Individual Variation in Movement Behavior of Stream-Resident Mediterranean Brown Trout (Salmo trutta Complex)

Abstract

Understanding individual movement patterns in stream-resident salmonids is critical for conservation and river management, particularly in Mediterranean streams characterized by high environmental variability. We tagged 997 Mediterranean brown trout (Salmo trutta complex) and conducted an 11-month mark–recapture study using Passive Integrated Transponder (PIT) technology to assess movement behavior in the Flamisell River (Ebro Basin, northeastern Iberian Peninsula). Movements followed a leptokurtic distribution, with 81.8% of the individuals classified as sedentary (median movement = 24.9 m) and 18.2% as mobile (median movement = 376.2 m). Generalized linear models revealed distinct drivers of fish movement for each group. In sedentary trout, movement was mainly influenced by mesohabitat type, season, sex, and body size, with males and larger individuals moving farther. In mobile trout, mesohabitat type, density, and body size were key predictors. Movement patterns were repeatable over time, indicating consistent behavioral tendencies. These results support a bimodal movement strategy and highlight the importance of individual variation. Conservation planning should account for both sedentary and mobile groups to preserve functional and genetic connectivity and improve resilience of Mediterranean streams.

1. Introduction

Movement is a fundamental behavioral trait in stream-dwelling fishes. It critically enables individuals to locate and exploit essential habitats throughout their life cycles, respond to changing environmental and biotic conditions, and ultimately optimize fitness components such as growth, survival, and reproductive success [1,2,3]. Therefore, understanding fish movement patterns is essential for identifying appropriate spatial scales for effective population management and conservation strategies. In riverine systems, fish movements are governed by a dynamic interplay of extrinsic and intrinsic factors. These include water temperature [4,5], hydrological variability [6], habitat structure and resource availability [7,8], competitive interactions [9], age [10], and sex [11]. This ecological complexity generates substantial inter- and intra-specific variation in movement behaviors across both spatial and temporal scales [12,13].

Among salmonids, the brown trout (Salmo trutta complex) is a particularly well-studied model species, largely due to its ecological plasticity and its importance to both recreational and commercial fisheries [14,15]. Stream-resident brown trout populations exhibit a wide range of movement strategies, from sedentary individuals with home ranges of less than 50 m to highly mobile conspecifics capable of displacing several kilometers, particularly in response to seasonal shifts in the environment (e.g., water flow, temperature, food availability) or reproductive cues [16,17,18]. This variation in mobility often results in leptokurtic movement distributions, where the majority of individuals are sedentary and a minority exhibit large displacements [12,19]. The intrapopulation variability is influenced not only by extrinsic environmental factors but also by intrinsic traits such as body size, sex, and behavioral phenotypes, including boldness and exploratory tendency [20,21]. Previous studies have demonstrated that such behavioral traits often correspond to stable, repeatable movement patterns, underscoring the role of individual personality in shaping dispersal strategies [22]. Integrating these behavioral traits with environmental drivers is essential for understanding fish movement, assessing functional connectivity, predicting dispersal patterns, and developing conservation strategies to preserve both ecological functionality and genetic diversity [23].

In Mediterranean streams, brown trout exhibit genetically differentiated lineages within the Salmo trutta complex, reflecting a highly structured evolutionary history shaped by long-term geographic isolation and local adaptation [24]. Mediterranean populations are subjected to additional selective pressures such as highly variable hydrology and the increasing frequency of climate anomalies [25]. These environmental stressors have been associated with heightened dispersal tendencies and movement patterns that diverge from those populations in the more hydrologically stable systems of northern Europe. For example, a genetic study conducted in a Mediterranean river in northeastern Spain identified trout family groups dispersing over several kilometers, suggesting an adaptive response to harsh and unpredictable environmental conditions [26]. Nevertheless, studies directly monitoring individual movement in Mediterranean brown trout populations are scarce.

Consequently, we studied the spatial dynamics of individually tagged brown trout with Passive Integrated Transponder (PIT) tags over a year in a headwater stream on the Iberian Peninsula. Our primary objective was to identify the drivers of movement and to examine the extent of inter-individual variation within the population. The specific objectives of this study were (i) to quantify movement distances and their temporal variation; (ii) to evaluate the degree and consistency of behavioral variation among individuals; and (iii) to determine the relative contributions of environmental factors (e.g., mesohabitat type and population density) and individual traits (e.g., sex and body length) to observed movement patterns.

