European eel distribution and body condition in a river floodplain: effect of longitudinal and lateral connectivity

We studied eel population characteristics (size classes, densities and body condition) in the lower Loire River floodplain (France) to evaluate the effects of longitudinal and lateral gradients. A total of 36 sites were electrofished in June 2005. The sites were grouped first into three river segments according to the distance inland corresponding to a longitudinal gradient and secondly into three lateral connectivity categories. Results indicate that small eels, especially those ≤300 mm, were very abundant downstream but density rapidly decreased upstream. In addition, eels ≤150 mm tended to be most abundant in connected waterbodies. Conversely, the distribution of larger eels was quite uniform across both dimensions. Eel condition decreased upstream. In parallel, in downstream river segment, eels had a lower condition in disconnected waterbodies than in connected ones. Eel specialists still have only a longitudinal perception of eel habitat. Our results suggest that lateral gradients should be also taken into consideration

European eel distribution and body condition in a river floodplain: effect of longitudinal and lateral connectivity Introduction The European eel Anguilla anguilla (L.) is highly threatened throughout its distribution range (Moriarty & Dekker 1997) and management actions are needed to restore the population (Feunteun 2002;Baisez & Laffaille 2005; Commission of the European Community 2007). However, several studies have emphasised that the quality of the stock should be also taken into account (EELREP 2005;Acou et al. in press). For instance, pollution or parasitism might severely lower individual fitness by reducing the fecundity or the probability of reaching the Sargasso Sea (Robinet & Feunteun 2002;EELREP 2005;Acou et al. 2008). Thus, restoration studies should focus on the relationship between the environment (i.e., quantity and quality of habitats) and the eel population characteristics (i.e., eel distribution pattern and quality) during the growth phase (i.e., the yellow eel stage) (Baisez & Laffaille 2005).
Most studies that deal with yellow European eels in large river systems show that distance from the sea (Ibbotson et al. 2002;Feunteun et al. 2003;Lasne & Laffaille 2008) and ⁄ or the presence of barriers to upstream migrations (Feunteun et al. 1998;Briand et al. 2005;Lasne & Laffaille 2008) are the most important determinant factors for density and size structure of the population Acou 2006). Generally, density decreases upstream as mean size increases. In addition, eel life-history traits usually co-vary with this spatial pattern because most of the environmental factors (temperature, salinity, productivity, etc.) that influence eel growth, sex determination, size and age at maturity (Krueger & Oliveira 1999;Acou et al. 2003;Feunteun et al. 2003;Davey & Jellyman 2005;Melià et al. 2006) are structured along the longitudinal gradient (Amoros & Petts 1993). Typically, males occupy the most downstream reaches, grow quickly and mature early at a smaller size, whereas females develop slowly in upstream reaches and mature later at a larger size. Thus, in large systems, river segments can be separated on the basis of both eel density and the dominant phenotypes.
However, the actual decrease in the eel stock together with the reduction in longitudinal connectivity in anthropised systems means that eels are now mainly located in the most downstream areas (Lasne & Laffaille 2008). In such a situation, a focus on patterns and processes in downstream areas, especially in large river floodplains, became very relevant. Curiously, unlike most other specialists of river ecology since the development of the Flood Pulse Concept (Junk et al. 1989), eel specialists still continue to largely ignore the process occurring in the lateral dimension of river floodplains. However, very strong environmental gradients might exist there, and could influence eels patterns . First, variable hydrological connectivity along the floodplain lateral gradient leads to great habitat heterogeneity (Amoros 2001;Amoros & Bornette 2002;Tockner et al. 1999;Lasne et al. 2007). In relatively pristine systems, it is possible to find a variety of aquatic habitats ranging from lotic (main channel and side arms) to lentic (disconnected waterbodies) that lead to contrasting environmental conditions (e.g., in terms of water velocity, aquatic vegetation cover, substrate size, temperature or conductivity). Numerous studies have shown that fish distribution across the floodplain is highly heterogeneous and related to individual species requirements, with species velocity preference being crucial (Copp 1989;Aarts et al. 2004;Lasne et al. 2007). Secondly, connectivity with the main channel is likely to lead to variable habitat accessibility. For instance, in an Amazonian floodplain, Granado-Lorencio et al. (2005) showed that the connectivity level influences the distribution patterns of migratory fishes. Thus, the lateral distribution of eels across a large floodplain is likely to be heterogeneous. Moreover, thermal conditions and food availability (especially invertebrates; Amoros & Petts 1993;Garcia & Laville 2000;Reckendorfer et al. 2006) along the lateral gradient can lead to fish having contrasting growth rates (Ribeiro et al. 2004). With eels, some studies have suggested that variable growth rates caused by habitat heterogeneity also exist (Chisnall 1989;Domingos et al. 2006). Therefore, it is likely that contrasting eel characteristics could be observed in large floodplains.
In this study, we investigated eel distribution patterns in a 140 km-long section of the lower Loire River floodplain. A recent study carried out in this floodplain showed that both habitat features and fish assemblages were very contrasted in relation to a connectivity gradient (Lasne et al. 2007). More particularly, isolated waterbodies had a lower species richness and hosted mainly stress-tolerant (i.e., eurytopic) species. Therefore, we first tested whether various connectivity levels influence eel distribution. More specifically, it was supposed that lower connectivity levels should reduce habitat accessibility and ⁄ or quality and consequently eel density. Secondly, we examined whether hydrological connectivity influences eel body condition (a proxy of fitness). It was expected that eels in isolated waterbodies have a lower condition, in relation to harsher conditions and ⁄ or reduced movement opportunity.

