Habitat selection patterns of a species at the edge – case study of the native racer goby population in Central Europe

Invasive alien species are regarded a nuisance. This extends into a lack of conservation efforts in their native range. As a consequence, conservation of e.g. range-edge populations is neglected. Gobiidae have many representatives of alien species in European freshwaters, and therefore they have a bad reputation. Objectives of this study were to: define the habitat selection patterns of a species at the edge, and examine the ontogenetic variation in its distributions, i.e. spatial distribution of different size classes. A racer goby Babka gymnotrachelus (syn. Neogobius gymnotrachelus) population was selected for the model. In numerous European river basins, Ponto-Caspian racer goby has been an invasive alien species of interest to researchers for many years. Recently, however, native populations of the species have been described in the Polish tributary of the upper Dniester River (Black Sea basin). We used habitat data and densities of racer goby to disentangle the habitat selection patterns of the species at a river reach at the edge of its native range. Evident preferences towards habitats with large submerged objects serving as hiding places were characteristic of the largest gobies. Adult, largest gobies were very likely to choose the ‘boulders’ site, while forcing smaller individuals to occupy places with faster water current, i.e. less suitable in terms of saving energy. At a larger geographic scale, a significant portion of the submountain river was unsuitable for racer gobies. At the edge of the racer goby range, patches providing habitats suitable for the species were scarce and scattered. With regard to invasive populations, the presence of stony bottoms, quite certainly cannot be considered as a factor excluding potential colonisation by racer goby, and in submountain rivers it might be the preferred kind of bottom.


Results
The bottom of the river reach consists of boulders and smaller stones, with the remainder consisting of gravel and sand. These substrate fractions show patchy distribution on the river bottom. Based on the bottom substrate fractions (see Supplementary Fig. S1), sampling sites (1 m x1 m squares, Fig. 2) were divided into three main habitat types. In the first habitat type (habitat types were distinguished on the basis of cluster analysis -please see the method section), the dominant bottom substrate fraction was gravel (habitat type: gravel, GR). In the second habitat, the most numerous fractions were stones of various sizes (habitat type: cobbles and pebbles, PE), and the third habitat type was dominated by boulders and large cobbles (habitat type: boulders, BO). Significant differences were observed in the shares of distinguished bottom substrate fractions in these three habitat types. All of the compared pairs of habitat types were significantly different (see Supplementary Table S1), with the largest proportion having gravel, followed by large cobbles and boulders (see Supplementary Table S2).
In the surveyed river reach, shallow fragments were present in zones near the banks and in the upper zone. However, in the middle part of the river reach, deeper regions (depth exceeding 50 cm) were observed. The water current varied; zones with slow flowing or even stagnant water as well as sections with water currents exceeding 1 m s −1 were observed. In sections with fast flowing water, boulders dominated the bottom substrate (habitat type: BO) (Fig. 3). The water temperature of the surveyed reach ranged from 18.6 °C to 20.4 °C (SD = 0.66 °C), and the dissolved oxygen content exceeded 9.5 mg L −1 (range 9.66-10.03 mg L −1 , SD = 0.14 mg L −1 ). Water conductivity ranged from 305 μS to 473 μS (SD = 43.9 μS). No statistically significant differences were determined between mean values of water temperature, dissolved oxygen, and water conductivity in three types of habitats.
At four additional sites (S1-S4) above the analysed river reach, racer goby was scarce, and its density varied from 3 to 12 individuals per 100 m2 of the area of the river (Table 1). Site S1 was the highest located fragment of the Strwiąż River where the species was recorded 35 . In the analysed river reach (SR), the density of racer goby was 98 ind. 100 m −2 . All three size classes of the fish were also abundantly represented (Table 1). A total of 431 racer goby specimens were caught, including 145 belonging to the SF size class, 215 belonging to the MF size class, and 71 large individuals (LF class). The distinguished habitat types differed significantly among the racer goby population size classes (see Supplementary Table S3, Table 2), with the smallest individuals (SF) contributing most to the dissimilarity (see Supplementary Table S4). A considerable habitat overlap exists between SF and MF fishes, but not between LF and the other two size classes (the degrees of habitat overlap (CZ) were as follows: size classes GLM showed that the density of the smallest racer goby (SF) was significantly influenced by the water current velocity and presence of a PE habitat. A similar relationship was found for fish from the MF size class. For adult fish (LF class), the water current velocity as well as the depth and presence of boulders in the substrate (habitat type BO) significantly influenced their numbers ( Table 2). The selected water current velocity for the racer goby SF size class was 0.057 m s −1 ± 0.005. The selected current speeds were 0.039 m s −1 ± 0.015 S.E. and 0.022 m s −1 ± 0.020 S.E for the MF and LF size classes, respectively. (Fig. 4a). The selected depths for the racer goby SF, MF and LF classes were 25.5 cm ± 2.624 S.E., 24.7 cm ± 1.906 S.E. and 37.7 cm ± 4.425 S.E., respectively (Fig. 4b). The sites where the LF fish were present differed significantly in depth from sites where the fish of the other two size classes were found. No significant differences occurred in terms of water current velocity between the LF and MF sites and between the SF and MF places (Fig. 4c,d).

