Daily nocturnal homing reveals goal-directed navigation in a wild teleost Authors: Milan Říha, Ruben Rabaneda-Bueno, Ivan Jarić, Marie Prchalová, Petr Blabolil, Marek Šmejkal, Tomáš Jůza, Michal Tušer, Jan Kubečka, Jiří Peterka Published: 2026 ABSTRACT SummaryUnderstanding how free-ranging animals navigate large, low-visibility environments remains an important challenge in movement ecology. Using high-resolution acoustic telemetry, we investigated the movement behavior of adult pikeperch (Sander lucioperca) in a large freshwater reservoir. Individuals repeatedly left a stable resting place to forage across extensive areas and reliably homed back even after multi-kilometer nocturnal excursions lasting several hours. Return trajectories were initiated under dark or low-light conditions and were strongly oriented toward the refuge from the onset of movement. Bearings during the initial phase of return showed low deviation from the direct direction to the goal, and trajectories were characterized by stable swimming speeds and low tortuosity. Final approach distances were small relative to excursion scale, and returns were achieved without extended search behavior. Together, these patterns demonstrate consistent and repeatable daily homing under low-visibility conditions and are consistent with goal-directed navigation toward a fixed spatial target.Subject areas: Biological sciences Introduction To navigate successfully, animals must determine their position relative to a goal and select an appropriate heading, a process often conceptualized as a “map and compass” framework.1,2 Navigation can rely on multiple spatial strategies, ranging from egocentric mechanisms to allocentric strategies based on stable external references and stored spatial memory.3,4,5 These processes support movement both within familiar environments, where individuals return to previously learned sites, and from unfamiliar locations, where large-scale gradients or map-and-compass mechanisms may be required.6,7 Yet, the relative contributions of these strategies and the extent to which animals depend on internal spatial representations remain challenging to disentangle empirically.8 Various navigation strategies have been well documented in mammals and birds or other animals.9,10 In fishes, however, these abilities have received far less attention particularly under natural field conditions.11,12 Laboratory experiments with teleosts, particularly goldfish (Carassius auratus), show that they can combine egocentric and allocentric frames of reference in maze tasks and adjust to rotated environments using external cues.13 Fish are also capable of piloting by learning and recalling sequences of landmarks, and can integrate multiple cues to navigate flexibly when routes are altered.3,14,15,16 Evidence further suggests that some fish species can perform path integration, encoding both distance and direction traveled relative to a home site.17 Although laboratory experiments show that fish can use multiple navigational strategies, these studies are typically conducted in spatially constrained and structurally simplified environments that differ fundamentally from natural habitats in both scale and complexity.18 Field observations indicate that navigation in nature often involves repeated movements between ecologically important locations.3,14,15 Many species exhibit strong site fidelity and homing, repeatedly returning to specific refuges, spawning grounds, or foraging areas after displacement or prolonged absences.19 Such behavior is consistent with the use of spatial memory and orientation processes that allow individuals to relocate ecologically important sites across time. These abilities likely contribute to efficient resource use, refuge selection, and predator avoidance.20,21 However, despite well-documented homing behavior, the navigational processes supporting such performance under natural conditions remain poorly understood. Existing field-based knowledge of fish navigation is dominated by studies of large-scale migrations in species such as salmonids, tunas or coral reef fish larvae, where orientation across vast distances is often supported by non-visual mechanisms, including magnetoreception, celestial compass and olfactory or auditory cues.12,22,23,24,25,26,27 Conversely, fine-scale homing is studied in visually structured environments, such as coral reefs, and is typically supported by abundant, stable visual landmarks.28,29,30,31 Comparatively, little is known about how fish achieve consistent homing within structurally simpler or more homogeneous freshwater systems. In turbid or dark environments, visual landmarks are reduced, and the availability and reliability of alternative orientation cues may differ from those in long-distance migratory contexts or visually rich habitats. Despite the ecological importance of these habitats, quantitative field-based analyses capturing the onset and directional structure of homing trajectories under reduced visual guidance are missing. Here, we use the pikeperch (Sander lucioperca) as a model to investigate homing behavior in a large, free-ranging freshwater predator operating under low-visibility conditions. This ecologically and commercially important nocturnal species exhibits strong site fidelity in turbid freshwater systems,32,33 making it well suited for examining repeated homing under visually degraded environments. Using long-term, high-resolution acoustic telemetry, we tracked adult individuals during nocturnal foraging excursions and quantified the orientation, onset, and precision of their return trajectories. By analyzing the directional structure and temporal consistency of these movements, we test whether homing behavior is consistent with goal-directed navigation within a familiar environment under natural conditions. Results Tracking overview and resting place Individual #1 exhibited complete fidelity to the primary resting place (RP) throughout the study, completing 101 foraging excursions. Individual #2 used the same site for most of the season but temporarily switched to alternative RPs for approximately one month before returning, resulting in 41 recorded excursions. In total, 142 excursions were documented across both individuals (Figure 1). RP locations and temporal patterns of their use by two pikeperch individuals (A and B) The spatial distribution of all detected RPs along the longitudinal axis of the Římov Reservoir. Green circles represent daily RPs used once, and red symbols indicate RPs