© S. Tomanova et al., Published by EDP Sciences 2025

1 Introduction

Integrating streams and rivers into road or railway projects has long been thought of as a simple matter of satisfying hydraulic issues and flood risks. These intersections between networks were developed as hydraulic structures (e.g. culverts) guaranteeing the sustainability of roads or railways, with little concern for other issues. In particular, ecological functions provided by watercourses (migratory corridors, living environments for many species, etc.) were neglected. As a result, numerous road-watercourse intersections contribute to habitat fragmentation (Forest Service Stream-Simulation Working Group, 2008; Goerig and Castro-Santos, 2017) threatening freshwater biodiversity (Fuller et al., 2015; Brauer and Beheregaray, 2020).

Hydraulic structures, built under the transport infrastructures, can act as physical and/or behavioral barriers for aquatic species, particularly for fish. Low water levels, high current speeds and/or significant falls at the downstream outlet can constitute physically impassable barriers for many fish species or individuals (Larinier, 2002; Forest Service Stream-Simulation Working Group, 2008; Burford et al., 2009; Newbold et al., 2014; and see citations in Goerig and Castro-Santos, 2017). Strong changes in luminous intensity at the entrance (and/or exit) of a hydraulic structure can act as a behavioral barrier, particularly when the structure is long and presents a very dark central part during the day. These conditions can discourage or frighten some fish species with diurnal activity (Keep et al., 2021), and result in hesitation by fish at the entrance/exit, in long passage times or in progress failure for some individuals.

In Europe, good ecological status for rivers (Directive 2000/60/CE, 2000) requires good ecological continuity conditions and, consequently, numerous culverts and hydraulic structures impacting free fish movements need to be improved in terms of fish passage. Some hydraulic structures under transport infrastructures have already been equipped with devices facilitating fish passage. Baffles and other types of deflectors are usually used in culverts (Larinier, 2002; Gigleux and De Billy, 2013) to ensure a sufficient water depth at any time, allowing the progression of large fish, along with low velocity areas where the fish can rest before moving through the next high velocity zone (Newbold et al., 2014). However, the biological efficiency of such devices, potentially able to solve or limit the impacts of such physical barriers (but not the behavioral impact), is still poorly documented in situ (but see Baker, 2014; Caudron et al., 2020).

To fill this gap, we used PIT (Passive Integrated Transponder) telemetry to monitor the movements of brown trout (Salmo trutta fario, Linnaeus, 1758) individuals through a 73 m long hydraulic structure equipped with baffles perpendicular to the flow and presenting alternating notches. Two methodological approaches were used to evaluate the passage success of trout through this device: (1) the translocation of individuals that were captured upstream of the culvert, tagged and then released downstream of the culvert, (2) the monitoring of naturally migrating individuals that had been living downstream of the culvert. Fish translocation has been previously used in several studies to evaluate the effectiveness of fishways (Weibel and Peter, 2013; Dodd et al., 2017; Ovidio et al., 2020). This approach can efficiently increase the number of migrating individuals (a high return rate can be observed especially for trout, Caudron et al., 2020) and usually provides the results in a very short time after releasing the fish. The concomitant use of these two methodological approaches allowed us to easily compare their respective results.

2 Material and methods

2.1 Study site

The Clidane River, with its source at an altitude of 951 m, is a 22 km long small-sized river in the Dordogne catchment (France). It is inhabited by brown trout (Salmo trutta fario), bullhead (Cottus sp.) and minnow (Phoxinus sp.) in its downstream part, and by brown trout alone in its upstream part. The Clidane River runs under the D2068 road (dual carriageway, 45°39'1.86”N, 2°38'37.54”E) and a small country road through a culvert 73.60 m long, approx. 5 m high and 4.80 m wide (Fig. 1) with a 1.6% slope. To allow fish passage, the culvert is equipped with 10 identical notched baffles (see Fig. 1), 7.8 m distant from each other, which raise the water level and reduce flow velocities by forming successive pools. This type of device concentrates the flow in the notches at low discharges, stabilizes the bottom substrate and provides resting areas for fish in the pools during their upstream movements. For example, in June 2018, at a discharge of 330 L/s, the water level ranged between 30 and 45 cm in the pools (depending on patches of sediment deposition), 26 cm on average in the notches, with 13–20 cm drops between pools, and the mean water velocity was 0.96 m/s in the notches. These conditions correspond to very low volumetric power dissipation (<50 W/m³) and are not limiting for trout passage (Larinier et al., 2006; Baudoin et al., 2015). A two-level shelf runs along the left bank to allow small wild fauna to pass through the structure.

