendre, P., O’Hara, R., Simpson, G., Solymos, P., Stevens, H., & Wagner, H. (2011). Vegan: community ecology package (pp. 11–17). R package version 1. http://CRAN.R-project.org/package0vegan Olin, M., Malinen, T., & Ruuhijarvi, J. (2009). Gillnet catch in estimating the density and structure of fish community— comparison of gillnet and trawl samples in a eutrophic lake. Fisheries Research, 96, 88–94. Power, M. E. (1987). Predator avoidance by grazing stream fishes in temperate and tropical streams: importance of stream depth and prey size. In W. C. Kerfoot & D. C. Heins (Eds.), Predation: direct and indirect impacts in aquatic communities. Hanover: University Press of New England. Prchalová, M., Kubecka, J., Hladik, M., Hohausová, E., Cech, M., & Frouzová, J. (2006). Fish habitat preferences in an artificial

Page 11 of 11 364 reservoir system. Verhandlungen des Internationalen Verein Limnologie, 29, 1890–1894. Prchalová, M., Kubecka, J., Vasek, M., Peterka, J., Seda, J., Juza, T., et al. (2008). Distribution patterns of fish in a canyonshaped reservoir. Journal of Fish Biology, 73, 54–78. R Development Core Team (2012). R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing ISBN 3-900051-07-0. Romare, P., Berg, S., Lauridsen, T., & Jeppesen, E. (2003). Spatial and temporal distribution of fish and zooplankton in a shallow lake. Freshwater Biology, 48, 1353–1362. Simmonds, J., & MacLennan, D. (2005). Fisheries acoustics: theory and practice, Fish and aquatic resources series 10 (2nd ed.). Oxford: Blackwell Scientific Publications. Søndergaard, M., Jensen, J. P., & Jeppesen, E. (2005). Seasonal response of nutrients to reduced phosphorus loading in 12 Danish lakes. Freshwater Biology, 50, 1605–1615. Soudoud. (2006). Base de données spécifique des grands barrages en Tunisie. Ministère de l’agriculture et des ressources hydrauliques, Direction générale des barrages et des grands travaux hydrauliques. Specziar, A., Tibor, E., Agnes, I. G., Tatrai, I., & Bıro, P. (2009). A comparison between the benthic Nordic gillnet and whole water column gillnet for characterizing fish assemblages in the shallow Lake Balaton. Annales de Limnologie— International Journal of Limnology, 45, 171–180. Tlili, S., Jebali, J., Banni, M., Haouas, Z., Mlayah, A., Helal, A. N., et al. (2010). Multimarker approach analysis in common carp Cyprinus carpio sampled from three freshwater sites. Environmental Monitoring and Assessment, 168, 285–298. Toujani, R., Missaoui, H., & Romdhane, M. S. (2000). Cycle sexuel des femelles de sandre (Stizostedion lucioperca) dans la retenue du barrage de Sidi Salem. Bulletin Institut National Sciences Technologies Mer, 27, 85–96. Vašek, M., Kubečka, J., Matěna, J., & Seda, J. (2006). Distribution and diet of 0+ fish within a canyon-shaped European reservoir in late summer. International Review of Hydrobiology, 91, 178–194. Vašek, M., Kubečka, J., Čech, M., Draštík, V., Matěna, J., Mrkvička, T., et al. (2009). Diel variation in gillnet catches and vertical distribution of pelagic fishes in a stratified European reservoir. Fisheries Research, 96, 64–69. Winfield, I. J., Fletcher, J. M., Ben James, J., & Bean, C. W. (2009). Assessment of fish populations in still waters using hydroacoustics and survey gill netting: experiences with Arctic charr (Salvelinus alpinus) in the UK. Fisheries Research, 96, 30–38. Wood, S. N. (2006). Generalized additive models: an introduction. New York: CRC. Züur, A., Ieno, E., Walker, N., Saveliev, A., & Smith, G. (2010). Mixed effects models and extensions in ecology with R. Statistics for biology and health. Springer: New York.