used repeatedly. The large red square marks the main RP used for movement analyses. Red numbers indicate the frequency of use for each repeatedly used RP during the study period. (C and D) Temporal patterns of RP use from June (Jun) to September (Sep). Each point represents a daily RP position, with red symbols indicating repeatedly used RPs. For all analyses, we included only excursions associated with the main RP. Following quality control, outbound and return segments were filtered independently based on acoustic coverage and track continuity. Segments with substantial gaps that prevented reliable path reconstruction were excluded. This resulted in 84 outbound and 89 return trajectories for individual #1, and 30 outbound and 37 return trajectories for individual #2. Findings were further corroborated by data from the remaining three individuals (#3–5). Despite providing fewer complete excursions, their movement patterns were consistently similar to the primary individuals, exhibiting comparable site fidelity and homing precision (see supplementary material S1). Outbound vs. return movement characteristics Foraging activity followed a clear diel pattern (Figure 2). Most departures occurred shortly before sunset, while 30%–40% of departures initiated within the first half an hour after sunset. Foraging typically continued throughout the night, most returns started before sunrise and finished around sunrise. Both individuals exhibited this pattern, although individual #2 occasionally departed earlier and returned over a broader time window. This consistent timing ensured that both outbound and return movements occurred predominantly under low-light or dark conditions. Timing and temporal distribution of key behavioral events (A and B) Key behavioral events for individuals #1 and #2, respectively, plotted against time of day (0–24 h, x axis) across successive days of the study period (y axis). Blue dots represent departures from the RP, green dots mark the onset of foraging, orange dots indicate the end of foraging (start of return), and red dots mark arrivals back at the RP. Gray shading indicates the nighttime period between sunset and sunrise. (C) The temporal distribution of events occurring within each behavioral phase in relation to sunset/sunrise. Return trajectories were significantly longer and of greater duration than outbound ones in both individuals (Table 1). Returns frequently exceeded 1 km with a maximum of 4.8 km and 214 min in individual #2. Shortest waterway distances were also greater during returns, reflecting that individuals often finished foraging farther from the RP and thus had to cover longer direct distances back (Table 1). This asymmetry results from the way movement phases were defined. Outbound segments were defined as directed movements from the RP to the first onset of foraging behavior, whereas return trajectories represent the final homing phase from the last foraging location back to the RP. Because individuals often continued foraging and moving away from the RP after the outbound phase ended, return trajectories frequently began at greater distances from the RP and were therefore longer and of greater duration. Table 1Summary of movement metrics for outbound and return segments in the two pikeperch individualsFishPhaseN eventsTrajectory length (m)Shortest waterway distance (m)Duration (min)Detour ratioShoreline ratioStraightness index#1Outbound84849 ± 394560 ± 27035 ± 171.5 ± 0.30.4 ± 0.30.7 ± 0.1(71–1,727)(48–1,243)(5–82)(1.1–2.6)(0–1)(0.3–0.96)–Return891,102 ± 653∗∗701 ± 427∗41 ± 23∗1.6 ± 0.4ᴺˢ0.6 ± 0.3∗∗∗0.7 ± 0.1ᴺˢ(99–3,400)(84–2,052)(5–124)(1–3.5)(0.1–1)(0.4–0.94)#2Outbound30709 ± 373462 ± 20931 ± 151.5 ± 0.40.3 ± 0.20.7 ± 0.2(16–1,496)(153–1,083)(7–64)(1.1–2.8)(0–0.7)(0.3–0.96)–Return371,233 ± 882∗∗∗797 ± 524∗∗53 ± 39∗∗1.6 ± 0.5ᴺˢ0.5 ± 0.2ᴺˢ0.7 ± 0.2ᴺˢ(172–4,766)(150–2,270)(7–214)(1–3.2)(0–1)(0.4–0.95)Metrics include total trajectory length, shortest waterway distance (direct water) distance, excursion duration, detour ratio (trajectory length/shortest waterway distance), shoreline ratio (proportion of positions ≤50 m from the shoreline), and straightness index (SI). Values represent mean ± SD (standard deviation), with ranges in parentheses. Asterisks indicate significant differences between outbound and return phases based on two-sample t tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). Summary of movement metrics for outbound and return segments in the two pikeperch individuals Metrics include total trajectory length, shortest waterway distance (direct water) distance, excursion duration, detour ratio (trajectory length/shortest waterway distance), shoreline ratio (proportion of positions ≤50 m from the shoreline), and straightness index (SI). Values represent mean ± SD (standard deviation), with ranges in parentheses. Asterisks indicate significant differences between outbound and return phases based on two-sample t tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). Trajectory length and excursion duration were strongly correlated (linear model, LM; R2 = 0.90–0.97; supplementary material S1, Figure S3), indicating nearly constant swimming speeds across excursions. Estimated mean swimming speeds were 0.75 ± 0.03 (estimate ±SE) and 0.92 ± 0.03 body lengths per second (BL s−1) for outbound and return movements of individual #1, and 0.80 ± 0.04 and 0.77 ± 0.02 BL s−1 for individual #2. Detour ratio (∼1.5) and straightness index (∼0.7) did not differ significantly between phases. Shoreline use increased during returns and was significant in #1 (two-sample t tests, p < 0.001) but not in #2. Depth profiles (supplementary material S1, Figure S4) showed that both outbound and return movements occurred close to the surface (approximately 1 to 5–7 m). The offshore part of returns occurred far above the bottom and did not spatially coincide with bottom contours. Route similarity and corridor use Correlations of grid-cell densities (50 × 50 m) between outbound and return phases were moderate to high (Spearman’s ρ = 0.61 for individual #1 and 0.52 for #2; both p < 0.001), indicating partial reuse of movement corridors—particularly in individual #1 (Figure 3). To visualize these patterns within the actual three-dimensional topography of the reservoir, we provide an interactive 3D bathymetric model with selected return trajectories (Data S2). Spatial patterns of movement during outbound and return phases (A) Location density maps (50 × 50 m grid) show spatial use during outbound and