Fig. 1 Study site on the Clidane River. The fished zones (upstream and downstream of the culvert) are delimited by the red dots. The dimensions of the baffles are given at the top. The two photos show the inside view of the culvert from downstream and its downstream entrance. The locations of the 3 PIT-detection flatbed antennas (River A, Entrance A and Exit A) are given in the right-hand box.

2.2 Fish and environmental survey

Fish movements through the culvert were studied using PIT telemetry (following the recommendations of NF EN 17233, 2021). Three flatbed half-duplex PIT detection antennas (connected to CIPAM or Oregon receivers with storage capacity) were installed at the site (Fig. 1): one on the river bottom downstream of the culvert (River antenna, detecting upstream migrating fish), one at the culvert downstream entrance (Entrance antenna, just upstream of the first baffle) and one at the culvert upstream exit (Exit antenna, detecting fish successfully passing through the entire device). Fish detection above the antenna cable was effective within a range of 23–30 cm (depending on the antenna) for 12 mm PIT tags and exceeded 80 cm for 23 mm PIT tags. During high flow events, the antennas could not detect 12 mm PITs in the whole water column. However, we considered them still efficient to detect upstream movements because it's very likely that small fish, tagged with these small PITs, used the more favorable hydraulic conditions of the near-bottom layer to move upstream.

The fish were captured on two occasions (May 2019 and May 2020) by electrofishing (DREAM Electronique© backpack electrofisher). Two different strategies were applied: (1) fish were captured upstream of the culvert, tagged, translocated and released 300 m downstream of the culvert, (2) fish were captured between 250 and 1000 m downstream of the culvert, on eight river sectors (of about 100 m in length, except sector #5 which was shorter, see Fig. 1), tagged and released on their capture sector. All captured individuals in good general condition were anaesthetized (3–5 min in a bath with 1:10 clove oil ethanol solution, in concertation of 3–4 ml/10 L of water), measured (total length and weight) and tagged with a PIT tag inserted into the body cavity (12 mm PITs were used for fish up to 130 mm in total length, 23 mm PITs for larger ones). The surgical procedure lasted a few seconds. Tagged fish were then kept in a pool with oxygenated water, where they recovered and swam normally after a few minutes. All the experimental procedures of this study were validated by the Ethics Committee No.073 and received the authorization of the French Ministry for Research.

The fish movements on the study site were recorded from May 2019 to June 2021. During the fish survey, the water levels were recorded every 15 min upstream of the culvert by an ultrasonic probe (Ijinus ©) (Fig. 2). The water temperature was also measured but unfortunately, due to the malfunction of the probe, the data are not available. During the study, several breakdowns impacted the fish monitoring, with full PIT-detection antennas functioning 77% of the time (Fig. 2). Consequently, complementary mobile fish detections (with an Oregon © RFID Backpack Reader and a GPS device to locate detections) were carried out on the upstream section of the culvert to check if some individuals had gone through the structure without being detected. Four mobile surveys were carried out: on 06/25/2019 and 06/17/2020 (following an antenna malfunction that occurred after the fish tagging), on 01/23/2020, and at the end of the study on 04/27/2021. This improved the quality of data by detecting two additional individuals that had crossed the culvert without being recorded on the Exit antenna.

Fig. 2 Water level variation upstream of the culvert and periods with all PIT-detection antennas operating (red line) during the study.

2.3 Data analysis

All results were considered separately for translocated and naturally migrating fish. The migration rate was assessed as the proportion of tagged fish detected in upstream movement, i.e. on the River antenna (defined as “available fish” following NF EN 17233, 2021). The potential difference in total length between migrating and non-migrating (undetected) individuals during the study were tested with a Wilcoxon test. Note that, when the fish length was used in an analysis, we used only records from the first year after fish tagging in order to limit the bias due to fish growth during the study.

