Influence of the Salmo (trutta) trutta on the population structure, the growth, and the habitat preference of a Cottus gobio population

The bullhead Cottus gobio is a small‐sized fish whose range extends over most of the European continent, and it is listed in Annex II of EU “Habitat” Directive for its great conservation interest. In the last decades, bullhead populations suffered local decline. Among the factors that negatively affect bullhead populations, major threats are pollution, habitat deterioration, and the massive introduction of salmonids. This study aims to better understand in which way the presence of the Salmo (trutta) trutta affects Cottus gobio populations. The investigation was carried out in two stretches of a stream located in the Orobic Alps (Italy). The downstream stretch hosts a fish assemblage constituted of both bullhead and brown trout, while in the upstream stretch only bullhead is present. An insurmountable barrier isolates the upstream population of C. gobio from trout, while the environmental conditions of the two stretches proved to be fully comparable. We evaluated population structure, habitat preference, and body shape of bullhead populations in both stretches: the results indicate that the presence of trout decreases the number of bullhead adults, reduces the average adult body size, and induces a bullhead suboptimal habitat occupation. However, both populations of C. gobio showed a well‐structured population and good performance indexes, so trout do not seem to be a threat to the population survival.


| INTRODUCTION
The bullhead Cottus gobio, Linnaeus, 1758, is a small-sized, bottomdwelling freshwater fish species belonging to the Cottidae family.
In the last decades, however, local extinction phenomena occurred, and the species is currently fragmented in different populations whose grade of isolation is growing (Legalle, Santoul, Figuerola, Mastrorillo, & Céréghino, 2005;Van Liefferinge et al., 2005). In Italy, the species is widespread throughout streams and rivers in the Alps (up to 800-1,200 m a.s.l.) and in the springs of the pre-alpine region, while in central Italy the range distribution has undergone a contraction (Franchi, Pompei, & Barbaresi, 2012;Gandolfi, Zerunian, Torricelli, & Marconato, 1991) so that the presence of the bullhead is now supposed to be restricted to a few watercourses of the Tiber River basin, the Tuscan-Emilian Apennines and the Marche region (Freyhof & Kottelat, 2011;Lorenzoni, Ghetti, Carosi, & Dolciami, 2010). The factors that mostly constitute a threat to the species are water pollution, habitat deterioration, predation, and competition with alien species.
The brown trout (Salmo (trutta) trutta) is a European species of salmonid fish, native to northern and central Europe widespread from Ural Mountains to Iberian Peninsula (Figure 1). Over time the brown trout has been widely introduced worldwide in North and South America, Asia, and Australia (MacCrimmon & Marshall, 1968). It is a very adaptable species, which can colonize a wide variety of environments, and has high economic value for food and angling purposes. In Italy the Atlantic haplotype Salmo trutta was probably introduced in the XIX century from North European fish hatcheries (Bettoni, 1895) and at present, it is extremely common and widespread throughout the peninsula, in Sicily and Sardinia, so much that it threats the Italian indigenous trouts (Splendiani, Palmas, Sabatini, & Caputo Barucchi, 2019;Zerunian, 2002): the Adriatic trout (Salmo ghigii), the marble F I G U R E 1 Geographical range of Cottus gobio and Salmo trutta (data provided by IUCN Red list) [Color figure can be viewed at wileyonlinelibrary.com] (Salmo marmuratus) and the Mediterranean trouts (Salmo macrostigma) as well as some very local endemisms (carpio, Salmo cenerinus, Salmo cettii, and Salmo fibreni) of which the systematic status is still today very debated (Splendiani et al., 2019). Since the middle of 19th century massive stocking activities with the Atlantic strain of S. trutta have been carried out to improve angling opportunities and human consumption so that this species is now considered as one of the world's 100 most invasive species (Jonsson & Jonsson, 2012) imperiling native species by predation, habitat and food competition, and hybridization (Budy et al., 2013).
Indeed, the massive release of brown trout in rivers and streams for angling purposes greatly increases the predation pressure causing a decline in bullhead populations through the predation of juveniles (Lorenzoni et al., 2018;Marconato, 1986;Simon & Townsend, 2003;Zerunian, 2002). Furthermore, the similar ecological preferences and the similar feeding habits between brown trout and bullhead can also lead to interspecific competition phenomena (Elliott, 2006;Holmen et al., 2003;Louhi, Mäki-Petäys, Huusko, & Muotka, 2014). For its great conservation interest, the bullhead is listed under the Annex II of the Council Directive 92/43/EEC (European Commission, 1992) and in the general action plan for Italian freshwater fish conservation (Zerunian, 2003). Thus, while it is classified as a "Least concern" species in the IUCN Vertebrates Red List (Rondinini, Battistoni, Peronace, & Teofili, 2013), the occurring fragmentation of populations and the reduction of the distribution range raise conservation issues in Italy.
The autoecology of the bullhead, such as poor vagility and the requirement of good environmental quality, together with the fact that it is not a species subject to manipulation (i.e., breeding and release for angling purposes, or relocation), could make the bullhead occurrence a useful indicator for evaluating the integrity and the conservation status of freshwater ecosystems (Charles, Subtil, Kielbassa, & Pont, 2008;Tomlinson & Perrow, 2003;Utzinger, Roth, & Peter, 1998), but more knowledge is needed about the causal links among the different stressors and their effects on populations. Thus, based on the hypothesis that the introduction of brown trout can negatively affect bullhead populations due to the increased predation and/or interspecific competition, the aims of the present research were: i. to present a case study of coexisting and long-term noncoexisting populations of bullhead with introduced trout in the same stream of the Orobic Alps (northern Italy); ii. to compare the population status of bullhead with the presence and the absence of trout, to understand in which way, S. trutta affects C. gobio populations in terms of density, growth performance, population structure, and habitat preference.