return phases for individuals #1 (top) and #2 (bottom). For each individual, central figures show whole-reservoir location density distributions for outbound and return movements, while the left and right provide expanded views of the outbound and return corridors around the main RP (white circle), respectively. Warmer colors indicate higher positional density. (B) Difference in location density maps (Δdensity = return − outbound) around the main RP highlight areas used more during returns (red) or departures (blue). Route similarity was even stronger near the RP (ρ = 0.71 for #1 and 0.54 for #2; both p < 0.001), confirming repeated use of a similar approach corridor, especially by individual #1. Both individuals consistently departed along a south-eastern corridor from the RP, whereas return trips more frequently followed a southern offshore path, combined with a western nearshore approach shortly before arrival (Figure 3). Bearings during the initial departure (first 100 m) and final arrival (last 100 m before reaching the RP) were both strongly non-random (MANOVA, both p < 0.001), aligning with established movement corridors (Figure 4). Departure bearings did not differ significantly between individuals (Pillai’s trace = 0.04, F2,110 = 2.36, p = 0.099), indicating similar departure headings between individuals. Orientation during the initial and final segments of trajectories relative to the RP (A) Bearings of the initial 100 m (outbound) and (B) final 100 m (return) trajectory segments for individual #1 (left) and individual #2 (right). Dots indicate bearings of individual excursions. Black arrows show mean bearing direction, and blue sectors represent the 95% confidence interval. Rose plots (15° bins) summarize the distribution of bearings across events. Outbound and return segments were filtered independently based on acoustic coverage and track continuity; consequently, sample sizes (n) differ slightly between outbound and return phases. In contrast, arrival bearings differed significantly between individuals (Pillai’s trace = 0.08, F2,121 = 5.21, p = 0.007), suggesting individual-specific approach routes to the RP. A hierarchical model comparison confirmed that navigation phase (departure vs. arrival) strongly influenced orientation (Pillai’s trace = 0.63, F2,233 = 201.5, p < 0.001), demonstrating that arrival trajectories were not simple reversals of departure headings. A significant interaction between phase and individual identity (F2,232 = 5.93, p = 0.003) further indicated that the relationship between departure and arrival orientation differed between individuals. Inshore-offshore context of outbound vs. return movement near the resting place Departures from the RP were characterized by a diverse range of movement types; while over half of the excursions involved direct transitions into open water (combined “offshore to shore” and “only offshore” categories; Figure 5), a substantial proportion began with a nearshore phase (“shore to offshore” or “only shore” categories; Figure 5). In contrast, return segments showed a more consistent pattern: the vast majority were dominated by an “offshore to shore” transition, as individuals typically entered a nearshore approach corridor before reaching the RP. Only a smaller fraction of returns was completed directly from offshore (“shore to offshore” or “only offshore” categories; Figure 5). These cases occurred when the final foraging bout ended in open water, or after a nearshore transit, followed by a direct final approach to the RP from the offshore side. Proportion of different movement types observed during the initial (outbound) and terminal (return) phases of foraging events for individuals #1 and #2. Movement types are based on transitions between nearshore (≤50 m) and offshore (>50 m) zones, shore to offshore, offshore to shore, shore only, and offshore only. Percentages are calculated from the total number of segments in each phase. Orientation and accuracy during return movement Orientation at the onset of return trajectories was strongly non-random, with a high degree of clustering toward the direct direction to RP (MANOVA, Pillai’s trace = 0.85, F2,124 = 314.6, p < 0.001; Figure 6). Mean angular deviations were 12.6° (95% CI, 4.6°, 19.9°) for Fish #1 and 8.3° (−3.7°, 19.9°) for Fish #2. No significant difference in orientation was detected between individuals at the start of return (Pillai’s trace = 0.009, F2,123 = 0.50, p = 0.60), indicating similar angular deviations from the direct direction to RP between individuals at the start of return. Orientation accuracy during the return phase relative to the RP (A) Bearings measured at the beginning of the return (first 100 m after leaving the last foraging block). (B) Bearings during the final offshore transition before entering the nearshore zone. Angles are expressed as deviations from the direct bearing to the RP (0° = direct, positive = right, negative = left, 180° = opposite). Dots indicate bearings of individual trajectories. Black arrows show mean direction with length proportional to mean vector length, and blue sectors represent the 95% confidence interval. Sample sizes (n) differ between figures because bearings in (B) could only be calculated for returns that included a distinct offshore segment. Returns that occurred entirely along the shoreline lacked this offshore transition and therefore could not be included in this analysis, resulting in lower sample sizes. Orientation at the start of the final offshore segment (positions and details of the final offshore segment are in the supplementary material S1, Figure S5) was likewise strongly non-random and directed toward the RP (Pillai’s trace = 0.88, F2,110 = 403.6, p < 0.001; Figure 6). Mean deviations were −3.1° (−10.7°, 4.7°) for Fish #1 and 8.1° (−4.0°, 19.9°) for Fish #2. No significant individual differences were detected at the start of this offshore phase (MANOVA, Pillai’s trace = 0.03, F2,110 = 1.63, p = 0.2). Across phases, orientation remained tightly clustered toward the RP with no detectable effect of individual identity or navigation phase, demonstrating consistent directional alignment toward the RP throughout return trajectories. Final approach accuracy Miss-point distances defined as the distance between the last offshore location and the RP were generally small (Figure 7). The median errors were 87–115 m for individual #1 and 77 and 90 m for individual #2, with the majority of return trajectories ending within 250 m of the RP. Return accuracy