Discussion
The upper Strwiąż River can be used to obtain information regarding how the racer goby, a species typically found in lower river sections, manages in the submountain river. In addition, the Strwiąż River runs the upper range edge of this species 30,35 . Species at the edge of their range are called peripheral 43 . Therefore, studying the racer goby population in this river provides additional information about the species functions at its range edge/ periphery, where the persistence of a stable population is particularly difficult 44,45 . The Strwiąż River in the territory of Poland, is an example of such a peripheral basin that has survived with a good ecological status (as defined in the Water Framework Directive (WFD) 46 because of effective land management and various forms of implemented protection (i.e., Natura 2000, landscape park) 35 . Due to the maintenance of the natural characteristics of the river channel, racer gobies in the Strwiąż River showed mosaic distribution in habitat patches meeting its   www.nature.com/scientificreports www.nature.com/scientificreports/ requirements. The presence of fish species in a habitat patch, and thus, their distribution in the river, is most influenced by abiotic factors, such as the water current, substrate granulation and water depth 1,39,47 . Individual species have different requirements in this respect, and their mosaic distribution is thus often observed in rivers 48,49 .
The mosaic fish distribution in rivers also applies to specific size classes (usually corresponding to age) due to different requirements at various stages of an individual's life cycle and the tendency to reduce intraspecific competition via the spatial distribution of habitats 45,49 . Evident preferences towards habitats with large submerged objects serving as hiding places were characteristic of the largest gobies. They chose places with boulders, and they were predominant in this type of habitat. A small number of smaller individuals in the habitat suggests that larger fishes won in competition with individuals of smaller sizes. Boulders with spaces in between with low water current velocity values seem to be the most hydrodynamically suitable habitat allowing for saving energy. Abiding in strong water current is costly in terms of energy, but boulders provide shelter from strong water current 50 . Behind boulders and between them, at the bottom, water current reached the zero value. Therefore, adult, largest gobies were very likely to choose the 'boulders' site in Strwiąż, while forcing smaller individuals to occupy places with faster water current, i.e. less suitable in terms of saving energy.
In the Strwiąż River, larger gobies inhabit large cobbles and boulders, while in lowland rivers, racer gobies prefer bottom substrates comprising sand and fine sediments containing large, immersed objects that serve as useful hiding places 21,38 . In submountain rivers, this type of bottom substrate is not frequent. Kakareko et al. 21 data on racer goby requirements suggest that a significant portion of the surveyed Strwiąż reach should be unsuitable for racer gobies, mainly due to excessively fast water current, but most likely also due to bottom substrate granulation. Depending on the body size, the species was divided between particular habitats. In the Strwiąż River, significant differences in the densities of large and small racer gobies found in areas with various bottom substrates were observed. The occurrence of large gobies was associated with boulders, while small gobies were more numerous in zones with finer bottom fractions (Fig. 5). A certain association with hard substrate was also observed in situ conditions by Krpo-Ćetković et al. 16 in the Danube (Serbia), where the highest occurrence of the racer goby was on pebble bottom. The authors, however, did not differentiate between age classes.
Gobies respond to the presence of other individuals of their species, as well as other species 51 . Visual isolation is very important for territorial fish species. Physical structures provide visual isolation from other fish, reducing territorial needs 52,53 . These factors may be even more important than hydrodynamic-related energy savings, because it allows individuals to avoid antagonistic intra-and interspecific interactions 54 . When sufficiently suitable habitats are scarce and do not occur continuously, habitat niches occupied by different size classes can partially overlap, creating strong intraspecific competition (Fig. 5) that potentially results in ontogenetic changes in habitat selection. Large racer goby specimens are aggressive towards other fish and try to occupy the most favourable habitats for themselves 51,55 , and smaller fish were forced to occupy worse habitats. The reproductive biology of the racer goby (speleophilic species) may also be the reason why males search for appropriate places, such as niches under stones, to deposit female eggs, and the males guard the eggs until hatching 17,36 .  Racer goby size categories

MF SF
intraspecific competition Figure 5. Conceptual model showing habitat preferences and relationships between racer goby size classes in the submountain river at the edge of the species range; depth min -minimum depth at which racer gobies were observed; V opt -optimal velocity for racer gobies.