Fish passage was considered successful if the individual reached the upstream-most antenna (Exit antenna) before the end of the study. Three efficiency metrics, defined in the European guideline (NF EN 17233, 2021), were calculated:

• Attraction efficiency (A_eff): the ratio between the number of fish entering the culvert (Entrance antenna) compared to the number of migrating fish (detected on the River antenna) (N.B.: attraction efficiency in the present work is in fact a combination of the attraction and entrance efficiencies defined in the European guideline);

• Passage efficiency (P_eff): the ratio between the number of fish leaving the culvert (Exit antenna) and those that entered it (Entrance antenna); and

• Overall efficiency (overall P_eff): the ratio between the number of fish leaving the culvert (Exit antenna) and the number of migrating fish (River antenna).

For each detected individual, three passage times were computed:

• Approach time (T_app): the duration between the first detection at the River antenna and the first detection at the Entrance antenna;

• Passage time (T_pas): the duration between the first detection at the Entrance antenna and the first detection at the Exit antenna;

• Overall passage time (T_all): the duration between the first detection at the River antenna and the first detection at the Exit antenna.

The potential effects of fish size and water level (hourly averaged) on passage success (or failure) were assessed by non-parametric tests (Wilcoxon or Kruskal-Wallis test). Their respective influences on the different passage times (T_app, T_pas and T_all) were examined using a Pearson correlation. The individual total length considered in this analysis was the value measured at the time of tagging. To avoid any bias due to fish growing, individuals that crossed the structure the second year after their tagging were excluded from this analysis. For each individual, the time of day was assigned to each first detection at the River, Entrance and Exit antenna according to the geographical coordinates of the site and the date and time of the recording using the get_DayTimeLimits function of the RchivalTag R package. Four periods were considered: dawn (between nautical dawn and sunrise), day (between sunrise and sunset), dusk (between sunset and nautical dusk) and night (between nautical dusk and dawn). The distributions of passages during these periods at the different antennas (River, Entrance, Exit) were examined and tested with a Chi2 test. We also used a Kruskal-Wallis test to see if there was a difference in approach time or in passage time with respect to the period of day during which the fish moved (when the fish was first detected at the River or Entrance antenna respectively). All statistical analyses and graphical outputs were carried out using the statistical program R (R Core Team, 2019).

3 Results

3.1 Migration rate and periods

A total of 358 brown trout were tagged during the study (Tab. 1): 137 individuals (38%) captured in the downstream section and 221 (62%) in the upstream one (before being translocated and released downstream). Tagged fish measured between 82 mm and 255 mm in total length, and were smaller in the downstream section (mean body length of 141 mm for downstream individuals against 152 mm for upstream ones). Migrating fish (i.e. detected on the River antenna) did not differ in size compared to undetected individuals (t test: t = –0.65, p = 0.52 for translocated individuals, and t = 0.36, p = 0.72 for individuals in natural migration).

89.6 % of translocated individuals, between 97 and 255 mm of total length, were detected at the downstream end of the culvert (i.e. by the River antenna), usually a very short time after their release (Fig. 3a): half of them less than 2 days later, 75% of them less than 34 days. 42.3% of tagged individuals in the downstream section, between 82 and 224 mm of total length, were detected in upstream migration over the study period. Their movements were mainly observed in autumn (between September and December, Fig. 3b), which corresponds to the reproduction period of the species.

Fig. 3 Timing of the first fish presentation downstream of the culvert after tagging, for a) translocated fish and b) fish in natural migration, with indication of fishing/tagging days.

3.2 Passage efficiency and time

The two methodological approaches provided fairly similar efficiency results (Tab. 2): A_eff was 94.4% and 93.1% for translocated and naturally migrating individuals respectively, P_eff was 94.7% and 83.3% respectively, resulting in overall P_eff of 89.4%, and 77.6%. Fish length and water level did not affect the overall efficiency (in terms of success vs. failure) for translocated individuals (Wilcoxon tests, p > 0.05). For individuals in natural migration, we observed an effect of the fish size (Kruskal-Wallis test to manage ex-aequo cases, χ² = 5.23, p = 0.02), with larger individuals being more likely to abandon their upstream progression. Water level had no effect on the passage success of individuals in natural migration (Kruskal–Wallis test, p > 0.05).

