Size-related and diel variations in microhabitat use of three endangered small fishes in a Mediterranean coastal stream M. C L A V E R O *†‡, F . B L A N C O -G A R R I D O †, L . Z A M O R A * J. PRE NDA † AND *Institut d’Ecologia Aqua`tica, Universitat de Girona, Facultat de Ciències, Campus Montilivi, 17071 Girona, Spain and †Departamento de Biologı´a Ambiental y Salud Pública, Universidad de Huelva. Campus Universitario de El Carmen, Avda. Andalucı´a s/n, 21071 Huelva, Spain This study analysed the microhabitat use of three endangered fish species, Andalusian toothcarp Aphanius baeticus, Iberian loach Cobitis paludica and sand smelt Atherina boyeri, in a coastal stream stretch. Plastic minnow traps were set both during daytime and at night on the bottom and at the surface. Depth and presence of effective refuge were recorded for each trap. To assess size-related changes in microhabitat use individuals of each species were classified in three size classes. The three species preferentially used bottom positions in the water column, though this behaviour was more evident in the case of Iberian loach. While large Iberian loach remained active at night Andalusian toothcarp and sand smelt were strongly diurnal, especially larger individuals. The three species showed a clear ontogenetic change in microhabitat preferences towards deeper waters. Small Andalusian toothcarp and medium-sized Iberian loach used deeper microhabitat in the presence of refuge. Large Andalusian toothcarp consistently preferred exposed microhabitat at any time. Andalusian toothcarp using refuge were smaller at any time, while Iberian loach followed this pattern only at night. The differential vulnerability of these species to different predators (aerial and aquatic; diurnal and nocturnal) could explain the observed patterns in microhabitat use. Fish tended to co-occur in microhabitats either due to habitat characteristics independently of species or due to species independently of habitat. Andalusian toothcarp segregated spatio-temporally from sand smelt and Iberian loach, but these species occurred independently of each other. According to these co-occurrence patterns, Andalusian toothcarp would be more sensitive than Iberian loach or sand smelt to interspecific interactions. Key words: activity; depth; habitat use; interspecific interactions; stream fishes; vegetation. INTRODUCTION Fish habitat requirements are determined by factors operating at multiple scales, running from hundreds or thousands of kilometres (e.g. drainage characteristics) to centimetres (e.g. presence of macrophytes) (Poizat & Pont, 1996). At the microhabitat scale of observation (Dungan et al., 2002), fishes have been shown to select habitat mainly as a function of food availability and predation ‡Author to whom correspondence should be addressed. Tel.: þ34 972 418467; fax: þ34 972 418150; email: miguel.clavero@udg.es risk (McIvor & Odum, 1988). Both factors vary during a fish’s ontogeny, causing changes in habitat preferences (Rosenberger & Angermeier, 2003; King, 2004). On the other hand, both food availability (Copp et al., 2005) and predation risk (Schlosser, 1988) can differ from daytime to night. In fact, the equilibrium of risks due to differential predation by piscivorous fishes and wading and diving terrestrial predators, which is often cited as the cause of the larger fish-deeper habitat pattern (Power, 1984; Harvey & Stewart, 1991), is likely to differ at night. It is therefore important to consider the possible ontogenetic and diel variations when analysing microhabitat use patterns by fishes (Copp & Jurajda, 1999). Few studies have analysed fish habitat use in Mediterranean streams and they have focused mainly on widespread cyprinid species (Grossman et al., 1987a, b; Rincon et al., 1992; Santos et al., 2004). This study analyses the microhabitat preferences of three small stream fish species: the Andalusian toothcarp Aphanius baeticus Doadrio, Carmona & Fernandez-Delgado, henceforth toothcarp, Iberian loach Cobitis paludica de Buen, henceforth loach, and sand smelt Atherina boyeri Risso. There is hardly any available information on the habitat requirements of these species at any scale of observation, and none on microhabitat use. Knowledge of fish habitat preferences is one of the main tools for efficient fish conservation and fisheries management. In the case of these three species the scarcity of knowledge is worrying, since they are endangered and should involve conservation actions (Doadrio, 2001). Loach and sand smelt are considered ‘Vulnerable’ (VU) in Spain following IUNC criteria (Doadrio, 2001). Toothcarp should be catalogued as ‘Critically Endangered’ (CR), since there are only eight extant populations, some of them suffering strong declines (Doadrio et al., 2002; Clavero et al., 2005a). The aim of this study was to analyse the microhabitat use of toothcarp, loach and sand smelt occupying sympatrically a coastal Mediterranean-regime stream. Passive capture methods (plastic minnow traps) were employed