2. Materials and Methods

2.1. Study Area

Brown trout movements were monitored in the Flamisell River (Figure 1), a tributary of the Noguera Pallaresa River in the Ebro Basin (northeastern Iberian Peninsula). It originates in the Pyrenees and flows southward for about 31 km draining a watershed of 583 km². The study reach (central point: 42°23′46″ N, 0°58′1″ E) was 3560 m long, with an average elevation of ca. 900 m above sea level (Figure 1). A weir at the upstream boundary prevents fish movement beyond the reach, which otherwise lacks physical barriers or major tributaries. The reach had a mean wetted width of 5.9 ± 1.7 m (µ ± SD) and an average depth of 35 ± 24 cm, with a channel slope of 53.3 m/km. Its morphology was primarily composed of pool–run–riffle sequences under a forested canopy. River hydrology is regulated by snowmelt in spring and increased rainfall in autumn, with lower flows in summer and winter (Ebro Water Authority; https://www.chebro.es, accessed on 07/04/2025). Brown trout of the Mediterranean lineage is the sole fish species present [27]. Recreational fishing pressure was negligible during the study period, with management regulations enforcing catch-and-release angling.

2.2. Fish Sampling and PIT Tagging

Brown trout were sampled in April 2013 from a 2100 m core tagging zone within the 3560 m study reach (Figure 1). The core tagging segment was flanked by two buffer zones extending 800 m upstream and 1060 m downstream, used to detect potential fish movement beyond the tagging area. The reach was divided into 72 discrete sections based on mesohabitat units (pools, riffles, and runs), which were classified according to the velocity-to-depth ratio [28]. Fish were captured by electrofishing (Hans-Grassl GmbH ELT60-IIH backpack electrofisher, DC pulsed, 1.3 kW) and temporarily held in section-specific keep-nets. A total of 997 individuals with a fork length (FL) range between 80 and 310 mm were selected for PIT tagging. Fish were anesthetized with clove oil (1 mL/20 L stream water), measured for FL (to the nearest mm), and weighed (to the nearest 0.1 g). Two sizes of half-duplex (HDX) PIT tags (Oregon RFID, Portland, OR, USA) were used: 12 mm tags (0.1 g) for trout 80–120 mm FL, and 23 mm tags (0.6 g) for trout >120 mm FL (Figure 2). Tags were inserted into the peritoneal cavity via a ~5 mm lateral-ventral incision using a sterilized needle and injector. Incisions were treated with topical antiseptic, and the adipose fin was clipped to enable visual identification upon recapture. No mortality was observed during fish handling and tagging procedure. Clipped fin tissue was preserved in 95% ethanol for molecular sex determination, returning 606 females (65.4%) and 320 males (34.6%). Thus, overall sex ratio was significantly skewed (G = 1194, df = 1, p < 0.0001) to females (1.89:1). Tagged fish were released at their point of capture after full recovery. A second electrofishing survey was conducted in November 2013 to assess tag retention and detection efficiency. All procedures adhered to applicable Spanish and EU regulations for animal research and welfare.

2.3. PIT Detection Surveys

The spatial locations of PIT-tagged brown trout were determined using a portable half-duplex (HDX) PIT tag detection system [29,30]. This system comprised an HDX Backpack PIT Reader (Oregon RFID, Portland, OR, USA) connected to a waterproof antenna mounted on a 2.5 m PVC pole. The maximum detection range varied between 30 and 55 cm, depending on PIT tag orientation. Detection surveys were conducted at approximately monthly intervals from June 2013 to February 2014 (N = 8 surveys). May survey was not possible due to high discharge from snowmelt, which precluded safe wading. In each survey, the entire 3560 m study reach was systematically scanned by wading upstream, ensuring full coverage of the wetted area. The operator maneuvered the antenna laterally across the channel, from bank to bank, covering the entire channel width. In deeper sections (>30 cm), the antenna was fully submerged to maximize detection efficiency. Upon detection of a PIT tag, the following data were recorded: unique tag code, date and time of detection, geographic coordinates (collected using a Leica Zeno GG03 GPS unit; Leica Geosystems, Heerbrugg, Switzerland), and mesohabitat type (pool, riffle, or run). To distinguish live fish from expelled tags, the operator disturbed the substrate near the detection point to flush fish from cover. Based on fish response, each detection was categorized as (i) live trout, the fish was visually confirmed or the signal ceased upon disturbance, indicating movement; (ii) possible live trout, the fish was not seen but the signal persisted following disturbance; or (iii) tag loss, the signal originated from clearly unsuitable habitat (e.g., shallow gravel bars or dry areas), suggesting mortality or tag expulsion. After each survey, detection histories were reviewed. Tags initially categorized as ’possible live trout’ were reclassified as ‘live trout’ if subsequent detections demonstrated active movement (i.e., a change in spatial location over time). Tags that did not meet this criterion, as well as those categorized as tag loss, were excluded from subsequent analyses.