did not differ significantly between trips ending before or after sunrise. These results indicate that individuals oriented toward the RP early in the return phase and reached the refuge with high spatial accuracy following the offshore segment. Final approach accuracy of returns relative to the RP (A) Distribution of miss-point distances—the straight-line distance between the last offshore position and the RP—for returns completed before and after sunrise. Diamonds indicate medians. (B) Spatial positions of the final offshore points colored by their distance from the RP. Integrated example of route flexibility and directed homing To integrate the findings on route use, orientation, and homing precision, we examined a six-day sequence of excursions by individual #1 (Figure 8A). During this period, the fish consistently exploited a stable foraging site located approximately 2 km from the RP, representing a temporary phase of spatial consistency within a broader pattern of multiple foraging areas used across the season. Despite repeated use of the same destination, both outbound and return trajectories varied markedly among days, demonstrating flexible route choice rather than stereotyped path following. Detailed inspection of representative return event (day 5, Figure 8B) showed that the return occurred largely offshore and did not involve systematic tracking of bottom topography. Movement was approximately parallel to the shoreline only during the initial phase of the return, followed by occasional approaches toward the shore. The return was characterized by a relatively stable cruising speed, with deceleration occurring only during the final approach to the RP. Overall, the trajectory followed a relatively direct course toward the RP, without extended directional reversals or persistent spatial alignment with shoreline or bottom contours. Repeated nocturnal foraging excursions and return movements of individual #1 (A) Six consecutive nocturnal excursions recorded between 30 June and 7 July. Excursion duration ranged from 7.8 to 9.5 h, and total distance traveled during individual excursions ranged from 8.3 to 11.2 km. During this period, the individual repeatedly used the same principal foraging area located approximately 2 km (straight-line water distance) from the RP (white point). (B) Movement descriptors for a representative return event (Day 5). Vertical position relative to bottom depth, distance to shoreline, alignment of movement direction relative to the nearest shoreline (0° = parallel, 90° = perpendicular), and swimming speed (BL s−1) are shown. The dashed purple line marks sunrise. Discussion Pikeperch exhibited highly consistent navigational behavior during diel commuting between a fixed daytime refuge and distant nocturnal foraging grounds. Both tracked individuals repeatedly left the same RP soon after sunset, foraged widely through the night, and returned around sunrise. Despite traveling several kilometers, their movements were highly structured and efficient: outbound and return trajectories followed consistent bearings, maintained steady swimming speeds, and showed low tortuosity. Return trajectories were typically longer, reflecting greater distances between final foraging locations and the refuge, yet individuals oriented rapidly and maintained a directed homeward course over kilometer scales. Together, these patterns indicate that pikeperch can reliably return to a fixed refuge within a structurally simple and turbid reservoir under low-light conditions. This study provides field-based evidence of consistent homing to a fixed refuge under conditions where continuous visual landmark guidance is limited. Previous research has demonstrated navigational abilities in fishes primarily in laboratory maze experiments,3,14,34 visually structured reef systems,29,30,31,35 and in large-scale migratory species such as salmonids, tunas, eels or coral reef fish larvae.22,23,24,36 However, these contexts differ fundamentally in spatial scale, sensory environment, and ecological function from the fine-scale, repeated homing documented here. Our results show consistent and repeated returns to a spatially fixed refuge within a turbid freshwater system, demonstrating that accurate goal-directed homing to a specific location can be maintained as part of routine daily movements between key sites in natural environments. Our data do not allow us to determine the precise navigational mechanism underlying homing. However, several simple explanations, when considered individually, appear insufficient to fully account for the observed patterns. Pikeperch physiological adaptations such as the tapetum lucidum may facilitate orientation under turbid or low-light conditions.37 However, the kilometer-scale homing precision observed here is difficult to attribute solely to visual mechanisms, particularly where stable landmark guidance is sparse and return trajectories frequently cross open water. Under such conditions, simple beaconing or local piloting based on proximal visual cues is also unlikely to explain the consistency of returns.3 Simple shoreline-following cannot explain many return trajectories, which frequently occur offshore and across open water. Likewise, the irregular and highly variable structure of movement trajectories further limits the plausibility of pure path integration across repeated multi-kilometer movements lasting several hours, as cumulative error typically degrades positional accuracy with distance and time.38,39 Other potential guidance cues also seem unlikely to provide consistent large-scale orientation independently. Water flow in the main body of the reservoir is negligible and directionally unstable, making flow-based orientation implausible. Stable chemical or olfactory gradients are unlikely to provide reliable positional information over kilometer scales given their pronounced spatial and primarily temporal variability.40,41 Acoustic cues may be present only intermittently and locally near the dam, making them neither stable nor consistently available, and they cannot explain homing by individuals whose RPs are located far from the dam. Social guidance is also unlikely, as adult pikeperch are largely solitary,42 and prey distributions in the reservoir are highly dynamic and therefore unlikely to provide stable spatial reference points.43 Taken together, these considerations suggest that homing in this system most likely relies on the flexible integration of multiple orientation cues rather than any