Study area
The study was undertaken in the downstream reaches of the Loire River, France (Fig. 1). The position of the Loire catchment on the Atlantic coast enables it to be extensively colonised by glass eels (Baisez & Laffaille 2005). In this system, before entering freshwater glass eels have to pass up a long estuary (55 km 2 including 32 km 2 always inundated) where they are subjected to intensive fishing pressure. In this river system, eels can find growing areas in brackish estuarine reaches and therefore, they might never penetrate into freshwater (Daverat & Tomas 2006).
The most downstream section of the study area was subjected to tidal influence and the downstream limit was located near the limit of salt water intrusion in the main channel (this limit corresponds to extreme and therefore rare events, i.e., low flow, high tide ⁄ spring tide). This sector roughly matches with the tidal freshwater estuary (Daverat & Tomas 2006). Conversely, the rest of the study sector was outside tidal influence. Water within the study sites was fresh (i.e., salinity close to 0) including the two lower ones located just downstream the limit of salt water intrusion. Migrating eels are free to move upstream as there are no barriers affecting the study area (Lasne & Laffaille 2008). The floodplain is large, relatively little impacted and has numerous kinds of waterbodies more or less connected to the main channel (Lasne et al. 2007).
We sampled a total of 36 waterbodies (Fig. 1). In order to test the effect of the longitudinal gradient, we divided the study area into three segments (Fig. 1). Segment A included all sites located downstream of the limit of tidal influence, i.e., in the tidal freshwater part of the river. Segment B was located between the tidal limit and the confluence with the Maine River, near Angers. The upstream limit of segment C was located at the confluence with the Vienne River, near Saumur.
To test the effect of the position of the waterbodies along the lateral gradient, sampling sites were classified according to three levels of connectivity, which were based on (i) their type of connection to the mainstem, (ii) the frequency of connection and (iii) their accessibility during the sampling period (Figs 1 and 2; modified after Lasne et al. 2007). Waterbodies of type 0 and 1 were disconnected from the main channel at sampling period (early June 2005). However, type 0 corresponds to sites only connected during high winter flows and water levels, whereas type 1 corresponds to sites highly accessible during a large period of the year. This type corresponds to secondary channels or backwaters that have temporary disconnections at low water levels. Finally, type 2 corresponds to sites highly accessible, i.e., eels are free to move in and out of the waterbody.
Such a connectivity gradient across the floodplain might regulate not only the movement opportunity of fish, especially migratory species such as eel, but also environmental conditions. Lasne et al. (2007) showed for instance that aquatic vegetation cover or substrate size varied according to connectivity, with connected sites having little vegetation cover and having sandy substrate, whereas isolated sites tend to have high levels of vegetation and silty substrates. Incidentally, the sandy-bottom connected sites would be expected to be more suited to cryptic behaviour in small, compared with large, eels. However, shelters for large eels (i.e., boulders, riprap banks, roots or woody debris) might also be present in waterbodies irrespective of the connectivity level.