2.4. Detection Efficiency Assessment

Detection efficiency was evaluated during the November survey by electrofishing (two-pass removal method) immediately after PIT scanning in two different 200 m sections of the study reach. Estimates of capture probability, derived from the electrofishing data, were applied to the number of recaptured tagged fish to estimate the total number of tagged individuals present in each section. Detection efficiency was then calculated as:

$$Detection efficiency={\displaystyle \frac{Number of detected tags}{Estimated number of tagged fish present}}$$

2.5. Data Analyses

Fish population size (N) for stream sections was estimated using a two-pass removal electrofishing method during the April tagging survey, following a method described by Seber and Le Cren [31]:

$$N={\displaystyle \frac{C_1^2}{C_1{-C}_2}}$$

where C₁ and C₂ are the number of fish captured in the first and second pass, respectively. Fish density (individuals/m²) per section was calculated by dividing N by the sampled stream area.

Fish movement distances were quantified in QGIS [32] by measuring the longitudinal distance (m) between successive detection positions along the stream centerline. Movement direction was indicated by sign, positive values denoted upstream movement, and negative downstream. Absolute movement distances were used to quantify total displacement. Directional bias in fish movement (i.e., whether the mean displacement differed significantly from zero) was assessed with one-sample t-tests. To evaluate individual variability in movement behavior, the distribution of movement distances was tested for normality using the Shapiro–Wilk test, and kurtosis was calculated following the method described by Turchin [33]. Elevated kurtosis indicates evidence of leptokurtic (i.e., fat-tailed) distributions, and hence disproportionate contributions from highly mobile individuals [19].

Fish dispersal was modeled using the R package fishmove v0.3-3 [34]. This package implements a leptokurtic dispersal model based on a double-normal probability density function, enabling the estimation of the mean movement distance for the sedentary (σ_sed) and mobile (σ_mob) components, as well as the proportion of sedentary individuals within the total population (p). The fishmove package also predicts movement parameters for different species using multiple regression models that incorporate fish length, caudal fin aspect ratio, stream order, and time between observations [34]. Movement predictions were generated for a mean fork length of 165.1 mm (derived from April and November electrofishing surveys), Salmo trutta fario as species, stream order of 3, and time interval based on the interval between tagging and detection. Predicted values were compared to field estimates; thus, non-overlapping confidence intervals were considered indicative of significant differences.

Differences in maximum linear displacement (MLD, the distance between the most distant detection positions) between sex groups were analyzed with analysis of covariance (one-way ANCOVA), using fork length as covariate. Analysis of variance (two-way ANOVA) was used to compare trout fork length between mobile groups (i.e., mobile vs. sedentary), and sex groups. The proportion of gender categories was compared between mobile groups with a G-test of independence.

The effects of both environmental factors and fish traits on trout movement were analyzed with Generalized Linear Models (GLMs). Only those fish detected at least three times during the study period were included in the analysis. An information-theory approach was implemented to find the best predictive models following the methodology described by Burnham and Anderson [35]. GLMs were built including all possible combinations of independent variables (i.e., initial and final habitat, initial and final density, behavioral group, sex, fork length, and month), but excluding interactions due to the large number of variables included. Initial (habitat and density) refers to the condition of the section from which the individual departed, and final to the section to which the individual moved; the behavioral group classifies individuals as sedentary or mobile, and fork length was measured when tagging. Only those models performing significantly better than the null model and with a variance inflation factor ≤ 5, to avoid multicollinearity effects in regression models [36], were considered as candidate models. The degree of support for each candidate model was assessed with the second-order Akaike Information Criterion (AICc), rescaled to obtain ΔAICc values (ΔAICc = AICci − minimum AICc) [37]. Finally, since models with ΔAICc > 4 have less support, they were omitted from further consideration. The relative plausibility of each candidate model was assessed by calculating Akaike’s weights (w_i), which range from 0 to 1 and can be interpreted as the probability that a given model is the best model in the candidate set. Because there was not a clearly best model (i.e., w_i ≥ 0.9), we averaged regression coefficients (β_i) by weighing selected model coefficients by model w_i. The relative importance of each variable (i.e., the selection probability, SP) was also calculated by the sum of w_i for all models in which a given variable occurs, which estimates the importance of an independent variable for differentiating the response variable (see Burnham and Anderson [35]). In addition to SP, we also provide standardized estimates of β_i based on partial standard deviations for their variables following Cade [38]. See [18,39] for more details regarding model criteria selection. Model selection was performed with R software version 4.5.0 [40]; MuMIn 1.48.11 was used for multi-model inference analysis; car 3.1-3 was used for VIF analysis of each of the candidate models; and Nortest 1.0-4 was used for normality test analysis.