to assess size-related and diel variations in the microhabitat preferences of the three species, as well as the patterns of coexistence between species and size classes. MATERIALS AND METHODS S T U D Y A RE A This study was conducted in the La Vega River (southern Spain), the last water course of the Iberian Atlantic slope, which flows to the sea through a common estuary with the adjacent La Jara River (Fig. 1). La Vega River is only 13 km long and has a drainage area of c. 20 km2, is free of important effluents and does not have regulation infrastructures, thus featuring a natural flow cycle (Clavero et al., 2005b). Due to its small size, this river suffers extreme flow changes following a typical Mediterranean cycle (Gasith & Resh, 1999), being reduced to a few isolated pools during the summer. Fish microhabitat was analysed in a 600 m stretch located just upstream of the tidal section (Fig. 1). Water flow ceases each year around the beginning of July. After that moment fresh water is retained only in five pools, some of which are occasionally filled with sea water (Clavero et al., 2005a). The studied stretch covers all or most of the stream section occupied sympatrically by toothcarp, loach and sand smelt. While the tidal waters mark the lower limit of loach distribution, toothcarp was not detected in the upper pool of the study stretch until the last two surveys. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 Study site Tarifa FIG. 1. Location of the study area. The presence of both estuarine and freshwater species favours the presence of a rich ichthyofauna, with at least 11 species, among which there are no introduced species (Clavero et al., 2005b). Among these species the eel Anguilla anguilla (L.) is the most abundant piscivorous fish. Other abundant fish predators include viperine snakes Natrix maura, egrets Bubulcus ibis and Egretta garzetta and kingfishers Alcedo athis. The otter Lutra lutra also occurs but none of the three studied species is important otter prey (Clavero et al., 2004). FISH SAMPLING Eleven surveys were performed between October 2002 and August 2004. Fish microhabitat preferences were studied using cylindrical plastic minnow traps (240 mm long, 95 mm wide and 21 mm at the mouth), henceforth traps (Hubert, 1996). Traps were fixed to the stream bottom with a metal stick, through small holes (4 mm). In the last six surveys traps were set in pairs, with one trap touching the substratum and the other at the water surface. The depth (cm) of each trap and the presence of surrounding refuge (submerged or emergent macrophytes, or submerged riparian vegetation) was noted (Prenda et al., 1997). Traps were set both during daytime and at night. Night traps were set from sunset to early morning (around sunrise). On some occasions a large amount of captures caused night traps to be set some hours in the morning. Considering all captured fishes in these traps as ‘night captures’ is a conservative assumption, however, since possible morning captures would tend to ‘homogenize’ daytime and night results. Each captured fish was identified, measured for total length (LT, mm) and released. DATA ANALYSIS Captures were expressed as catch per unit of effort (CPUE), defined as the number of individuals captured per trap per hour. To assess possible size-related changes in microhabitat use, fishes were classified into three size classes (I, II and III). Intervals used in this classifications were different for toothcarp (<22, 23–28 and >28 mm) and for loach and sandsmelt (<30, 30–45 and >45 mm), since the last two species reach larger sizes. The following variables were calculated for each trap: the CPUE of each species and size class, the CPUE of each species (all sizes), the CPUE of each size class (all species) and the # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 total CPUE (all fishes). Depth measurements were converted to 5 semi-quantitative categories for the analyses (<20, 21–30, 31–40, 41–50 and >50 cm). This enabled the analysis of the role of depth in the variation of CPUE and fish size without assuming linear relationships, as would have been the case if depth had been used as a continuous variable. Patterns in microhabitat use were analysed using generalized linear models (GLMs) for CPUE (Crawley, 2002) and ANOVA for fish size (Copp & Jurajda, 1999). Since CPUE is a variable derived from count data, Poisson distribution as error and log link function were used. Significance of the different explanatory variables was checked through comparisons of the change in deviance and d.f. with a w2 distribution (Crawley, 2002). In order to use similar terms for ANOVA and GLM results, henceforth the term ‘variation’ is used instead of ‘deviance’. To allow comparable interpretations of the effects of the different independent variables among species and sizes, every independent variable and all the second order interactions were included simultaneously in each model. First, the influence of trap position (bottom-surface) and depth in the CPUE and size