Predators are also a biotic factor that strongly affect the distribution of fish in rivers 56 . In the Strwiąż River, large chub (Squalius cephalus) and brown trout (Salmo trutta) are piscivorous fish 35 , and zones with shallow water offer shelter to smaller individuals against large predatory fish 57,58 . Most small-sized racer gobies were caught in the shallowest zones. On the other hand, although less exposed to attack by large predatory fish, larger gobies selected deeper places (but only where boulders and large cobbles were present, allowing hiding and forage). But although escaping to shallow water is an effective way to avoid predatory fish, the threat from piscivorous birds increases in this case. In the Strwiąż River, grey herons (Ardea cinerea) and black storks (Ciconia nigra) were often observed foraging in the shallows (Kukuła and Bylak unpubl. data). The black stork is a predator posing one of the serious threats to the ichthyofauna of mountain streams in the Carpathians 59 . The stork's preference for shallow reaches of the streams with stony bottom, inhabited by bottom-dwelling fish, is linked to the visual fishing method of storks 60 .
The habitat suitabilities for racer gobies of different age classes are characterized by two environmental factors previously discussed, i.e., the type of bottom substrate and depth, verified by the water current. In experimental and field studies in lowland rivers, racer gobies avoided water currents exceeding 0.1 m s −1 38 . The water current strongly influences the availability of other environmental elements and strongly limits the space available for racer gobies. In submountain rivers, habitats especially desirable for larger individuals seem to be boulders and large cobbles (Fig. 5). At a larger geographic scale, however, considering the entire submountain river, such habitats are found mainly in zones with a water current velocity too fast for racer gobies. Therefore, the species was not found in a significant portion of the river studied 35 . Zones with more turbulent water flow occur more frequently in submountain rivers than in lowland rivers 61,62 . Kakareko et al. 21 reported that in lowland rivers, racer gobies mainly reside in habitats off the main channel that have slow water velocities and both soft and hard bottoms, using various submerged objects as shelters. Few of these habitat qualities are found in submountain rivers. Therefore, racer gobies had few suitable places to inhabit in the Strwiąż River. At a larger geographic scale, the most suitable racer goby habitats occurred in patches separated from each other 35 , but the population is maintained, because the survival of the population at the edge of the species range largely depends on the heterogeneity of habitats on a small scale 35 . The presence of all size classes suggests that the species finds suitable feeding grounds, shelters, and spawning grounds, as well as places of growth of fry there.
Preservation of the river channel's natural characteristics with the diverse habitat mosaic ensures that the population remains stable even at the edge of the species range. Therefore, protection of the species at the range edge protects not only the species itself, but also the specific genetic features of the population. Such features of peripheral populations, different than those in the centre of the range of the species, may serve as a preadaptation preceding future environmental changes 44 . The exclusion of peripheral taxa from protection programs could result in a significant loss of overall genetic resources 43 . Therefore, we believe that there is an urgent need for action to protect river fish species populations at the edge of their range. Protection of such populations is recommended, because populations functioning near the edge of the species range appear to have the highest potential for speciation due to their exposure to variable environmental conditions at the boundaries of species' tolerances 63 . Peripheral areas of river basins, where fish species exist on the edge of their range, are often located in zones with relatively low levels of human pressure 30,35 . Due to this, the populations could have survived in good state, and are worth a more thorough insight, and consequently deserve to be included in local fish protection plans.
On the other hand, referring the obtained results to alien species, even when separated by long sections of unfavourable habitat parameters, patches of suitable habitats can form a stepping stones that allow significant extension of the species range, even to areas that are far from optimal 64 . The construction of dam reservoirs is frequently accompanied by the development of habitats suitable for gobies. Dam reservoirs are mentioned as sources of secondary invasion of the species. Dam reservoirs with shallow areas near the shores offer extensive areas with habitats suitable for the goby 65 . Until now, forecasts of e.g., racer goby expansion routes in areas wherein the fish is considered an alien invasive species have been limited to lowland rivers, particularly those on which dam reservoirs are constructed 66 . Our data showed that racer gobies also cope well in difficult environmental conditions in submountain rivers. In such rivers, relatively small patches of suitable habitats, considerably smaller than those offered by dam reservoirs, ensure fish recruitment, and in the case of invasive populations could become a source of further expansion. Therefore, our results can also be used to develop a framework for the risk assessment of alien racer gobies that may extend their range through river habitat networks. Forecasting threats related to alien gobies should include areas that seem suboptimal.