The assessment of the passage times differed between the two methodological approaches. The T_all required to enter and go through the culvert (Tab. 2) were shorter for individuals in natural migration. 25% of these individuals passed through the structure in less than 40 minutes, 75% in less than 42 hours. Although T_pas did not differ significantly between translocation versus natural migration (median T_pas = 7 h 02 min vs. 5 h 38 min respectively, Wilcoxon test, p > 0.05), a significant difference was found in T_app (median T_{app =} 1 h 50 min for translocated individuals vs. 11 min for naturally migrating ones, Wilcoxon test, W = 4463, p < 0.001). Translocated fish usually spent more time downstream of the culvert before entering it, which had a direct consequence on T_all which was longer for translocated individuals (median T_all = 24 h 40 min for these vs. 11 h 16 min for naturally migrating fish, Wilcoxon test, W = 3583, p = 0.01).

Considering translocated fish only, large individuals generally passed through the culvert more quickly than small ones (Pearson correlation coefficient r = –0.27, p < 0.001). The water level (min–max = 156–507 mm, recorded at the time of the first individual presentation on the River antenna) influenced the passage time: T_app decreased with increasing water level (r = –0.34, p < 0.001) resulting in a decrease of T_all (r = –0.26, p < 0.001). These effects were not significant for individuals in natural migration (p > 0.05, but note a lower number of observations).

The first approach downstream of the culvert (first fish detection at the River antenna) was recorded during the day for 75.4% of translocated individuals and for 78.3% of individuals in natural migration (Fig. 4). 13.4% and 17.4% of first approaches, respectively, were recorded at night, 11.2% and 4.3% at dusk or dawn (note that the “dawn” and “dusk” periods were grouped together due to the low number of observations). The first fish detections on the other antennas were distributed in a fairly similar way without any significant difference (Chi2 tests on the distribution of detections between River vs. Entrance, Entrance vs. Exit and River vs. Exit, all with p > 0.05). We only noted that fish in natural migration, in comparison with translocated ones, were rarely recorded at dusk and dawn. The approach time into the culvert T_app did not differ following the period of the first fish presentation on the River antenna (dawn, dusk, day, night, Kruskal-Wallis tests, with p > 0.05). Likewise, the period at which the fish entered the culvert did not influence the passage time T_pas through the culvert (all Kruskal–Wallis tests, with p > 0.05).

Fig. 4 Proportions of the first individual detections (%) at the different antennas, according to the period of day, for translocated fish (left) and fish in natural migration (right).

4 Discussion

4.1 Functionality of the fish crossing device

Hydraulic structures can act as physical barriers, creating sections with high water speeds and/or shallow water depths, impeding fish progression. Gregory et al. (2004) had already shown the benefit of the presence of baffles in a hydraulic structure, placed perpendicularly, at 30 or 45° relative to the flow direction, on the progression of salmonids, but without being able to compute the exact gain in fish passage efficiency. Studies by Newbold et al. (2014) in an outdoor flume showed that baffle installation in a culvert can considerably improve overall passage efficiency for eels (an increase from 28% to 84% was observed during their experiments). Our field study, based on the monitoring of 198 translocated trout and of 58 individuals in natural migration, of various body lengths, completed previous works by assessing the passage efficiency through a long hydraulic structure (73 m) equipped with perpendicular baffles. The observed A_eff (94.4% after translocation and 93.1% for natural migrating behavior), P_eff (94.7% and 83.3%, respectively) and overall P_eff (89.4% and 77.6% respectively) reached high values, whichever the methodological approach used (Tab. 2). The most rapid passage times were only a few minutes, but overall, individuals in natural migration crossed the structure more quickly (median time 11 h) than translocated fish (median time 24 h). These results clearly confirm the advantage of this type of equipment for making hydraulic structures passable for trout.