of the different species was analysed. All the data from surveys in which surface traps had not been set were excluded from this analysis. A second analysis assessed the role of depth and presence of refuge in the CPUE of the different size classes of the three species and in their size structure across microhabitat, both at daytime and at night. This second analysis included only data from bottom traps. By including time (day-night) the possible diel variations in the activity of fishes could be assessed, assuming that captures in passive traps would increase with increasing fish activity. The analysis of microhabitat preferences can show that certain species and size classes are using similar microhabitats. This does not imply co-occurrence, however, since similar habitats could be used at different moments. To assess possible patterns of cooccurrence among the studied species and size classes a principal components analysis (PCA) was applied to a matrix of traps x CPUEs of the different species and size classes, thus treated as nine different ‘species’. Surface traps and traps in which no individual of any of the three species had been captured were excluded from this analysis. The interpretation of principal components was made through Pearson correlation and oneway ANOVA analyses. Prior to PCA and subsequent correlation analyses CPUE data were logarithmically transformed [log10 (x þ 1)]. Only those principal components (PCs) with eigenvalues >1 were interpreted (Kaiser criterion). RESULTS SUMMARY OF CAPTURES During the study period 857 bottom traps and 395 surface traps were set, for a total sampling effort of 11 989 traps h—1. Overall 3699 toothcarps (total CPUE 0•31 individuals trap—1 h—1), 1262 loaches (0•10 individuals trap—1 h—1) and 614 sand smelts (0•05 individuals trap—1 h—1) were captured. These three species constituted 98•7% of total captures. Other species occasionally captured in traps were Iberian chub Leucisus pyrenaicus Gunther eel, grey mullets Liza spp. and Mugil cephalus L., and common goby Pomatoschistus microps Krøyer. Captured toothcarps were on average smaller (mean ± S.E. 24•5 mm ± 0•1, range 10–56 mm) than sand smelts (mean ± S.E. 33•3 mm ± 0•3, range 12–70 mm), which were smaller than loaches (mean ± S.E. 38•6 mm ± 0•3, range 14–95 mm) (ANOVA, P < 0•001). POSITION IN THE WATER COLUMN The three studied species were captured more often in bottom traps than in surface ones (Fig. 2). This difference was especially strong in the case of loach, since the GLM showed that >20% of the variation in the CPUE of the species could be # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 Log10 CPUE (individuals trap–1 h–1) 0·25 Toothcarp Loach Sand smelt 0·2 0·15 0·1 0·05 0 LT (mm) 60 n = 152 n = 28 n = 602 n = 38 n = 1993 n = 495 45 30 15 ≤20 25 35 45 >50 ≤20 25 35 45 >50 ≤20 25 35 45 >50 Depth (cm) FIG. 2. Differences in CPUE and fish size (mean ± S.E.) between bottom (*) and surface ( .) traps at different depths. Numbers of captured fish are also given. explained in terms of the position of traps. In contrast, this value was <10% for toothcarp and sand smelt. The CPUE of loach and sand smelt in surface traps did not show any clear pattern in relation to depth, but toothcarp were more frequently captured at the surface when traps were set in shallow waters (Fig. 2). Sand smelt and toothcarp occupying surface positions were significantly smaller than individuals trapped at the bottom (ANOVA, P < 0•001 in both cases), but there was no such difference in the size of captured loach. The smaller size of toothcarp and sand smelt captured at the surface was a pattern maintained through all depth classes (Fig. 2). Toothcarp and loach occupying surface positions were larger at night (24•5 and 44•8 mm, respectively) than during daytime (20•0 and 28•5 mm) (toothcarp F1,494, P < 0•001; loach F1,36, P < 0•01). Sand smelt did not show such a difference (F1,26, P ¼ 0•8). The relatively large number of toothcarp captured in surface position (495 individuals) allowed a confident comparison of the relation between fish size and depth among bottom and surface captures. While the size of toothcarps using microhabitats near the bottom showed a clear increase with increasing depth, this pattern almost disappeared in surface captures, which did not present important size differences among depth classes (Fig. 2). A C TI V I TY , D EP T H A N D R E F U G E Toothcarp and sand smelt were captured more frequently in daytime traps than in those set at night, suggesting a reduced nocturnal activity (Table I). In # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 TABLE I. Results of the GLM on the influence of time (night or day), presence of refuge (Ref) and depth in the CPUE of the different species and size classes Factor d.f. ToI ToII ToIII Time 1 7 •1 9 • 2 Ref 1 0 •3 0 • 1 Depth 4 13•1 6• 5 Time x Ref 1 0 •2 0 • 0 Time x depth 4 1 •1 0 • 1 Depth x Ref 4 2 •4 3 • 8 Error 835 Total % deviance 24•2 19•9 Dispersion parameter 1 •1 0 • 9 11•0 5 •8 4 •8 0 •1 0 •4 0 •7 22•9 0 •7 LoI LoII LoIII 2 •1 0 •3 2 •6 0 •5 1 •2 3 •4 13•8 10•1 0 • 4 0 •2 2• 8 0• 7 7• 0 1• 5 0• 9 0• 8 SaI SaII SaIII 1• 4 0• 0 7• 6 0• 3 2• 0 0• 5 6•2 11•1 0 • 0 3 •2 9 • 5 4 •3 0 • 6 0 •4 0 • 2 0 •3 0 • 9 0 •8 11•3 0•2 3•8 0•0 0•2 1•3 11•9 0• 3 17•4 20•1 0 • 2 0 •4 16•8 0•1 To, toothcarp; Lo, loach; Sa, sand smelt. Size classes- I, II and III. Values are the percentage of the total deviance explained by each factor. Significant results (P < 0•05) appear in bold. both species these differences in activity between day and night increased from size I (explaining 6–7% of the total CPUE variation) to size III fish (>11% of the variation). Size III loaches did not show any significant activity reduction at night, while the two smaller size classes were slightly more active during daytime. All toothcarp size classes showed significant preferences for certain depths, although the strength of this preference decreased from small to large fish (Table I). During the day size I toothcarps were almost absent at depths >40 cm, and strongly preferred the shallowest microhabitat. Larger size classes used relatively deeper microhabitats (Fig. 3). Size III toothcarps clearly preferred exposed microhabitats, while the presence of refuge allowed smaller size classes to occupy relatively deeper waters, as shown by the interaction between depth and refuge (Table I). When using refuge microhabitat at night the three size classes of toothcarp were more or less equally distributed among depths (Fig. 3). Loach also showed a change in depth preference towards deeper microhabitats with increasing fish size. Depth explained a significant amount of the variation in CPUE only in the smallest and largest loach size classes (Table I), size II loaches showing intermediate depth preferences (Fig. 3). Loach size classes, however, did not show absolute avoidance of any depth class as those observed in small toothcarp. Size II loach occupied deeper positions at any time when using refuge microhabitats (Fig. 3), though the interaction between depth and refuge was not significant in the GLM due to a strong underdispersion of the data (Table I). Data sets on the CPUE of the different size classes of sand smelt were also severely underdispersed, increasing the probability of type II error (Table I). Captures of this species were also very low at night, and reliable patterns could be extracted only from daytime traps (Fig. 3). As with the other species, sand smelt changed depth preferences from shallow to deeper microhabitats with increasing fish size. As observed for toothcarp, however, the depth preferences were stronger in small individuals than in larger ones (Table I). Size II sand smelts significantly avoided refuge microhabitats. There were clear direct relationships between fish size and depth in the three species, but there were also some changes in this pattern related to time and # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 Toothcarp Day 0·2 Size I Size II Size III Size I Loach Size II Size III Size I Sand smelt Size II Size III ≤20 25 35 45 >50 ≤20 25 35 45 >50 ≤20 25 35 45 >50 0·1 Night 0 0·2 0·1 0 Day 0·05 0 0·1 Night Log10 CPUE (individuals trap–1 h–1) 0·1 0·05 0 Day 0·1 0·05 Night 0 0·1 0·05 0 Depth (cm) FIG. 3. CPUE (mean ± S.E.) of the different size classes of the three studied species in exposed microhabitats (*) and refuge ones (.) at different depths and times of the day. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 presence of refuge (Fig. 4). Toothcarp using refuge microhabitat were significantly smaller than those in exposed microhabitat at any time. While during the day toothcarp in refuge microhabitats were larger in deeper waters this relation disappeared at night. During the day there were no differences in the size of loach using refuge or exposed areas, but at night they were larger in the latter. As with toothcarp, loach did not display a clear size structure in relation to depth when using refuge microhabitats at night. Loach were also larger in exposed microhabitat at night than during the day, a pattern that could be observed at different depths (Fig. 4). Day 55 Night n = 1038 n = 1114 n = 605 n = 443 n = 207 n = 238 n = 281 n = 490 n = 243 n = 132 n = 62 n = 101 Toothcarp 45 35 25 15 55 Loach LT (mm) 45 35 25 15 55 Sand smelt 45 35 * * 25 15 ≤20 25 35 45 >50 ≤20 25 35 45 >50 Depth (cm) FIG. 4. Mean ± S.E. total length in exposed microhabitats (*) and refuge ones ( .) at different depths and time of the day. Numbers of captured fish are also given. *, less than five valid cases. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 PATTERNS OF CO-OCCURRENCE (b) 1·5 (a) PC4 (12·6%) III 1 0·5 0 1·5 (c) 1 0·5 0 PC2 (17·8%) III II III II I PC1 (22·7%) II I I PC2 (17·8%) Log10 CPUE (individuals trap–1 h–1) Log10 CPUE (individuals trap–1 h–1) Log10 CPUE (individuals trap–1 h–1) PCA was applied to a matrix of 481 rows (traps) x nine columns (species and size classes) and produced four principal components (PCs) with eigenvalues >1. PC1 ordered traps as an inverse function of the number of captures [Fig. 5(a)], was positively correlated with trap depth (r ¼ 0•21; P < 0•001) and its scores were higher in night traps (ANOVA, P < 0•001). CPUEs of all species/size classes were negatively correlated with this component, except that of size III loaches (r ¼ 0•23, P < 0•001), which was the species and size class that had the maximum nocturnal activity and used the deepest microhabitats among the different species and size classes. PC4 constituted a gradient running from traps dominated by size I catches to traps dominated by size III catches [Fig. 5(b), (d)]. It was positively correlated with depth (r ¼ 0•21; P < 0•001) and was therefore related to the size-related change towards deeper microhabitat recorded in the three species (Fig. 3). PC2 and PC3 scores, however, were not related to habitat use, but with the specific identity of fishes, independently of fish size. In fact, these two gradients in trap catches represented trends towards a spatio-temporal segregation between toothcarp and sand smelt (PC2) [Fig. 5(b), (c)] and between toothcarp and loach (PC3), while no clear association or avoidance was observed between 1·5 (d) 1 0·5 0 PC4 (12·6%) FIG. 5. Results of the PCA applied to a matrix of CPUEs of the nine different species and size classes in 481 plastic minnow traps. (a) Relationship between PC1 (eigenvalue ¼ 2•04) and total CPUE (including all individuals of the three species) (r ¼ —0•91, P < 0•001). (b) Projection of the loadings of the different species (*, tooth carp, x, loach and ., sand smelt) and size classes (I, II and III) on the spaced formed by PC2 (eigenvalue ¼ 1•60) and PC4 (eigenvalue ¼ 1•13). (c) Relationship between the PC2 scores of each trap and the CPUE of toothcarp (*) (r ¼ —0•48, P < 0•001), loach ( ) (r ¼ —0•09, P ¼ 0•03) and sand smelt (.) (r ¼ 0•80, P < 0•001) independently of size classes. (d) Relationship between the PC4 scores of each trap and the CPUE of size I (*) (r ¼ —0•66, P < 0•001), size II (x) (r ¼ 0•05, P ¼ 0•22) and size III fish (. ) (r ¼ 0•40, P < 0•001), independently of fish species. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 loach and sand smelt. All toothcarp size classes were negatively correlated with PC2 (P < 0•001 in all cases), while all sand smelt size classes were positively correlated with it (P < 0•001 in all cases). While CPUE of loach was not related to PC2, it had a strong correlation with PC3 (r ¼ 0•81, P < 0•001), which was at the same time negatively correlated with all toothcarp size classes (P < 0•001). DISCUSSION POSITION IN THE WATER COLUMN A strong preference for bottom positions in relation to surface positions could be observed in the three species under study. This pattern was obvious in the case of loach that, as other Cobitis species, is a bottom-dwelling fish lacking a swimbladder (Perdices & Doadrio, 1997), but to date there are no available data on this issue concerning toothcarp and sand smelt. Preference for bottom positions can be linked to feeding behaviour. For example, it has been shown that Andalusian toothcarp’s ‘sister’ species, Aphanius iberus (Valenciennes), has a diet based mainly on benthic crustaceans (Vargas & de Sostoa, 1997). Rincon et al. (2002) also noted in experimental conditions that A. iberus tended to occupy positions near the bottom of aquaria. These previous observations fit well with toothcarp’s preference for positions near the stream bottom. Though sand smelt is often considered a planktivorous species, it have been shown to be an opportunistic feeder, switching to benthic prey when zooplankton is not available (Vizzini & Mazzola, 2002). It is probable, therefore, that in stream environments where, as far as is known the diet of sand smelt has never been studied, the species would feed preferentially on benthic organisms. On the other hand, predation risk could also limit the use of surface positions by toothcarp and sand smelt, since in this situation fishes would be more vulnerable to avian predators like egrets and kingfishers. As birds are sizeselective predators preferentially consuming larger fishes (Britton & Moser, 1982), the increased predation risk in surface positions could account for the smaller size of toothcarps and sand smelts occupying them. Also, since egrets and kingfishers are exclusively diurnal predators, the release of bird predation pressure at night would allow larger toothcarps to move to the surface. D E P T H A ND R E FU G E The results presented here show that the three studied species select microhabitat as a function of depth and presence of refuge, and that these preferences vary with fish size and time of day. In general, the three species changed depth preferences towards deeper areas at different stages during ontogeny, as has been recorded in other fishes (Mallet et al., 2000; Rosenberger & Angermeier, 2003). These spatial segregations of size classes can at the same time reduce intraspecific competition (Heggenes, 1988) as well as reduce predation risk from a combination of aquatic and