single simple mechanism acting alone. Our field data indicate that navigation in pikeperch likely involves the integration of multiple orientation strategies similar to those documented in laboratory studies. Experimental work has shown that teleosts can combine compass-based orientation with the use of familiar spatial reference systems when navigating to known locations.3,14,34 Although the precise mechanisms cannot be determined here, the movement patterns observed in this study are most consistent with goal-directed orientation toward a memorized refuge location, likely relying on reliable directional information and complemented by opportunistic use of structural cues when available. Individuals frequently initiated return movements with bearings closely aligned to the RP from multiple starting locations, indicating that navigation was not restricted to retracing previously traveled routes. Although parts of return trajectories followed the shoreline, fish were able to maintain directed offshore movements, demonstrating that continuous landmark tracking was not necessary for accurate homing. The ability to follow different shoreline sections on different days further indicates flexibility rather than stereotyped corridor use. Together, these patterns imply that fish may retain spatial information about the refuge location that allows them to generate an appropriate homeward bearing from multiple starting points, while structural features may serve as stabilizing or auxiliary cues rather than primary navigational guides. Although alternative mechanisms such as path integration or sequential use of familiar routes cannot be entirely excluded, they appear insufficient alone to explain the full range of observed behaviors. Our results document not only consistent nightly homing but also patterns consistent with long-term spatial memory. Both individuals repeatedly returned to the same refuge over several months despite undertaking multi-kilometer nocturnal foraging trips along variable routes. Remarkably, one individual abandoned the main refuge for more than a month and used alternative resting sites elsewhere in the reservoir before returning with high spatial accuracy to the original location. Such prolonged absences followed by accurate returns are consistent with the retention of spatial information about key areas rather than reliance on short-term route familiarity alone. Long-term homing supported by spatial memory is well documented across fishes inhabiting a wide range of environments, including coral reefs, freshwater systems, and large marine pelagic ecosystems.3,22,29,30,31,44,45,46,47 Our observations add a field-based example of persistent, fine-scale homing in a resident freshwater predator operating under low-visibility conditions. Accurate returns to a stable refuge combined with flexible use of foraging areas likely provide clear ecological benefits. Memory-supported movements can reduce search and travel costs, promote efficient reuse of space, and enable predators to balance exploration of shifting resources with reliable access to shelter.20,48 In dynamic environments such as large turbid reservoirs, durable spatial representations may therefore allow individuals to exploit transient opportunities without losing track of key resting locations. Advances in high-resolution acoustic telemetry are expanding opportunities to study navigation and spatial memory in free-ranging fishes.49,50,51 Our results illustrate how dense positioning arrays can reveal fine-scale homing and persistent site fidelity under natural conditions that are difficult to capture in laboratory settings. Such systems also offer a promising framework for future work combining long-term tracking with experimental manipulations or environmental data to better link movement patterns with underlying navigational processes.18,50 Our findings provide field-based evidence that pikeperch can navigate with high spatial precision under low-visibility conditions in a natural freshwater system. The persistence of accurate homing, even following month-long absences, is consistent with the use of long-term spatial memory and flexible orientation strategies in a free-ranging teleost. Evidence of consistent homing across diverse fish taxa suggests that such spatial abilities may be widespread among fishes inhabiting both marine and freshwater environments.29,47,52 More broadly, our results illustrate how high-resolution telemetry can reveal fine-scale navigational structure and memory-informed movement in free-ranging animals and provide a foundation for future field-based tests of spatial behavior in aquatic systems. Limitations of the study While our high-resolution analysis focuses on a large number of trajectories (n = 142 homing events) from two intensively monitored individuals, the consistency and repeatability of these movements are unlikely to reflect purely idiosyncratic behavior. This is further supported by observations from three additional individuals (supplementary material S1), which exhibited qualitatively similar movement patterns. The exceptional spatiotemporal resolution of our data allows for a level of behavioral characterization that is often lost in larger-scale studies with lower sampling frequencies. However, the sex of tracked individuals was not determined, and potential sex-related differences in homing behavior could therefore not be evaluated. Future work should extend sampling across individuals and environmental contexts and combine high-resolution tracking with targeted experimental manipulations to test underlying navigational mechanisms more directly. Environmental variables such as light regime, water clarity, and bathymetry also warrant explicit investigation to better understand how spatial information is acquired and used in turbid freshwater systems. Resource availability Lead contact Further information and requests should be directed to and will be fulfilled by the lead contact, Milan Říha (milan.riha@hbu.cas.cz). Materials availability This study did not generate new unique reagents. Data and code availability Processed data reported in this article are available from Zenodo at https://doi.org/10.5281/zenodo.19885546. Analysis code reported in this article is available from Zenodo at https://doi.org/10.5281/zenodo.20331980 and from GitHub at https://github.com/rihamilan/Daily-nocturnal-homing-reveals-goal-directed-navigation-in-a-wild-teleost.git. Acknowledgments The authors would like to thank the Fishecu team for their invaluable assistance during the fieldwork and Zuzana Sajdlová for drawing the graphical abstract. This study was supported by project no. QL26020013 (“Long-term sustainable and profitable production of pikeperch juveniles in recirculating facilities leading to higher adaptability of stocked fish and improved water quality in reservoirs”) of the National Agency for Agricultural Research MZe. Author contributions Conceptualization and design, M.