To assess consistency in individual movement behavior, a partial correlation analysis was conducted between sequential movement distances for fish with ≥2 recorded movements, controlling for fork length [41]. Prior to analysis, quantitative variables were log-transformed to improve linearity and homoscedasticity.

3. Results

3.1. Detection Efficiency and General Movement Patterns

A total of 997 individuals were captured and fin-clipped during the study. Tagged fish had a mean FL of 142.3 ± 36.7 mm (µ ± SD) and a mean weight of 39.1 ± 34.4 g. In the November electrofishing survey, 131 of 135 recaptured trout had retained their PIT tags, resulting in a tag retention rate of 96.3%. All recaptured fish exhibited complete healing at the tag insertion site, with no signs of external infection or abnormalities. The average capture probability of electrofishing was estimated at 0.52, while the mean detection efficiency for PIT-tagged individuals was 0.51. The number of detected individuals decreased over the study period, from 377 in July 2013 to 184 in February 2014 (µ ± SE: 299 ± 24 detections per sampling event). Over the entire study period, 70.4% of the tagged fish were detected at least once.

Overall, brown trout in the Flamisell River exhibited limited movement throughout the study. For individuals with three or more detections (N = 389), the mean maximum linear displacement (MLD, the distance between the most distant detection positions) was 130.01 ± 15.03 m, ranging from 0 to 2013 m. Of these, 96 individuals (24.7%) exhibited displacements of 20 m or less, 313 individuals (80.5%) displaced 100 m or less, and only 43 individuals (11.1%) moved more than 200 m. The median MLD across all tagged individuals was 40 m. The MLD was not significantly correlated with fork length (ANCOVA, F_{1, 386} = 1.78; p = 0.18), but it differed significantly between sexes (F_{1, 386} = 9.20; p = 0.003). Males moved significantly more than females, with mean displacements of 154.2 ± 27.2 m and 117.1 ± 17.9 m, respectively.

Between consecutive tracking surveys, the median distance moved by tagged individuals ranged from 0 to 1 m in all sampling events, except for June, when the median movement increased to 8 m (Figure 3). Directional movements (defined as upstream = positive, downstream = negative) were predominantly upstream across most sampling periods, except in December and February, when movements were primarily downstream. Mean directional movement reached a maximum in October (44.51 ± 16.60 m) and a minimum in December (−2.89 ± 9.28 m) (Figure 3). However, directional movement was significantly different only in October (Student’s t = 2.68, df = 143, p = 0.004), when it was mostly upstream (Figure 3).

3.2. Mobile vs. Sedentary Individuals

The distance-frequency distribution derived from tracking data was highly right-skewed and leptokurtic, indicating a significantly greater than expected frequency of both short- and long-distance displacements relative to a normal distribution. Overall, the kurtosis value was 101.9 ± 0.16 (µ ± SE), exhibiting marked temporal variation. It ranged from a minimum of 7.91 ± 0.65 in December to a maximum of 151.2 ± 0.34 in August (Figure 4). The two-group exponential model for stream-resident brown trout, implemented using the fishmove package, provided a good fit to the observed dispersal frequency distribution (Figure 5). The estimated mean dispersal range (calculated as 1/λ) was 35.97 m for the sedentary group (representing 81.8% of the population) and 542.8 m for the mobile group (18.2% of the population). However, model parameters and derived estimates varied over the course of the study (

Table 1. Mean movement distances (in meters) and standard errors (SE) for both the sedentary (σ_sed) and the mobile (σ_mob) components of trout population, estimated from the two-group exponential model along the study period. The proportion of the sedentary group (p) is also shown.