terrestrial predators (Power, 1984, 1987). Depth preferences were stronger during daytime than at night in most species and size classes (Fig. 3), suggesting that the diurnal bird predation could be an important factor causing the observed depth preference patterns (Schlosser, 1988). # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 Refuge (vegetation or other structures) can be an effective protection for fishes both from terrestrial (Valdimarsson & Metcalfe, 1998) and from aquatic predators (Orth et al., 1984; Gotceitas, 1990). The responses of prey fishes in relation to these two types of predation should be different, however, since piscivorous fishes are gape-limited (Nilsson & Bronmark, 2000) and birds are size-selective predators (Britton & Moser, 1982). Among the studied species there was a stronger selection of depths than that observed for refuge, since only size III toothcarps and size II sand smelts showed significant preferences related to the presence of refuge (Table I). In the presence of refuge small toothcarp and size II loach, however, occupied preferentially deeper microhabitats than in exposed areas. The presence of refuge also had a strong influence in the depth-size structure of toothcarp and loach populations (Fig. 4). During the day, the loach population had a strong depth-size structure, with no size differences recorded between refuge and exposed microhabitats, while at night toothcarp were smaller and did not follow the larger fish-deeper water pattern in refuge microhabitats. As a cryptic and benthic species, loach is probably less vulnerable to visual predators, such as birds and diurnal piscivorous fishes (Armbruster & Page, 1996). But loach can be also especially vulnerable to predation by eels, a species that, when feeding on fishes, feeds mainly upon benthic species (Barak & Mason, 1992). Eels remain inactive in sheltered habitats during daytime and forage at night (Baras et al., 1998; Schulze et al., 2004). Predation risk due to eel activity is probably linked with the differences in loach size between refuge and exposed habitat observed at night. Toothcarp were larger in deeper microhabitats and at the same time smaller in refuge than in exposed microhabitats during the day, but at night there was no size-depth relation for toothcarp in refuge microhabitat. The diurnal pattern seems to be influenced by an aquatic gape-limited predator, as refuge is occupied by smaller individuals. But since eels are nocturnal, what is that diurnal aquatic predator? One possibility is that small toothcarp are seeking refuge from viperine snakes. Snakes are not truly gape-limited predators, but the higher energetic cost of capturing large prey cause a positive selection of smaller or medium-sized prey (Mehta, 2003; Moore et al., 2004). Another possibility is that some of the invertebrate-feeding fish species occupying the stream stretch, mainly sand smelt or flathead grey mullet M. cephalus, would be consuming small fishes. In fact, some authors (Rosecchi & Crivelli, 1992; Bartulovic et al., 2004) have found that fish larvae or juveniles are occasional prey of sand smelt, which is therefore capable of becoming a small fish predator in certain conditions. At night the presence of the otter in the study stretch, being nocturnal predator (Beja, 1996) frequently consuming eels (Clavero et al., 2004), probably preclude larger eels from occupying shallow areas. This predator exclusion could be linked to the nocturnal maintenance of the size-depth relation in exposed microhabitats observed in both toothcarp and loach (Fig. 4). CO-OCCURRENC E Results from the PCA showed that some of the patterns on the co-occurrence of species and size classes could be related to differential habitat use (PC1 and PC4). The PCA also showed clear segregations between toothcarp and sand # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 smelt (PC2) and between toothcarp and loach (PC3), however, which could not be linked to any microhabitat characteristic. These observations suggest that toothcarp tends to occupy microhabitats that are not being simultaneously used by the other two species. One possible explanation for this pattern would be that toothcarp excludes other small fishes through aggressive interactions. Aphanius toothcarps, especially male individuals, have been shown to display aggressive behaviours (Rincon et al., 2002). During the fieldwork male toothcarps were seen defending territories by chasing and nipping other individuals (unpubl. data). These aggressive interactions, however, were mainly directed to conspecifics, and would hardly explain the exclusion of loach and sand smelt. The spatio-temporal segregation between toothcarp and both loach and sand smelt could be also produced by a higher incidence of interspecific interactions on toothcarp’s microhabitat use. Some cyprinodontid species in North America have been shown to be very sensitive to interspecific interactions (Echelle et al., 1972), often occupying harsh environments where they lack competitors