Ř.; methodology, M.Ř.; project administration, M.Ř.; data collection, M.Ř., P.B., M.Š., T.J., M.T., J.K., and J.P.; data and statistical analysis, M.Ř., R.R.-B., M.P., and I.J., manuscript drafting, M.Ř.; manuscript review and editing, all authors; funding acquisition, M.Ř., T.J., and J.P. Declaration of interests The authors declare no competing interests. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work the authors used ChatGPT and Gemini in order to polish R code and improve the readability and language of the manuscript. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERExperimental models: Organisms/strainsPikeperch (Sander lucioperca)Římov Reservoir, Czech RepublicN/ASoftware and algorithmsUMAP acoustic positioning software (v.1.4.3)Lotek Wireless Inc., Canadahttps://www.lotek.com/R (version 4.4.2)R Foundation for Statistical Computinghttps://www.r-project.orgRcppRoll packageCRANhttps://cran.r-project.org/package=RcppRolldbscan packageCRANhttps://cran.r-project.org/package=dbscangdistance packageCRANhttps://cran.r-project.org/package=gdistanceraster packageCRANhttps://cran.r-project.org/package=rastersf packageCRANhttps://cran.r-project.org/package=sfsuncalc packageCRANhttps://cran.r-project.org/package=suncalcrKin packageCRANhttps://cran.r-project.org/package=rKINgeosphere packageCRANhttps://cran.r-project.org/package=geospherecircular packageCRANhttps://cran.r-project.org/package=circularDirectional packageCRANhttps://cran.r-project.org/package=DirectionalMVN packageCRANhttps://cran.r-project.org/package=MVNOtherAcoustic transmittersLotek Wireless Inc., CanadaModel: MM-M-11-28-TPAcoustic hydrophonesLotek Wireless Inc., CanadaModel: WHS3250Data resourcesProcessed acoustic telemetry datasetThis studyhttps://doi.org/10.5281/zenodo.19885546Analysis codeThis studyhttps://doi.org/10.5281/zenodo.20331980; GitHub: https://github.com/rihamilan/Daily-nocturnal-homing-reveals-goal-directed-navigation-in-a-wild-teleost.git Experimental model and study participant details Study species and ethics statement This study utilized acoustic telemetry data from a broader multi-species tracking project described in detail by.53 Twenty adult pikeperch were captured in the Římov Reservoir between 18 and 25 April 2017 using boat electrofishing, anesthetized with 2-phenoxyethanol, and surgically implanted with depth- and temperature-sensitive acoustic transmitters (Lotek Wireless MM-M-11-28-TP; 65 × 12 mm; 13 g in air; 15 s burst rate). Following recovery, individuals were released at their respective capture sites. Pikeperch sex was not determined at tagging because external sex determination is unreliable outside the spawning period and no lethal sampling was performed. Consequently, potential sex-related differences in movement and homing behavior could not be assessed and are acknowledged as a limitation. Animal treatment was performed in accordance with guidelines from the Experimental Animal Welfare Commission under the Ministry of Agriculture of the Czech Republic and with permission (Ref. no. 310/7387) from the study sites managers, Povodí Vltavy s.p. Method details Study site The study was conducted in the Římov Reservoir, a deep, narrow, canyon-shaped freshwater system located in South Bohemia, Czech Republic (48°51.00978′ N, 14°29.47462′ E; supplementary material S1, Figure S1). Constructed in 1978 on the Malše River for drinking water supply and flood protection, the reservoir covers 210 ha, holds a total volume of 33.1 × 106 m3, and reaches maximum and average depths of 45 m and 16 m, respectively. The main basin extends approximately 8.5 km along its longitudinal axis and features a simple V-shaped cross-section with steep banks and narrow littoral zones.43 These geomorphological characteristics, combined with seasonal water level fluctuations, completely inhibit the growth of submerged macrophytes. The steep slopes and absence of littoral vegetation result in structurally simple shoreline zones with relatively few stable visual features that could serve as local reference points for aquatic organisms. The reservoir is dimictic, with summer stratification typically occurring from April to September.53 Its trophic status ranges from mesotrophic to eutrophic, with phosphorus and chlorophyll-a concentrations decreasing from the tributary toward the dam.40 The transparency of water fluctuates usually between 1.5 and 2.2 m during summer.54 Water flow within the main basin is minimal and directionally unstable, with current velocities typically in the order of millimetres per second, reflecting the lentic character of the reservoir (J. Hejzlar, pers. comm.). The reservoir is dominated by cyprinid species,55 which are suitable prey for pikeperch, and has a long-term stable pikeperch population.56 Telemetry and data acquisition Tagged fish were monitored using a MAP high-resolution acoustic positioning system (Lotek Wireless Inc., Canada), consisting of 90 hydrophones (WHS3250, frequency 76 kHz) spaced 80–300 m apart to provide near-complete coverage of the reservoir (approximately 92% of the reservoir area covered by the positioning system). Only a small bay near the dam and part of the tributary remained outside the detection range (Supplementary Material Figure S1). Horizontal positioning accuracy was assessed using 22 stationary reference tags and validated through boat tow tests (see53 for further details). For this study, analyses were restricted to the period from late June to late September 2017 to avoid potential behavioral effects related to spawning and seasonal activity shifts at the beginning of autumn.53 Horizontal positions were calculated using UMAP v.1.4.3 software (Lotek Wireless Inc., Canada), while depth was derived from the tag’s internal pressure sensor (resolution 0.7 m). Raw positional data were filtered to remove erroneous estimates using GAM-based smoothing and visual inspection of movement tracks, following the protocol of.53 Although data availability was limited by post-release mortality and tag expulsion, sufficient tracking data were obtained for five