Ohayon and Stepien 19 raised the need for further studies to assess the racer goby's spreading abilities. Others, for the purpose of recording and monitoring of racer goby, recommended intensification of research in soft-bottom habitats as places particularly preferred by the species 67 . Recently, Kakareko et al. 21 suggested that the plasticity of the species' habitats reflects its ability to occupy sub-optimal environments, but with less preferred hard substrates. Larger gobies, however, in areas of soft substrata, were observed to excavate cavities underneath stones or pieces of wood actively creating their own refuges from elevated water velocities, and smaller ones have been seen to use small stones as shelters 21 . Our research significantly broadens the knowledge concerning the ecological abilities of this species. With regard to invasive populations, the presence of stony bottoms, quite certainly cannot be considered as a factor excluding potential colonisation by racer goby, and in submountain rivers it might be the preferred kind of bottom. The density of racer goby in the submountain river was even higher than at other sites in the zone of its native range (e.g. in the Dniprodzerzhynsk Reservoir on the Dnieper River in Ukraine 68 ). It turned out that hard substrate does not preclude the functioning of a vivid, abundant, and self-sustaining population of racer goby. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods ethics statement. A sampling permit (No. RG-IX.7143.6.2015.MS) was issued by the Marshal Office of the Podkarpackie Voivodeship following approval by the Regional Directorate for Environmental Protection. Research project was approved by the Department of Biology and Agriculture's Committee for Research Ethics. The research was conducted under license to operate electroshocking tools and license to perform animal investigations according to legislation on the protection of animals and the recommendations of the International Council for Laboratory Animal Science (ICLAS).

Study area.
The headwaters of the Strwiąż River are located in Poland, the Eastern Carpathians and the Sanocko-Turczańskie Mountains, and the Strwiąż River basin is part of the upper (Carpathian) Dniester River basin (Fig. 2a). The river is 94 km long, and the basin has an area of 955 km². The territory located in Poland includes the upper part of the Strwiąż River, which is 17.5 km long and has a basin area of approximately 200 km². The Strwiąż is a submountain river with a strongly variable channel: typical montane sections with rapid water currents and stony bottoms alternate with sections of slow water flow and sandy and sandy-muddy sediments. The river reach (SR, length 115 m) studied was located in the lower part of the Polish section of the Strwiąż River. The river reach covered all types of habitats occurring in the submountain portion of the Strwiąż, i.e., riffles with a stony bottom and fast flowing water, runs and pools with deeper water and a pebbly or pebbly gravel bottom. This reach of the river was selected for detailed research, because it was the highest located fragment at the upper edge of range of the species in this catchment where gobies belonging to all three size classes occurred in high abundance. Above this reach, the abundance of gobies varied from several to a dozen individuals per 100 m2 of the area of the river (Table 1), reaching zero at a distance of approximately ~6 km above the analysed river reach 35 . Sampling design. First, 103 line transects at right angles to the main axis of the water current spaced every 1 metre were determined using a laser rangefinder with a tripod. Each transect was stabilized in the field with wooden stakes driven into the river banks. In the designated stripes, 2 ropes were stretched between the pairs of stakes, and the stripe transect was determined. On the ropes, one-meter-long sections were marked with colourful ribbons, which allowed for the precise determination of 1194 squares (1 m x 1 m), equalling 638 sites in which racer gobies were caught (Fig. 1b). The racer gobies were caught in late summer/early autumn (from 10 to 16 September 2015), which is when young-of-the-year (YOY) fish are readily identified 69 .