and major aquatic predators (McMahon & Tash, 1988). Iberian Aphanius species are also very sensitive to the presence of other fish species, both native and introduced (Rincon et al., 2002; Prenda et al., 2003), and the most numerous Andalusian toothcarp populations occupy hypersaline streams where there are no other fish species (Doadrio et al., 2002). Being a critically endangered species, the possible influences of these interactions on toothcarp populations should be further investigated. To do so, the results presented here, which are from a diverse fish community and the first available on the species’ microhabitat, should be compared with locations where toothcarp do not coexist with any other fishes. We greatly acknowledge the assistance and company during the field work provided by L. Fernandez, M. Narvaez, L. Barrios, J. Alvarez, E. Deuan, L. Dominguez and E. Perez. Two anonymous referees made very useful comments on an early version of this work. This study is part of the project ‘Biotic integrity and environmental factors of watersheds in south-western Spain. Application to the management and conservation of Mediterranean streams’ (Ministerio de Ciencia y Tecnologıa, REN2002–03513/HID) References Armbruster, J. W. & Page, L. M. (1996). Convergence of a cryptic saddle pattern in benthic freshwater fishes. Environmental Biology of Fishes 45, 249–257. Barak, N. A. E. & Mason, C. F. (1992). Population density, growth and diet of eels, Anguilla anguilla L., in two rivers in eastern England. Aquaculture and Fisheries Management 23, 59–70. Baras, E., Jeandrain, D., Serouge, B. & Philippart, J. C. (1998). Seasonal variations of time and space utilisation by radio-tagged yellow eels Anguilla anguilla in a small stream. Hydrobiologia 371/372, 187–198. Bartulovic, V., Lucic, D., Conides, A., Glamuzina, B., Dulcic, J., Hafner, D. & Batistic, M. (2004). Food of sand smelt, Atherina boyeri Risso, 1810 (Pisces: Atherinidae) in the estuary of the Mala Neretva River (middle-eastern Adriatic, Croatia). Scientia Marina 68, 597–603. Beja, P. R. (1996). Temporal and spatial patterns of rest-site use by four female otters Lutra lutra along the south-west coast of Portugal. Journal of Zoology, London 239, 741–753. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 Britton, R. H. & Moser, M. E. (1982). Size specific predation by herons and its effect on the sex-ratio of natural populations of the mosquito fish Gambusia affinis Baird and Girard. Oecologia 53, 146–151. Clavero, M., Prenda, J. & Delibes, M. (2004). Influence of spatial heterogeneity on coastal otters (Lutra lutra) prey consumption. Annales Zoologici Fennici 41, 545–549. Clavero, M., Blanco-Garrido, F., Rebollo, A. & Prenda, J. (2005a). El salinete (Aphanius baeticus) en el rıo de La Vega: caracterizacion de la poblacion y factores de amenaza. Almoraima 31, 87–92. Clavero, M., Blanco-Garrido, F. & Prenda, J. (2005b). Fish-habitat relationships and fish conservation in small coastal streams in southern Spain. Aquatic Conservation: Marine and Freshwater Ecosystems 15, 415–426. Copp, G. H. & Jurajda, P. (1999). Size-structured diel use of river banks by fish. Aquatic Sciences 61, 75–91. Copp, G. H., Spathari, S. & Turmel, M. (2005). Consistency of diel behaviour and interactions of stream fishes and invertebrates during summer. River Research and Applications 21, 75–90. Crawley, M. J. (2002). Statistical Computing. An Introduction to Data Analysis using SPlus. Chichester: John Wiley & Sons. Doadrio, I. (Ed.) (2001). Atlas y Libro Rojo de los Peces Continentales de España. Madrid: Ministerio de Medio Ambiente. Doadrio, I., Carmona, J. A. & Fernandez-Delgado, C. (2002). Morphometric study of the Iberian Aphanius (Actinopterygii, Cyprinodontiformes), with description of a new species. Folia Zoologica 51, 67–79. Dungan, J. L., Perry, J. N., Dale, N. R T., Legendre, P., Citron-Pousty, S., Fortin, M.-J., Jakomulska, A., Miriti, M. & Rosenberg, M. S. (2002). A balanced view of scale in spatial analysis. Ecography 25, 626–640. Echelle, A. A., Echelle, A. F. & Hill, L. G. (1972). Interspecific interactions and limiting factors of abundance and distribution in the Red River pupfish (Cyprinodon rubrofluviatilis). American Midland Naturalist 88, 109–130. Gasith, A. & Resh, V. H. (1999). Streams in Mediterranean climate regions – Abiotic influences and biotic responses to predictable seasonal events. Annual Review of Ecology and Systematics 30, 51–81. Gotceitas, V. (1990). Plant stem density as cue in patch choice by foraging juvenile bluegill sunfish. Environmental Biology of Fishes 29, 227–232. Grossman, G. D., de Sostoa, A., Freeman, M. & Lobon-Cervia, J. (1987a). Microhabitat selection in a Mediterranean riverine fish assemblage: I. Fishes of the lower Matarrana. Oecologia 73, 490–500. Grossman, G. D., de Sostoa, A., Freeman, M. & Lobon-Cervia, J. (1987b). Microhabitat selection in a Mediterranean riverine fish assemblage: II. Fishes of the upper Matarrana. Oecologia 73, 501–512. Harvey, B. C. & Stewart, A. J. (1991). Fish