Month	σsed ± SE	σmob ± SE	p ± SE
June	20.25 ± 0.92	303.60 ± 36.47	0.918 ± 0.015
July	20.58 ± 0.77	408.98 ± 40.50	0.912 ± 0.013
August	27.94 ± 1.28	702.66 ± 84.09	0.912 ± 0.015
October	30.61 ± 1.22	923.97 ± 86.71	0.859 ± 0.018
November	31.00 ± 1.86	682.37 ± 63.21	0.787 ± 0.026
December	26.23 ± 2.75	695.36 ± 93.64	0.791 ± 0.039
February	22.72 ± 2.18	619.39 ± 115.40	0.823 ± 0.044

). The proportion of sedentary individuals (p) declined from June to November and December (Table 1), reflecting increased mobility in late autumn. The mean dispersal range of the sedentary group (σ_sed) increased from 20.25 m in June to a peak of 31.00 m in October, before decreasing to 22.72 m by February. Similarly, the dispersal range of the mobile group (σ_mob) showed a marked increase from 303.60 m in June to 923.97 m in October, followed by a decline to 619.39 m in February (Table 1). Despite generally limited movement, 66.5% of sedentary individuals were detected in two or more distinct mesohabitat units, suggesting that these fish were not strictly site-attached but regularly shifted among adjacent habitats. Model predictions generated by fishmove closely matched empirical observations, with overlapping confidence intervals (Figure 6). However, during the spawning period and preceding months, observed movement distances in the mobile group exceeded model predictions, probable due to behavioral changes associated with reproduction not fully captured by the predictive model (Figure 6).

Overall, the sex ratio of brown trout did not vary between mobile and sedentary groups (G = 1.61, df = 1, p = 0.21). Furthermore, fish fork length did not show significant differences between sexes (ANOVA, p = 0.95), mobile groups (p = 0.37), or their interaction (p = 0.29). Therefore, differences between mobile and sedentary groups could not be explained by differences in size-class or sex composition.

3.3. Influence of Intrinsic and Extrinsic Factors on Movements

Several intrinsic and extrinsic factors were associated with brown trout movement patterns. An information-theoretic approach applied to a set of candidate GLMs identified seven plausible models (ΔAICc ≤ 4) explaining variability in trout movements. The best explanatory variables were initial habitat, sampling month, sex, fork length, and group classification (sedentary vs. mobile), as indicated by standardized parameter estimates and model selection probabilities (SP and β_pSD;

Table 2. Results from the information-theoretic framework analysis (Generalized Linear Models) to assess the influence of both intrinsic and extrinsic factors on brown trout movements in the Flamisell River. Model-averaged coefficients (β) are parameter coefficients; selection probability (SP) and standardized estimates of β based on partial standard deviations (β_pSD) indicate the relative importance of each independent variable. Parameter bias is defined as the difference between model-averaged estimates and the corresponding full model coefficients. The table also includes the number (N) of plausible candidate models (∆AICc < 4) and the Pearson correlation coefficient (r) between observed and model-predicted values.

Model Parameter	Global Model (N = 7; r = 0.37)
	SP	β ± SE	Bias	βpSD ± SE	BiaspSD
Intercept	1.000	0.372 ± 0.481	0.231
HabitatIni—Run	1.000	−0.075 ± 0.087	−0.661	−0.075 ± 0.087	−0.075
HabitatIni—Riffle	1.000	−0.182 ± 0.092	−0.271	−0.170 ± 0.086	−0.053
July	1.000	−0.267 ± 0.054	0.021	−0.208 ± 0.042	−0.661
August	1.000	−0.206 ± 0.052	0.027	−0.170 ± 0.043	−0.271
October	1.000	−0.304 ± 0.058	0.012	−0.214 ± 0.041	−0.306
November	1.000	−0.125 ± 0.065	0.023	−0.077 ± 0.040	−0.058
December	1.000	−0.053 ± 0.088	0.026	−0.022 ± 0.037	−0.021
February	1.000	−0.370 ± 0.104	−0.011	−0.131 ± 0.037	0.023
Sex—Male	1.000	0.094 ± 0.037	−0.005	0.090 ± 0.036	0.026
Group—sedentary	1.000	−0.295 ± 0.070	0.007	−0.152 ± 0.036	−0.011
Log(Fork Length)	0.951	0.472 ± 0.198	−0.021	0.085 ± 0.036	0.021
Log(DensityEnd)	0.759	−0.961 ± 0.577	−0.306	−0.107 ± 0.064	0.027
HabitatEnd—Run	0.687	0.167 ± 0.074	−0.075	0.166 ± 0.073	0.012
HabitatEnd—Riffle	0.687	0.164 ± 0.080	−0.053	0.155 ± 0.075	−0.005
Log(DensityIni)	0.509	0.918 ± 0.513	−0.058	0.104 ± 0.058	0.007