Eel sampling
Sampling was conducted in early June 2005. At this time of the year, water levels are normally low, and heterogeneity across the floodplain is high (Ward & Tockner 2001;Lasne et al. 2007;Thomaz et al. 2007). Eels were collected by wading with electroshocker (FEG 8000, EFKO, Leutkirch, Germany; DC, 300-600 V, 6-8 A). Electrofishing by wading meant that deeper waters (>1.20 m) were not sampled, but this method is in any case poorly efficient in  such habitats. We used a 30 cm diameter anode on 2 m long pole following the point abundance samples (PAS) methodology (Nelva et al. 1979). At each site, a total of 25-35 PAS was made in the various habitats present to obtain a reliable estimate of the relative abundance and structure of the eel stocks (see Laffaille et al. 2005b). Fish were anaesthetised in the field with clove oil and total body length was measured to the nearest mm. Individuals ‡200 mm were also weighed (g ± 1) for body condition analysis. Smaller eels were not weighed because of potential errors in weight measurement on the field. According to the characterisation methods of Acou et al. (2005Acou et al. ( , 2006, none of the eels sampled could be considered as being silver eels so they were all in the colonising or growing stage. All individuals were released after collection of biological data.

Distribution patterns
The distribution of eels was analysed along the longitudinal and lateral gradients by grouping sites according to river segment (A, B or C) and connectivity (0, 1 or 2). Thus, we obtained nine groups: A0, A1, A2, B0, B1, B2, C0, C1 and C2 (Table 1). We used the mean number of eels per PAS (i.e., catch per unit effort; eel PAS )1 ) as an index of density. We first considered all size classes together. Secondly, based on the assumption that behaviour and habitat requirements change with eel size, we examined the distribution for different size classes as did Lasne & Laffaille (2008). However, given the low abundance of eel >650 mm, we pooled all individuals >450 mm together. Finally, the data set was then split into four size classes: £150, 151-300, 301-450 and >450 mm. Homogeneity of density among groups was evaluated using Kruskal-Wallis (KW) tests with a = 0.05.

Body condition
We followed the protocol of Vila-Gispert et al. (2000) to compare eel condition among groups (i.e., A0-C2). We preferred this method to the classical index of Fulton (Cone 1989) because it totally removes the effect of size on body condition (Cone 1989;Packard & Boardman 1999). Indeed, in most fish, including eels (Melià et al. 2006), the index of Fulton increases as length increases. In a recent study, the protocol of Vila-Gispert et al. (2000) has been successfully applied to make comparisons of eel condition (Acou et al. 2008). We, therefore, applied an ancova procedure with mass as the dependent variable and length as a covariate. Mass-length relationships were linearised by ln-transformation. First, we tested the homogeneity of slopes among groups. If the interaction between length (i.e., covariate) and group (factor) was not significant (P < 0.05), secondly, a standard ancova was applied to compare the y-intercept of the mass-length relationships among groups. Post hoc test (HSD Tukey) was used to identify groups with different intercepts. All analyses were performed using R software (Ihaka & Gentleman 1996).

Results
A total of 1006 PAS was performed, and 1029 eels ranging from 53 to 849 mm (average 234 ± 142 mm) were caught.