| Study area
The study site is represented by Nossana stream, a tributary of Serio River located in the Orobic Alps (Province of Bergamo, northern Italy) that flows on the surface for 500 m. The Nossana watershed covers an extension of about 80 km 2 and is constituted by a predominant drainage system towards the subsoil that feeds a spring (Vigna & Banzato, 2015), characterized by a high permeability due to a fracture system within two geological formations, limestone and dolomite (Ferlinghetti, Arzuffi, & Beretta, 2011). Nossana spring ensures the drinking water service to the city of Bergamo and the surrounding municipalities. Two stretches of the stream have been selected: Nossana upstream (NU) and Nossana downstream (ND).
The first is located close to the spring: it is characterized by the presence of a bullhead population because previously released trout have been removed since the early 2000s to avoid the potential proliferation of pathogens in a fish hatchery located downstream.
The second stretch is located before the confluence with the Serio River and is characterized by the presence of both bullhead and trout emispecies populations. Brown trout are periodically released from the hatchery or can go upstream from the Serio River. The two reaches are divided by an insurmountable barrier, constituted by a 7 m high artificial weir, and have the same environmental conditions in terms of width, water depth, channel substrates and habitat availability ( Figure 2).
Nossana stream has a torrential water regime, with a flow rate that can vary between 0.5 and 15 m 3 /s (flow data is assessed by the managing company of the drinking water service): however, a minimum flow of about 0.5 m 3 /s is guaranteed by law due to the presence of the spring water collection activity. The selected stream represents a suitable model to study the S. trutta and C. gobio interaction due to: (i) an intact and morphologically diversi-

| Data collection
Fish data were collected eight times from September 2016 to September 2020 in the two stretches. Each sampling was carried out by two-pass electrofishing using the removal method (Moran, 1951;Zippin, 1958). Fish were captured during low flow periods using electric current, and applying a similar fishing effort in NU and ND (Seber & Le Cren, 1982). For all the captured fish, the species was identified, and individuals were counted to determine the species abundance, in terms of density (number of individuals/m 2 ) and standing crop (g/m 2 ). The total length (TL) was measured to the nearest 0.1 cm for all specimens while the body weight (W) was measured to the nearest 0.1 g. All the fish were released back into their natural environment at the end of the fieldwork. In autumn 2020 we also monitored the habitat occupied by every captured bullhead, both in NU and ND stretches.
To assess the environmental quality and characteristics of the stretches, water temperature was measured in continuous by iButton devices (range À5 to 26 C, resolution: ±0.0625 C, measurement interval: 10 min) fixed in the riverbed of the two stretches from Autumn 2018 to after the end of the fieldwork. Physico-chemical parameters (i.e., electric conductivity, dissolved oxygen, and oxygen saturation) were measured in the field using a probe HACH-HQ40d while pH, nitric nitrogen, ammonia nitrogen, and phosphate were determined in the laboratory according to Standard Methods (APHA/AWWA/WEF, 2012). Macroinvertebrate samplings were also performed on the occasion of each fish sampling with a Surber net (0.10 m 2 , 500 μm mesh) following a standardized multi-habitat sampling procedure (Barbour, Gerritsen, Blaine, & Stribling, 1999). Ten replicates for each stretch were collected then merged in the field and preserved with 96% ethanol. In the laboratory, taxa were identified at the family level according to standard keys. Water quality was evaluated using the LIMeco index for water quality assessment (European Community, 2000; Italian Ministry of the Environment, 2010) while the ecological status was evaluated through the STAR ICMi indicator as reported in Fornaroli, Calabrese, Marazzi, Zaupa, and Mezzanotte (2019). A riverbed habitat map was created in QGIS based on hydraulic and geomorphological data collected for both reaches during the surveys. Water velocity, water depth and size of substrates were monitored at 1 m steps along transects perpendicular to the flow, placed every 5 m. Those data were firstly interpolated to produce thematic maps for each variable (i.e., water velocity, water depth, substrate composition) and then reclassified to create the habitat classification map as it is described in Fornaroli et al. (2016).