). In contrast, initial density and final habitat had the weakest relationship with trout movements. Fish group was a key factor explaining movements, indicating some differences between sedentary and mobile individuals. Thus, both groups were analyzed separately, improving model performance (see Pearson’s r values in Table 2 and

Table 3. Results from the information-theoretic framework analysis (GLMs) to assess the influence of both intrinsic and extrinsic factors on brown trout movements in the Flamisell River, for both the sedentary and mobile group. Model-averaged coefficients (β) are parameter coefficients; selection probability (SP) and standardized estimates of β based on partial standard deviations (β_pSD) indicate the relative importance of each independent variable. Parameter bias is defined as the difference between model-averaged estimates and the corresponding full model coefficients. The table also includes the number (N) of plausible candidate models (∆AICc < 4), and the Pearson correlation coefficient (r) between observed and model-predicted values.

Model Parameter	Sedentary Trout Model (N = 8; r = 0.43)					Mobile Trout Model (N = 7; r = 0.72)
	SP	β ± SE	Bias	βpSD ± SE	BiaspSD	SP	β ± SE	Bias	βpSD ± SE	BiaspSD
Intercept	1.000	0.505 ± 0.507	0.541			1.000	−3.994 ± 2.297	−0.233
HabitatIni—Run	1.000	−0.134 ± 0.071	−0.083	−0.141 ± 0.075	−0.083
HabitatIni—Riffle	1.000	−0.292 ± 0.080	−0.044	−0.290 ± 0.080	−0.044
HabitatEnd—Run	1.000	0.204 ± 0.073	−0.072	0.215 ± 0.077	−0.072	0.114	0.108 ± 0.304	−4.374	0.073 ± 0.205	−4.374
HabitatEnd—Riffle	1.000	0.245 ± 0.081	−0.050	0.247 ± 0.081	−0.050	0.114	−0.184 ± 0.310	0.141	−0.108 ± 0.183	0.141
July	1.000	−0.233 ± 0.053	0.007	−0.194 ± 0.044	0.007
August	1.000	−0.205 ± 0.051	0.016	−0.178 ± 0.044	0.016
October	1.000	−0.286 ± 0.057	−0.002	−0.215 ± 0.043	−0.002
November	1.000	−0.113 ± 0.063	0.004	−0.074 ± 0.041	0.004
December	1.000	−0.078 ± 0.086	0.013	−0.035 ± 0.039	0.013
February	1.000	−0.457 ± 0.105	−0.007	−0.167 ± 0.038	−0.007
Sex—Male	1.000	0.100 ± 0.037	−0.008	0.100 ± 0.037	−0.008	0.183	0.036 ± 0.195	1.911	0.024 ± 0.132	1.911
Log(Fork Length)	0.649	0.351 ± 0.195	−0.011	0.067 ± 0.037	−0.011	1.000	2.588 ± 1.079	−0.173	0.316 ± 0.132	−0.173
Log(DensityEnd)	0.411	−0.488 ± 0.465	−0.556	−0.058 ± 0.055	−0.556	0.920	−4.969 ± 2.336	0.042	−0.377 ± 0.177	0.042
Log(DensityIni)	0.323	0.301 ± 0.514	−0.997	0.037 ± 0.063	−0.997	0.419	3.602 ± 2.552	0.604	0.248 ± 0.175	0.604

, Figure 7).

The sedentary group showed a similar pattern to the global model (Table 2 and Table 3), since most of the individuals were sedentary (Table 1). Eight models were selected as candidates, in which habitat type, sampling month, and sex were again the most influential predictors for trout movement, while fork length, initial density, and final density showed intermediate to weak importance (Table 3). Conditional residual revealed that individuals previously detected in pools moved farther when transitioning to runs and riffles. Fork length was positively associated with movement, males movements were larger than females, and movements significantly varied seasonally, being greatest in June, November, and December, and lowest in February (Figure 8).

For mobile individuals, seven models were retained as plausible, and only final habitat, fork length, and final density were the most influential predictors, while initial density had moderate importance (Table 3) in trout movement. Fork length again showed a positive relationship with movement. A contrasting pattern was observed for density: movement increased with higher initial density but decreased with higher final density (Figure 9). Mobile trout did not exhibit significant seasonal variation in movement, suggesting a consistent mobility pattern across time.