Size class distribution patterns
The total density of eels (i.e., all size classes pooled) was significantly different among site groups (KW = 27.388; P < 0.001; Fig. 3). Densities were higher in river segment A irrespective of the connectivity level and decreased upstream. In A, densities tended to be lower in disconnected sites compared with connected ones, and above all, densities were more variable as shown by error bars. This was not true in other compartments.
When considering size classes separately, it appeared that total density patterns were mainly influenced by small eels (£150 and 151-300 mm) patterns which indicated a heterogeneous distribution (KW = 28.138; P < 0.0001 and KW = 27.628; P < 0.001, respectively) in the longitudinal gradient. For these two size classes, densities were highest in segment A and decreased upstream. In particular, eels £150 mm were almost totally absent from segments B and C, and 151-300 mm eels were also almost absent from segment C. Density of eels >300 mm was always low and constant over waterbody groups (KW = 9.915; P = 0.271 and KW = 11.742; P = 0.163, respectively, for 301-450 mm eels and for eels >450 mm). Only the density of eels £150 mm tended to be heterogeneous along the lateral gradient, but only in compartment A (Fig. 3). Indeed, density was always high in connected waterbodies, whereas it was on average lower and more variable in isolated ones.

Body condition patterns
Among the 1029 eels caught in the 36 sampling sites, only 510 eels ‡200 mm were individually weighed (Table 1) for mass-length relationship analyses. Unfortunately, not enough individuals (<10) were available in group C0 (Table 1) due to very small densities in this upstream compartment, especially in disconnected waterbodies (Fig. 3). Other groups had enough data and were therefore included in the ancova procedure. In the first model (model 1; Table 2), the interaction between covariate (i.e., length) and factor (i.e., group) was not significant, indicating that the slopes of the regressions were homogeneous. The second model (model 2, without interaction; Table 2) showed that y-intercepts of the mass-length relationship were different between groups and pairwise significant differences were found between groups (Fig. 4). When considering sampling sites of types 1 and 2, fish condition decreased significantly upstream. Regardless of the river segment, there were no significant differences between condition of eels in sites of connectivity type 1 and 2. However, in segment A,  eels in the most isolated sites (type 0) had a significantly lower condition, almost equivalent to the condition of eels found in the uppermost reaches (i.e., in compartment C).