individuals. Our primary analyses focus on two individuals (#1 and #2; total body length 495 mm and 490 mm, respectively) that provided uninterrupted, high-resolution tracking throughout the entire study period, yielding a complete set of foraging-to-homing events suitable for quantitative analysis of trajectories. This approach allowed for a high density of data points per individual, with each individual serving as its own replicate across dozens of individual homing events (total excursions = 142). Such longitudinal, high-frequency sampling provides a robust basis for assessing intra-individual behavioral consistency, which is particularly informative for evaluating navigational mechanisms compared to lower-resolution data from a larger cohort. Partial tracks from the remaining three individuals were used to corroborate the generality of the observed patterns and are presented in the supplementary material S1. Analytical framework To investigate how fish navigate between their resting place (RP) and foraging areas, we characterized both the spatial context and movement efficiency of outbound and return journeys. Movements were classified as nearshore when individuals were within 50 m of the shoreline and as offshore when beyond this threshold. Movement performance was evaluated based on route choice (nearshore or offshore), path efficiency, spatial consistency across trips, and the accuracy of the final approach to the RP. In the study reservoir, nearshore zones are more likely to provide stable spatial cues, whereas offshore areas are featureless. We analyzed heading orientation, preferred movement corridors, and final approach precision during the day-night transition. Using these metrics, we assessed route fidelity, use of nearshore or offshore pathways, and the directional structure of outbound and return trajectories. Resting place identification To establish a fixed spatial reference point for analyzing departures, returns, and movement patterns, we first identified the daily RP for each individual. For each fish, we calculated the distance and swimming speed between consecutive positions and smoothed the speed using a five-point right-aligned rolling mean using the RcppRoll package57 to reduce noise. Data were grouped by day. Positions with very low velocities (<0.015 m s−1) were marked as potential resting positions. These low-speed positions were clustered using the DBSCAN algorithm58 implemented in the dbscan package,59 applying a 20 m radius and a minimum of 50 positions per cluster. Days without a valid cluster were excluded from further analyses. The centroid of the largest cluster each day was defined as the RP. The speed threshold and DBSCAN parameters were chosen to maximize consistency of behavioral classification across excursions while minimizing sensitivity to positional noise, and were validated by visual inspection of representative trajectories. Identification of foraging excursions In the second step, we isolated complete foraging excursions from resting periods to focus the analysis on movements returning to the same RP. For each individual, we calculated the Euclidean distance between each position and the daily RP centroid. Positions more than 30 m from the RP centroid were classified as outside the RP, and a foraging excursion was defined as a continuous sequence of positions beginning with the first point beyond this threshold and ending with the first return within the RP. To ensure that individuals returned to the same refuge (i.e., demonstrating site fidelity), only excursions terminating within 50 m of the previous day’s RP centroid were retained. Both distance thresholds were selected to conservatively exceed the estimated positioning error of the acoustic system (approximately 10 m;60) and to accommodate fine-scale movements within the refuge structure, thereby minimizing the risk of misclassifying positional jitter or local activity as true excursions. Additionally, we restricted the dataset to excursions lasting between 3 and 23 h. Excursion segmentation To characterize movement patterns to and from foraging sites, each excursion was segmented into distinct phases based on local movement geometry and speed. A local straightness index (SI; straight-line to real path distance ratio, 0–1, with higher values indicating more direct movement) and local velocity were calculated within a moving window of ±5 fixed points. Threshold values were empirically determined through visual validation of trajectory segments to reliably distinguish between behavioral modes. Positions with SI < 0.75 and/or velocity <0.2 m s−1, excluding those within a near-home buffer (see below), were classified as foraging (see Figure S2 for segmentation examples). All other positions were considered as transition phases. This approach captures both tortuous, area-restricted foraging and slow, directional movements potentially associated with prey pursuit.61,62 This classification method allowed for consistent behavioral segmentation across excursions of varying duration without the need for individual model parameterization. Consecutive foraging points were grouped into foraging blocks, while the rest of the excursion was designated as transition. To avoid misclassifying slow or meandering movements near the RP, a 250 m buffer zone around the RP was always classified as transition, regardless of SI or velocity. In three excursions, this buffer was reduced to 90–110 m based on visual inspection, which indicated that foraging began unusually close to the RP. These adjustments affected fewer than 5% of excursions and did not alter the overall results. An example of a fully segmented excursion is provided in Figure S2 (supplementary material S1). In the following analyses we focused exclusively on the first outbound segment (from the RP to the first foraging block) and the final return segment (from the last foraging block back to the RP) of each excursion. Quantification and statistical analysis Segment-level analyses and statistics For each segment (the initial outbound and the final return), we derived a set of metrics capturing key aspects of individual movement: (i) path length and shortest waterway distance (direct water distance) quantified how far fish traveled from or to the RP; (ii) the detour ratio (path length/shortest waterway distance) and straightness index (SI) described travel efficiency in terms of path directness; (iii) the proportion of nearshore positions (≤50 m from the shoreline) characterized the use of