For individual fish, movements between consecutive tracking events were highly correlated (Pearson’s r = 0.26, N = 656, p < 0.0001), indicating consistent inter-individual differences in movement behavior. Partial correlation analysis (controlling for fish length) showed a significant positive relationship between successive movement distances for both females (r = 0.33; Student’s t = 6.91, df = 399, p < 0.0001) and males (r = 0.14; Student’s t = 2.97, df = 223, p = 0.04). These findings suggest that past movement predicts future movement, highlighting the presence of repeatable individual variation in mobility within the population.

4. Discussion

4.1. Overall Movement Patterns and Mobile Fish Ratio

Movement patterns of brown trout in the Flamisell River exhibited a markedly leptokurtic distribution, characterized by a high frequency of short-range displacements (<50 m, classified as restricted movement [12]) and infrequent long-distance movements. These longer movements were more common in October, preceding the spawning season, but occurred in other months. This movement pattern supports the coexistence of both sedentary and mobile individuals within the population, a phenomenon widely documented in stream-resident salmonids (e.g., [1,12,21]). The observed proportion of mobile individuals (18.2%) was consistent with previously provided estimates, such as the median value of 19% for salmonid species [12], and proportions of 13–20% reported for other brown trout populations [16,42]. However, the 45.7% reported in [43] suggests that local ecological context may influence the mobility ratio. The fishmove model provided a reliable framework for estimating movement parameters, with predicted values generally consistent with field estimates, both for the proportion of sedentary and mobile individuals and their displacement distances. The main deviation between predicted and observed values was in the mobile component during the pre-spawning period, when observed movement distances exceeded model predictions. This discrepancy reflects the heightened migratory activity associated with reproduction, in agreement with previous studies [6,44].

Our findings in the Flamisell River contrast with previous inferences based on genetic population structure from a nearby river in the same basin, which suggested notable dispersal by adult brown trout [26]. In that study, the random spatial distribution of related individuals in older age classes, as opposed to the localized clustering in juveniles, was interpreted as evidence of adult movement over several kilometers. This pattern was mediated by ecological constraints such as low productivity, variable flow regimes, and limited spawning habitat typical of Mediterranean streams. In contrast, our results show that most trout were sedentary, in concordance with other studies in Mediterranean systems [6,18]. The discrepancy between genetic and direct-tracking methods is common [45,46], and may stem from several sources. Genetic signals of long-distance movement may reflect rare or historical dispersal events across generations, while mark–recapture methods may miss early-life or non-detectable movements [46,47]. The relative contribution of these factors remains unknown, but the prevailing evidence from movement studies suggests that adult trout in Mediterranean streams are largely sedentary. However, rare long-distance movements likely play a critical role both in maintaining genetic connectivity and ensuring resilience to high and often catastrophic natural hydrological variability of Mediterranean streams.

Our results on trout movement are based on recaptured tagged fish (i.e., surviving individuals); thus, several potential sources of bias could have affected our results. For instance, tag insertion and adipose fin clipping may affect behavior or mortality rate compared to non-tagged individuals. However, several studies on salmonids have shown that tagging and adipose fin clipping do not affect condition, growth, or survival (e.g., [48,49,50]), and therefore the behavior of tagged fish is representative of the population. The 29.6% of the PIT-tagged fish were never detected. These individuals may have been in the study reach but not detected, may have died, or could have moved outside the study boundaries. Escaping while the operator is scanning for fish could be an important source of missing tags. Fish predation could also be important since several piscivorous birds and mammals (e.g., grey heron Ardea cinerea, great cormorant Phalacrocorax carbo, and river otter Lutra lutra) were frequently observed at the study site. Indeed, some PIT tags were recovered from otter spraints and a cormorant’s stomach contents, thus suggesting that predation was a major cause of PIT tag loss. In addition, the annual mortality rate of adult brown trout is estimated to range between 43 and 72% [51]. Although detection efficiency was relatively low, our estimate was similar to that reported in previous studies on the movement of salmonids in open systems (e.g., [50]). Furthermore, while tag loss biases abundance estimates, movement estimates are not affected since behavioral differences between tagged and tag-loss fish are not expected (e.g., [48,49,52]). Movements of tagged trout out of the study area could have occurred, but they were probably limited. The upstream end of the study reach is bounded by an impassable weir, and fish could only leave the study area downstream. However, less than 3% of the total PIT tag detections occurred outside of the central tagging section, suggesting that the length of the study reach was adequate for the detection of most movements and did not significantly affect our overall results.