Longitudinal patterns
Most studies that deal with European eel distribution at a catchment or river scale show a decrease in density or occurrence along the longitudinal gradient (Ibbotson et al. 2002;Feunteun et al. 2003;Briand et al. 2005). However, this pattern might not occur in very small systems where young recruits might invade the whole catchment and reach upstream reaches after their first colonisation season Acou 2006). In addition, the mean size of eels increases upstream as density of small eels decreases ). In our study, similar patterns were found despite the relatively small length of the study sector (140 km compared with the 1012 km of the total length of the Loire River), closeness to the estuary (segment A is subjected to the tidal influence) and the absence of barriers to migration. In segment A, most eels collected were £300 mm. According to Feunteun et al. (2003), the upper estuary roughly corresponds to the high-density area where young recruits accumulate before migrating to peripheral low density areas (i.e., compartments B and C). Under the assumption of density-dependent upstream migratory behaviour (Moriarty 1986;Feunteun et al. 2003), high densities in downstream areas should lead to a significant colonisation of upstream reaches (Ibbotson et al. 2002). Conversely, the reduction in recruitment and density downstream should reduce the migratory behaviour at both an individual and a population level (Edeline 2007). In our study, river segment C was located between km 40 and km 100 from the tidal limit but was free of small eels (e.g., £300 mm) whereas it is highly accessible for such individuals (based on a mean migration rate of 10-75 km a year; Aprahamian 1988; Baras et al. 1996;Briand et al. 2005). Although the colonisation of tributaries (including the large Maine River) probably contributes to the decreasing density along the main axis of the Loire and its floodplain, this result suggests that the fluvial recruitment in this catchment is low and the eel population in the downstream reaches of the Loire River is far below its potential carrying capacity.
The patterns described along the study sector are consistent with the results of Lasne & Laffaille (2008). Using long-term presence-absence data from the entire Loire catchment, these authors showed that occurrence probabilities of eels £300 mm, and especially of eels £150 mm, decrease very rapidly upstream (e.g., occurrence probability of £150 mm eels is null upstream of the confluence with the Maine catchment). Conversely, larger eels tended to go further inland. Although presence-absence data and catch per unit effort do not exactly provide the same information and sampling protocols of the two studies were quite different, our results indicate that the longitudinal structure of the stock of juvenile yellow eels (i.e., £300 mm eels) in the Loire floodplain provides a good indication of the colonisation level in the entire catchment. Quantitative sampling of eels is not possible in the deeper parts of large rivers like the Loire, but it appears that electrofishing in type 2 waterbodies can reveal useful information about colonisation by small eels -recognising that larger eels might be present in deeper waters and consequently, their density is probably underestimated. In addition, whilst studies of type 1 and 0 lateral waterbodies can also be useful, it needs to be pointed out that variability can be high, depending on the distance from the main channel, temporal connection history, waterbody permanency, etc. (e.g., see Lasne et al. 2007).
The analysis of the body condition revealed that eels were relatively fatter in the lower reaches of the study area which was located in the tidal freshwater zone of the estuary, whereas segments B and C were outside tidal influence. According to Mounaix & Fontenelle (1994), Acou et al. (2003) and Daverat & Tomas (2006), conditions in coastal and estuarine brackish waters lead to a higher and more constant eel growth than straight riverine reaches. Indeed, estuaries are generally highly productive, and in addition, offer buffered conditions, especially in terms of temperature (Mounaix & Fontenelle 1994;Morrison & Secor 2003;Costa et al. 2008). Although water was fresh in segment A, it is very likely that the higher body condition of individuals located in this segment is the result of the proximity of brackish waters which probably provides good environmental conditions for eel growth. Indeed, in segment A, most eels are £300 mm and considered as active colonisers Lasne & Laffaille 2008), therefore, probably arising straight from downstream productive areas and undertaking river colonisation with a high body condition. It is also possible that 'nomadic' (according to the description of Feunteun et al. 2003) eels in downstream locations move from the riverine compartment to the estuarine one and ⁄ or vice versa (Daverat & Tomas 2006;. For instance, eels in type 2 waterbodies might be able to disperse and forage in the main river channel. Conversely, upstream compartments are mainly occupied by large individuals that exhibit little migratory behaviour, eels described as 'sedentary' by Laffaille et al. (2005a).