Fish were caught using consistent methods with backpack electrofishing equipment (IG600T, Hans Grassl, GmbH, Germany; DC/AC; 650 W direct current; 1,200 W impulse current; 115-565 V). A sampling permit (No. RG-IX.7143.6.2015.MS) was issued by the Marshal Office of the Podkarpackie Voivodeship following approval by the Regional Directorate for Environmental Protection. Research project was approved by the Department of Biology and Agriculture's Committee for Research Ethics. The research was conducted under license to operate electroshocking tools and license to perform animal investigations according to legislation on the protection of animals and the recommendations of the International Council for Laboratory Animal Science (ICLAS). Each fishing crew consisted of one person operating the anode and three people capturing and measuring the fish. To avoid startling the fish, catches were brought in every second (width of one metre) stripe transect, wading from the right to the left river bank. The catches were conducted starting at the most downstream transect. At each sample point, an anode was immersed in the centre of the square for ~10 s, a period proven effective for electrofishing 70 . The electric field parameters were adapted to the water conductivity and the physical nature of the river. The applied voltage was reduced (effective electric field of approximately 1 m diameter). It was sufficient to avoid electrical disturbance of non-sampled areas. Such point abundance sampling is particularly recommended as a useful technique for studying habitat preferences of species 71 . Caught fish were identified and then released as soon as possible after completion of processing (i.e., measuring), approximately 20 metres below the most downstream fishing transect.
Based on literature data 21 , the most important environmental factors affecting racer goby distribution are presumably the water depth, water current velocity, and type of bottom substrate. In each 1 m × 1 m square (at each site) of the sampled river reach, the substrate composition was estimated as the percentage of the area covered by different particle size fractions. Six fractions of the bottom substrate were distinguished: boulders (>256 mm), large cobbles (256-131 mm), small cobbles (130-65 mm), pebbles (64-17 mm), gravel (16-2 mm), and sand (<2 mm). This division was based on the criteria proposed by Bain et al. 72 The water depth and water current velocity (~2 cm above the bottom; and in the case of boulders, between them) were measured along twenty-two line transects spaced every five metres. Measurements along the transects were made every 0.5 m, yielding 530 sampling points. The current velocities were measured using an acoustic Doppler velocimeter (Flowtracker, SonTek, San Diego, CA, USA). Spatial variations of the river water depth and velocity were estimated by the inverse distance weighting method at a resolution of 0.1 m. All spatial analyses were performed using ArcGIS 10.1 software with the spatial analysis extension 73 . In addition, water temperature, conductivity, and dissolved oxygen content were measured at 10 points along the entire river reach sampled using a multiparameter metre (6600 V2, YSI Incorporated, Yellow Springs, Ohio, USA). The characterisation of abiotic factors was done after sampling the fish in order to avoid startling the fish. Data analysis. Statistical data analyses were performed using STATISTICA 12 (TIBCO Software Inc., Palo Alto, CA, USA), and all multivariate analyses were performed using PRIMER v7 74 . Using cluster analysis (Ward's linkage), sampling sites were divided into habitat types (see Supplementary Fig. S1) differing in the percentage shares of individual fractions in the substrate. Percentage data were arcsine transformed. www.nature.com/scientificreports www.nature.com/scientificreports/ Using one-way permutational multivariate analysis of variance (PERMANOVA) with 999 permutations, three distinguished habitat types were compared in terms of their share of six substrate categories using the Bray-Curtis matrix of dissimilarities. Pairwise tests were used to compare the significance of differences between pairs of habitat types. The proportion of each fraction in the different habitat types was then determined using the SIMPER procedure.
One-way PERMANOVA was also used to compare three types of habitats in terms of their percentage of three distinguished racer goby size classes. Species data were log transformed [log(x + 1)], and percentage data were arcsine transformed. Because racer gobies were not found in many of the examined squares (sites), a 'dummy' species was incorporated for this analysis 59 . In addition, SIMPER was used to determine the contribution of each goby size class to the dissimilarity between the three habitat types.
Czekanowski's index (CZ) was used to estimate the degree of habitat overlap between the racer goby size classes with regard to the habitat type as follows: CZ = 1 − 0.5 (∑ |A i -B i |), where A i and B i are the numbers of fish from the compared racer goby size classes occupying habitat type i divided by the total counts of both classes in this type of habitat. CZ ranges from 0 (no overlap) to 1 (full overlap), with values > 0.6 assumed to be considerable overlapping 75 .
For comparisons of water current velocities at sites where racer gobies were found and water depth, nonparametric one-way ANOVA (Kruskal-Wallis test) and post hoc tests for Kruskal-Wallis ANOVA 61 were used. The comparison of water temperature, conductivity, and dissolved oxygen content measured in each type of habitat involved a one-way ANOVA test 76 .
The responses of racer gobies of different sizes (ages) to habitat type, current velocity, and water depth were also analysed using generalized linear models (GLM). The number of racer gobies per sample was treated as a Poisson-distributed response with a log-link function relating the racer goby abundance to the measured environmental variables 77 .

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.