4.2. Drivers of Movement Variability

The observed intrapopulation variability in movement appears to stem from the interaction between local environmental conditions and individual-specific traits. Among the extrinsic factors, mesohabitat type and population density, both varying at local spatial scales, influenced brown trout movement patterns in the Flamisell River, but differed between mobile and sedentary individuals. Sedentary fish occupying riffle and run habitats exhibited significantly shorter movement distances than individuals in pool habitats. This pattern was not observed in mobile fish, thus suggesting divergent drivers of movement in the two groups. The reduced mobility of sedentary individuals in faster-flowing mesohabitats may reflect adequate feeding opportunities and cover from predators in these environments. As a drift-feeding species, brown trout benefit from the continuous delivery of drifting aquatic invertebrates facilitated by increased current velocity, thereby increasing habitat suitability and feeding efficiency [53]. In contrast, the movement patterns of mobile individuals were not associated with specific mesohabitats, implying that other intrinsic or extrinsic factors govern their movement decisions. In addition to habitat features, fish density also emerged as a significant predictor of individual mobility, highlighting the role of social interactions in shaping movement behavior. Our model showed that individuals located in high-density sections during one sampling event were more likely to relocate to lower-density areas in subsequent surveys. This pattern is indicative of density-dependent dispersal, likely driven by competition avoidance. In salmonids, elevated conspecific density is known to promote redistribution and movement, functioning as a mechanism to reduce intraspecific competition and optimize habitat use [54]. In fact, both sedentary and mobile groups exhibited this response, although the effect was stronger in mobile individuals. This heightened responsiveness of mobile fish under crowded conditions may reflect also a subdominant status within the social hierarchy, as subordinate fish are more likely to be displaced from high-quality territories by dominant conspecifics [55].

Among intrinsic traits, sex was a significant predictor of movement primarily in sedentary individuals, with males moving approximately 1.4 times farther than females. In contrast, the effect of sex was weak among mobile individuals, although males in this group also tended to move farther than females. Sex-biased movement patterns have been observed in other salmonids (e.g., [11]) and are consistent with polygynous mating systems, where male reproductive success is often limited by access to mates, while female success is constrained by fecundity [56]. Increased male mobility may therefore represent an adaptive strategy to maximize reproductive opportunities across spatially structured environments. These differences in movement behavior have important implications for gene flow and recolonization dynamics. Greater male dispersal may facilitate genetic connectivity among subpopulations, whereas limited female movement could constrain demographic expansion and reduce the rate at which new or recovering habitats are colonized. Fish length was also significantly associated with movement distances for both sedentary and mobile individuals. In salmonids, numerous studies have reported a positive relationship between body size and mobility, with larger individuals typically displacing greater distances (e.g., [57,58]). This positive correlation between body size and mobility can be attributed to several factors; for instance, larger individuals often possess greater swimming capacity, enabling them to navigate through stronger currents and overcome physical barriers more effectively [59]. Additionally, larger salmonids may have higher energetic demands, prompting them to explore broader areas to meet their metabolic needs [60].

Consistent individual differences in movement behavior—evidenced by repeatable movement patterns across recaptures—indicate the presence of stable behavioral traits that may reflect aspects of fish personality [20,22]. These patterns were observed across both sexes and were independent of body size, suggesting that such behavioral consistency is not merely a byproduct of morphological or sex-related variation. Similar findings in other studies reinforce the interpretation of movement behavior as a personality-linked trait [61,62], which is likely to have a genetic basis [63]. Supporting this, metabolomic analyses in brown trout revealed distinct physiological profiles between sedentary and mobile individuals, particularly in energy metabolism and stress response [64].

5. Conclusions

Overall, this study adds to the growing body of evidence that stream-resident brown trout populations are composed of individuals exhibiting diverse movement strategies. This ecological complexity has significant implications for conservation and management. While sedentary individuals may be more susceptible to local habitat degradation, mobile individuals face greater risks from landscape-scale fragmentation. The presence of physical barriers may selectively disadvantage mobile individuals, potentially reducing their frequency and threatening long-term population resilience [65]. Even in populations with predominantly sedentary strategies, maintaining genetic and behavioral diversity, including mobility traits, is vital for ensuring population connectivity, recolonization potential, and adaptability to environmental change. Effective management must therefore adopt multi-scale approaches that preserve both local habitat integrity and regional connectivity to sustain the ecological and evolutionary viability of brown trout populations in dynamic stream environments.