Lateral distribution patterns
To date, no study has clearly addressed the question of eel distribution patterns along a lateral connectivity gradient. In the Loire floodplain, the density of eels was high in all connected waterbodies of the tidal freshwater sector of the Loire (segment A). However, this was mainly due to the density of small eels which was high in connected waterbodies (type 2) and tended to decrease in isolated ones (type 1 and then 0). Except for small eels, density of all size classes was quite homogeneous among connectivity types irrespective of the river segment. Two synergic but not exclusive mechanisms might explain such contrasting patterns. As mentioned earlier, there is a double gradient across large floodplains: a gradient of accessibility (Granado-Lorencio et al. 2005) and a gradient of habitat condition (Aarts et al. 2004;Tockner et al. 1999;Lasne et al. 2007) with each possibly playing an important role.
First, small eels might be less likely to colonise disconnected waterbodies because of a low accessibility. Research that deals with eel migratory patterns has shown that eel movements mainly concern small individuals and occur in late spring when water temperature increases (Moriarty 1986;Naismith & Knights 1988;Acou 2006). In general, during this period, eels arriving from downstream reaches cannot penetrate into disconnected waterbodies (type 0) because water levels are often low and these waterbodies are already physically isolated, resulting in a lower eel density. However, as shown in Fig. 3, standard errors of density of eels £300 mm and especially £150 mm in A0 were quite high indicating a large variability. This variability was mainly due to the most downstream site of group A0 which had a high density of eels (density of £150 mm = 2.5 eels PAS )1 ). The accessibility of this site was enhanced by the existence of a little outlet which flows into the Loire in the spring and therefore allows the smallest eels to colonise the waterbody irrespective of the water level of the Loire itself. Such a special case emphasises the complexity of measuring connectivity, especially for a highly mobile species such as the European eel, and the need to consider other kinds of connection. Unfortunately, such connections (e.g., temporary brook, ditch networks) might be difficult to identify and quantify. This site aside, the mean density of £150 mm eels in other A0 sites was very low (0.16 ± 0.09 eels PAS )1 ). Conversely, most large eels have a sedentary behaviour (Laffaille et al. 2005a) and are 'home range dwellers' . Such individuals are generally older and have several opportunities through their life to colonise waterbodies of types 1 and 0 during floods. Finally, their distribution is quite homogeneous across the floodplain. Thus, our results suggest that density patterns across the floodplain could result from variable temporal accessibility.
Secondly, small eels might avoid disconnected waterbodies because they present unsuitable habitat conditions. Eels are known to be largely ubiquitous and opportunistic (i.e., they are able to use a wide range of aquatic habitats; Feunteun 1994), but strong size-related habitat preferences might exist in rivers Domingos et al. 2006) and in wetlands (Laffaille et al. 2004). Hence, it is possible that heterogeneity of small eel distribution in the lateral dimension results from habitat selection. On the contrary, density of eels >150 mm was quite homogeneous between these different habitat types suggesting that the distribution of larger eels is not significantly influenced by these habitat conditions.
Our results also show that the mean body condition of eels decreases as lateral connectivity decreases in segment A whereas no clear trend appeared in segment B. Furthermore, there were too few eels collected from group C0 to include in data analyses. This suggests that disconnected habitats are less suitable than connected ones, either because environmental characteristics are less favourable than in more connected sites (habitat condition hypothesis), or because eels in these habitats have little opportunities to colonise it (habitat accessibility hypothesis). It is also possible that eels with low body condition stop in disconnected waterbodies because they do not have enough energetic reserves to start upstream colonisation. Such hypothesis is in accordance with Edeline et al. (2006) who found that migratory behaviour was stimulatedvia hormone secretion -by high energetic reserves.

Conservation implications
Today, eel distribution is being more and more restricted to downstream reaches of rivers, first because of the decreasing recruitment, and secondly because of longitudinal connectivity alteration. Thus, in the context of eel decline, downstream floodplains of large river systems are probably of great importance for the species. Our study is the first that reveals the influence of lateral hydrological connectivity across the floodplain on the European eel. Furthermore, we hypothesise that the environmental and demographic heterogeneity observed at the floodplain scale produces variable phenotypes and tactics. According to the role of energetic reserves on the determinism of phenotype (e.g., migratory vs. sedentary; see Edeline et al. 2006;Edeline 2007), the effect of connectivity on body condition shown in this study is in accordance with this hypothesis. One can also suppose that locally contrasted density patterns and condition would lead to spatially variable sex ratio, with males being mainly produced in high density waterbodies and females in low density ones.
Although the Loire River still has a near-natural flow regime and a large floodplain with high habitat diversity, this is far from being the case for most large European systems (Tockner & Stanford 2002). Indeed, floodplains are often highly altered, and lateral connectivity has been reduced by channelisation or flow regulation (Tockner & Stanford 2002). Considering the role of environmental and demographic conditions in downstream areas as determinants of various phenotypes and tactics (Edeline 2007), the consequences of floodplain habitat loss or modification for the eel population at both small (i.e., local or catchment stock) and large (i.e., whole population) scales need to be examined in detail in further research.