size and depth relationships in headwater streams. Oecologia 87, 336–342. Heggenes, J. (1988). Effect of experimentally increased intraspecific competition on sedentary adult brown trout (Salmo trutta) movement and stream habitat choice. Canadian Journal of Fisheries and Aquatic Sciences 45, 1163–1172. Hubert, W. A. (1996). Passive capture techniques. In Fisheries Techniques, 2nd edn (Murphy, B. R. & Willis, D. W., eds), pp. 157–192. Bethesda, MD: American Fisheries Society. King, A. J. (2004). Ontogenetic patterns of habitat use by fishes within the main channel of an Australian floodplain river. Journal of Fish Biology 65, 1582–1603. doi: 10.1111/j.1095.8649.2004.00567.x Mallet, J. P., Lamouroux, N., Sagnes, P. & Persat, H. (2000). Habitat preferences of European grayling in a medium size stream, the Ain River, France. Journal of Fish Biology 56, 1312–1322. doi:10.1006/jfbi.2000.1252 McIvor, C. C. & Odum, W. E. (1988). Food, predation risk, and microhabitat selection in a marsh fish assemblage. Ecology 69, 1341–1351. # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85 McMahon, T. E. & Tash, J. C. (1988). Experimental analysis of the role of emigration in population regulation of desert pupfish. Ecology 69, 1871–1883. Mehta, R. S. (2003). Prey-handling behaviour of hatchling Elaphe helena (Colubridae). Herpetologica 5, 469–474. Moore, R. D., Griffiths, R. A., O’Brien, C. M., Murphy, A. & Jay, D. (2004). Induced defences in an endangered amphibian in response to an introduced snake predator. Oecologia 141, 139–147. Nilsson, P. A. & Bronmark, C. (2000). Prey vulnerability to a gape-size limited predator: behavioural and morphological impacts on northern pike piscivory. Oikos 88, 539–546. Orth, R. J., Heck, K. L. & van Montfrans, J. (1984). Faunal communities in seagrass beds: a review of the influence of plant structure and prey characteristics on predator-prey relationships. Estuaries 7, 339–350. Perdices, A. & Doadrio, I. (1997). Threatened fishes of the world: Cobitis paludica (De Buen, 1930) (Cobitidae). Environmental Biology of Fishes 49, 360. Poizat, G. & Pont, D. (1996). Multi-scale approach to species-habitat relationships: juvenile fish in a larger river section. Freshwater Biology 36, 611–622. Power, M. E. (1984). Depth distribution of armored catfish: predator induced resource avoidance. Ecology 65, 523–528. Power, M. E. (1987). Predator avoidance by grazing fishes in temperate and tropical streams: importance of stream depth and prey size. In Predation: Direct and Indirect Impacts on Aquatic Communities (Kerfoot, W. C & Sih, A., eds), pp. 333–351. Hanover, NH: University Press of New England. Prenda, J., Armitage, P. D. & Grayston, A. (1997). Habitat use by the fish assemblages of two chalk streams. Journal of Fish Biology 51, 64–79. Prenda, J., Clavero, M. & Blanco-Garrido, F. (2003). Los peces continentales de la provincia de Cadiz. Revista de la Sociedad Gaditana de Historia Natural 3, 217–232. Rincon, P. A., Barrachina, P. & Bernat, Y. (1992). Microhabitat use by 0þ juveniles cyprinids during summer in a Mediterranean river. Archiv für Hydrobiologie 125, 323–337. Rincon, P. A., Correas, A. M., Morcillo, F., Risueno, P. & Lobon-Cervia, J. (2002). Interaction between the introduced eastern mosquitofish and two autochthonous Spanish toothcarps. Journal of Fish Biology 61, 1560–1585. doi: 10.1006/ jfbi.2002.2175 Rosecchi, E. & Crivelli, A. J. (1992). Study of a sand smelt (Atherina boyeri Risso 1810) population reproducing in fresh water. Ecology of Freshwater Fish 1, 77–85. Rosenberger, A. & Angermeier, P. L. (2003). Ontogenetic shifts in habitat use by the endangered Roanoke logperch (Percina rex). Freshwater Biology 48, 1563–1577. Santos, J. M., Godinho, F. N. & Ferreira, M. T. (2004). Microhabitat use by Iberian nase Chondrostoma polylepis and Iberian chub Squalius carolitertii in three small streams, north-west Portugal. Ecology of Freshwater Fish 13, 223–230. Schlosser, I. J. (1988). Predation risk and habitat selection by two size classes of stream cyprinids: experimental test of a hypothesis. Oikos 52, 36–40. Schulze, T., Kahl, U., Radke, R. J. & Benndorf, J. (2004). Consumption, abundance and habitat use of Anguilla anguilla in a mesotrophic reservoir. Journal of Fish Biology 65, 1543–1562. doi: 10.1111/j.1095-8649.2004.00565.x Valdimarsson, S. K. & Metcalfe, N. B. (1998). Shelter selection in juvenile Atlantic salmon, or why do salmon seek shelter in winter? Journal of Fish Biology 52, 42–49. Vargas, M. J. & de Sostoa, A. (1997). Life-history pattern of the Iberian toothcarp Aphanius iberus (Pisces, Cyprinodontidae) from a Mediterranean estuary, the Ebro delta (Spain). Netherlands Journal of Zoology 47, 143–160. Vizzini, S. & Mazzola, A. (2002). Stable carbon and nitrogen ratios in the sand smelt from a Mediterranean coastal area: feeding habits and effect of season and size. Journal of Fish Biology 60, 1498–1510. doi: 10.1006/jfbi.2002.2003 # 2005 The Fisheries Society of the British Isles, Journal of Fish Biology 2005, 67 (Supplement B), 72–85