nearshore versus offshore habitats; (iv) mean travel speed indicated swimming performance; (v) spatial overlap between outbound and return routes quantified route similarity; (vi) bearing deviation relative to the RP measured orientation accuracy; (vii) nearshore/offshore transition type characterized whether departures and returns followed the shoreline or direct offshore paths; and (viii) final approach accuracy quantified how closely individuals reached the RP after the offshore phase (miss-point distance between the last offshore position and the RP). Together, these metrics evaluated how consistently individuals departed from and returned to the RP, as well as how directly they traveled. In addition, they quantified how accurately individuals reached the RP after the offshore phase. Each position was classified as nearshore or offshore, and short fragments (fewer than six consecutive positions) were merged with adjacent classifications to reduce artifacts caused by positional noise. Path length was calculated as the cumulative sum of linear distances between consecutive positions, while the shortest waterway distance represented the shortest possible waterway path between the segment start and end. Shortest waterway distances were computed using the gdistance package63 on a rasterized water surface derived from the reservoir polygon using the raster and sf packages.64 Departure and return times were referenced to local sunset and sunrise using the suncalc package.65 We compared these movement characteristics between outbound and return phases using two-sample t-tests for each individual and metric. To visualize the relationship between fish depth and bottom depth in the outbound and return segments, we applied kernel isopleth analysis (25%, 50%, and 95% levels) in fish depth–bottom depth coordinate space using the rKin package.66 Separate kernels were computed for nearshore (≤50 m from shoreline) and offshore positions, allowing comparison of spatial extent and overlap within each phase. Spatial patterns of movement were examined by binning fish positions into a 50 × 50 m grid and normalizing counts by segment duration to account for differences in observation effort. For each individual, we generated density maps for outbound and return phases and their difference (return – outbound) to highlight corridor use. Spatial correlation between outbound and return paths was then quantified using rank-based correlation of corresponding grid cell densities, providing a measure of route similarity independent of absolute location usage. Mean travel speed for each phase was estimated by fitting linear models (LMs) with segment trajectory length as the response variable and segment duration as the predictor; the slope of the fitted line was interpreted as average speed, and R2 was recorded as a measure of model fit. To classify the nearshore/offshore context of initial departure and final approach, we examined the first two consecutive shoreline classes in the outbound segment and the last two in the return segment. Each sequence was assigned to one of four categories: shore_to_offshore (nearshore → offshore), offshore_to_shore (offshore → nearshore), only_shore (both nearshore), and only_offshore (both offshore). Orientation relative to the resting place (RP) was evaluated at four key navigation points. Bearings were calculated (i) during the first 100 m of departure from the RP, (ii) within the final 100 m before arrival, (iii) immediately after leaving the final foraging area at the onset of return, and (iv) at the beginning of the final offshore segment preceding arrival. Bearings near the RP (departure and final approach) were analyzed as absolute geographic azimuths (0–360°). Bearings during return phases were expressed as angular deviations from the direct bearing to the RP (−180°–180°, where 0° represents the direct heading). All bearings were calculated on the WGS84 ellipsoid using the geosphere package.67 Non-random orientation and differences between navigation phases and individuals were evaluated using a circular MANOVA framework following,68 with navigation phase, individual identity, and their interaction as explanatory terms. To determine whether orientation within each phase was significantly directed rather than random, we first fitted intercept-only MANOVA models testing whether the mean resultant vector deviated from the origin. Bearings were transformed into bivariate Cartesian coordinates (sin θ, cos θ) to allow analysis within a linear multivariate framework. MANOVA assumptions were assessed using Mardia’s test for multivariate normality and Box’s M test for homogeneity of covariance matrices. Given the variability typical of behavioral data, Pillai’s Trace was used for all multivariate tests due to its robustness to deviations from normality.68 Directional precision was quantified using 95% bootstrap confidence intervals for mean direction (2,000 iterations), enabling robust comparisons without assuming a specific underlying circular distribution. All analyses were conducted in R using the circular, Directional and MVN packages.69,70 Final approach accuracy was quantified as the miss-point distance, defined as the straight-line distance between the last offshore position of each return trajectory and the RP. This metric quantified how closely individuals oriented toward the refuge during the offshore phase, when stable external reference features were limited, and how close they approached the RP before transitioning to nearshore movement. All analyses and visualizations were conducted in R (version 4.4.2; R Core Team, 2024). Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.isci.2026.116364. Supplemental information Document S1. Figures S1–S9, Table S1, and supplementary material S1 Data S2. Interactive 3D bathymetric model with selected pikeperch return trajectories, related to Figure 3 and STAR methods References Associated Data Supplementary Materials Document S1. Figures S1–S9, Table S1, and supplementary material S1 Data S2. Interactive 3D bathymetric model with selected pikeperch return trajectories, related to Figure 3 and STAR methods Data Availability Statement Processed data reported in this article are available from Zenodo at https://doi.org/10.5281/zenodo.19885546. Analysis code reported in this article is available from Zenodo at https://doi.org/10.5281/zenodo.20331980 and from GitHub at https://github.com/rihamilan/Daily-nocturnal-homing-reveals-goal-directed-navigation-in-a-wild-teleost.git.