| Age and growth
The age of bullhead was determined throughout the analysis of the length-frequency distribution (Bagenal, 1978;Harvey & Cowx, 2000) using the "ELEFAN" function of the TropFishR package (Mildenberger, Taylor, & Wolff, 2017) within R project software (R Core Team, 2020) since that for bullhead is not possible the age determination using direct monitoring as the skin scales.
To estimate the infinitive length (L inf ) and the growth coefficient (K) we performed the moving average over 5 length classes (5 mm each); we identified the optimal L inf value among 30 steps in the range 12-14.5 and, similarly, among 20 steps in the range 0.3-1.5 for the optimal K. We specified a maximum age of the bullhead of 4 years.
The theoretical growth was estimated separately for the two bullhead populations, using the von Bertalanffy growth curve model (Von Bertalaffy, 1938) (Equation (1)) and the mean length of the different age classes was estimated and compared between the two bullhead populations.
where TL t is the total length of the fish at time t, L inf is the theoretical maximum length (cm), K is the rate of approach to L inf , and t 0 is the theoretical age at which TL t = 0.
The total Length-Weight Relationship (LWR) (Froese, 2006;Le Cren, 1951) was estimated for the upstream and downstream populations of bullhead by the least-squares method (Ricker, 1975) based on Equation (2) to compare the growth between the two bullhead populations.
where W = weight, L = length, a and b constants.
Furthermore, the index of growth performance was calculated by the equation of Pauly and Munro (1984) (Equation (3)) for both populations: where K and L inf are the growth parameters of the Von Bertalanffy model.
Multiple working hypotheses were evaluated requiring tests between groups of data classified based on the presence or absence of trout. Because normality and homogeneity of variance could not be achieved in most log-transformed data, differences in population characteristics were tested using nonparametric procedures. We used Wilcoxon Signed Rank test for paired data and Wilcox Rank Sum test (α < .05) to evaluate differences in population traits between treatments. We used the two stretches (NU and ND) as treatments in terms of (i) density of individuals, (ii) standing crop, (iii) body length of each age class, (iv) length-weigh curves, and (v) the frequency of individual of each length class. We used Spearman ρ statistic to estimate a rank-based measure of association among density of individuals of fish species and macroinvertebrate families over time. These comparisons allowed us to identify differences in the structure of the two populations of bullhead and understand the ecological implications of interspecific interaction.

| Habitat preference
Habitat use of bullhead was defined by measuring current velocity, water depth, and substrate type at each fish location. To identify the fish position we used backpack electrofishing, as direct observation via snorkeling was difficult because of water turbulence, and stream bank direct observation was also difficult because of the lack of suitable observation points (i.e., bridges) and small sizes of the fishes (Vismara, Azzellino, Bosi, Crosa, & Gentili, 2001).
Habitat availability in the two study stretches was derived from thematic maps produced for the whole stream with a cell dimension of 1 m 2 . Univariate suitability curves were defined, according to the procedure outlined by Boove (1986): (i) each variable was divided into classes, and frequencies of utilization and availability were computed; (ii) preferences for each class interval of the measured variable were computed from relative frequencies of utilization and availability as follows:

Maps
where P i : relative preference value of a target species for a specific interval of the measured variable, U i : % of utilization of a specific interval of the measured variable, A i : % of availability of a specific interval of the measured variable in the studied river sector at the time the organisms were sampled; (iii) preferences were then normalized to a maximum value of 1.
Univariate suitability curves for adult brown trout were not evaluated in this work due to the very low number of individuals but were available for the mainstem of Serio River, for stretches similar to the Nossana stream in terms of size, flow and morphology (Viganò et al., 2015).

| Environmental characterization and ecological status of the stretches
Water temperature was constant the whole year (8.3 ± 0.2 C) both in NU and ND, due to the presence and proximity of the spring.
Dissolved oxygen ranged from 10.50 to 12 mg/L, close to saturation (>95%). Electrical Conductivity was about 240 ± 40 mS/cm and pH 7.9 ± 0.3. The ecological status of both stretches resulted constantly high since LIMeco and STAR ICMi indicators resulted to be always "elevated" and "good" respectively (LIMeco = 0.859 ± 0.034 in NU and 0.866 ± 0.023 in ND while STAR ICMi = 0.79 ± 0.06 in NU and 0.75 ± 0.05 in ND for the quinquennium) as reported in Table 1.

| Density and standing crop
Salmo (  Wilcox Test showed statistically significant differences between the two bullhead populations (density: V = 28, p = .016; standing crop: V = 28, p = .016): density was 50% higher and standing crop was 20% higher in the NU population of bullhead compared to the ND one ( Figure 3a).

| Demographic characteristics
Five age classes (0þ-4þ) of bullhead were identified by the analysis of the length-frequency distribution and the Von Bertalanffy growth curve equations were extrapolated ( Figure 3b):  (Pauly, 1979).

| Population structure
The population structure was studied investigating the frequency of each dimensional class estimated both based on the length data collected in the field (Figure 3e) and on the age classes obtained by the analysis of the length-frequency distribution (Figure 3f). Both populations presented a good structure even if the ND population had a smaller proportion of 3+ adults (35% vs. 54%) and a higher proportion of 1+ young (27% vs. 15%) as underlined also by the age-length analysis ( Figure 3c).

| Habitat suitability
The bullhead is a benthic species that prefers cold, clear, fast-flowing small streams and middle-sized rivers. According to the meso-habitat survey, the habitat preference of bullhead was different between the two stretches. In ND, the bullhead preference for habitats of middle velocity

| DISCUSSION
The environmental characterization confirmed that Nossana stream represents an optimal environment for bullhead, especially for the presence of stable and diversified stony substrates, high concentration of oxygen and low water temperature. Many studies showed that springs and riverine reaches close to the source are among the habitats preferred by bullhead (Lorenzoni et al., 2018;Tomlinson & Perrow, 2003).  (Table 2).
Nevertheless, the change rate of body shape was about three (2.93 and 2.99 in NU and ND), indicating an isometric growth (Ricker, 1975). This slope was different from those recorded in central Italy (Lorenzoni et al., 2018) and in the Verbano Cusio Ossola Province (Foglini et al., 2018) that showed negative allometric growth.
Conversely, the Tiber basin populations showed a positively allometric growth with a slope value greater than 3 (Lorenzoni et al., 2018) ( Table 2). These deviations could be related to the differences in size structure and contribution of age classes among the compared bullhead populations due to different environmental conditions and predation pressure. For example, the non-optimal corpulence indicator of the Valgrande bullhead (2.05) was ascribed to a low productive environment and high predatory pressure (Foglini et al., 2018). The other Italian populations, included those of Nossana stream, showed a good body shape having a corpulence indicator included in the optimal range (2.6-3.4).
The analysis of the population structure, together with the analysis of length-frequency distribution, showed that, in presence of trout, 3+ bullhead specimens were shorter while 1+ bullhead ones were longer (Figure 3a). In addition, at the same time, 3+ downstream adults were a smaller percentage of the total population while 1+ were a larger one (Figure 3e,f) compared to the upstream population.
This suggests that the trout, besides decreasing the number of bullhead, could influence the population structure by lowering the life expectancy and promoting faster growth in younger individuals.
As shown through the habitat suitability analysis, the presence of trout induced a bullhead suboptimal habitat occupation: C. gobio preferences shifted toward faster flow velocities and shallower depths.
This probably occurs to avoid the salmonid predators, that are more skilled in deeper and slower water, and whose larger specimens tend to occupy those habitats. The two species may compete also for food availability (Irons, Sass, McClelland, & Stafford, 2007) and some diversifications in their diet and microhabitat preferences have been already reported in other studies (Elliott, 2006;Franchi et al., 2012).
The results suggest that interspecific competition phenomena occurred between brown trout and bullhead in the downstream stretch, as observed also in other countries by Elliott (2006), Holmen et al. (2003 and by Franchi et al. (2012) in Central Italy. However, the competition in Nossana stream does not seem particularly strong since the bullhead population keeps good demographic and structural population indexes.

| CONCLUSION
The present case study showed a marked interspecific competition between brown trout and bullhead populations. This hypothesis is

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.