Body length seemed to influence the passage success of individual fish in natural migration but in the opposite way to what was observed in previous works on other fishway solutions (Bravo-Córdoba et al., 2022; Tomanova et al., 2025). Smaller individuals passed through the culvert better than larger ones. Passage success was therefore not directly linked to the swimming abilities of individuals, known to be weaker for small fish (Ojanguren and Braña, 2003; Tudorache et al., 2008). On the other hand, water depth in the culvert did not seem to be limiting for larger individuals under the studied hydraulic conditions, since no influence of water level on passage success/failure was detected. The only hypothesis to explain this result is a lower motivation of large fish to migrate (Goerig and Castro-Santos, 2017). Knowing that favorable spawning areas were also present downstream of the culvert, it is possible that the larger individuals approached the structure for another reason (e.g. a movement within the home range). There is also a possibility of predation of the larger individuals before passing through the culvert by otters (Lutra lutra), a species detected within the study area, with dietary preferences for fish ≥120–250 mm (Sittenthaler et al., 2019).

Hydraulic structures can also act as behavioral barriers for fish exhibiting diurnal activity, due to potential strong changes in luminosity between the external environment and the entrance or the exit of the structure (Keep et al., 2021). In our study, as in Murphy (2020) on Brook trout, the majority of the studied individuals approached the culvert during the day (Fig. 4) but did not hesitate to enter or exit the culvert. The attraction efficiency A_eff of the structure was very high (>90%, Tab. 1) and the majority of trout approached, entered and exited the culvert during the day (Fig. 4). Fish approaching during the day did not show a longer approach or passage times than those approaching at night or at dawn/dusk. However, the rich riparian vegetation could have been helpful in our study, making the luminosity changes fairly gradual, which could have reduced or prevented any hesitation in the fish to enter or leave the culvert. The majority of crossing events observed during the day confirmed that migrating trout did not favor moments of total darkness to pass through the culvert, but seemed to prefer keeping visual cues to move along the structure (Fjeldstad et al., 2018).

4.2 Efficiency assessment of passage devices by translocation vs. natural migration

Two methodological approaches were applied in our study: (1) monitoring individuals translocated from upstream to downstream of the structure and (2) monitoring the natural migration of individuals living downstream, allowing a direct comparison of results provided by these two methods.

Fish translocation was previously used in several studies to evaluate the effectiveness of fishways (Weibel and Peter, 2013; Dodd et al., 2017; Ovidio et al., 2020). This approach generally ensures a sufficient number of tested individuals (a high return rate is usually observed, which was confirmed by the 90% return rate in our study) and provides the results in a very short time after the fish release (in the present study, 75% of the translocated individuals migrated within less than 34 days after the release). However, the major disadvantage of this method is that the environmental (e.g. hydraulic) conditions during these “forced” movements may not be representative of conditions during the natural migration period, potentially producing biased results. In the present study, we did not observe major differences between passage efficiency values using translocated individuals and values for naturally migrating ones. Nevertheless, the tested hydraulic conditions (summarized here by the water level) were comparable between these two experimental groups (min–max water level: 156–507 mm vs. 135–440 mm respectively, Wilcoxon tests, p > 0.05).

Significant differences however appeared on the calculations of the transit times (Tab. 2). T_app values (which reflect the attractiveness of the structure and possible fish hesitation at the entrance) were higher for translocated individuals (75th percentile = 24 h) than for naturally migrating ones (75th percentile = approx. 25 min). These results obviously impacted the T_all values of both groups. This fish hesitation at the entrance of the culvert could be linked to a post-fishing/tagging/translocation disorientation, and the resulting longer approach time could wrongly point to an attractiveness failure or a problem at the entrance of the culvert. Assessments of fish passage efficiencies based solely on the translocation of individuals should therefore consider this possible bias.

4.3 Conclusion and perspectives

The studied series of baffles in a long culvert was quite well designed to facilitate trout upstream migration. Such feedback must now be gathered from other types of fish passage devices installed inside diverse hydraulic structures, in order to provide good dimensioning criteria for these devices, adapted to the local context (e.g. discharge, species of interest). It would also be very useful to extend these studies to include other potamodromous species, in particular species with lower swimming capacities than trout, a species which presents very good swimming and jumping capabilities.

Acknowledgments

The authors thank Mr. Battut, Departmental Council of Puy-de-Dôme, and the Auvergne Rhône-Alpes Regional Division of the OFB for their kind welcome and help in the field. The authors also thank to John Kerr for proofreading the article, and to two independent referees for their valuable comments, which improved the original version of the manuscript.

References

All Tables

All Figures

	Fig. 1 Study site on the Clidane River. The fished zones (upstream and downstream of the culvert) are delimited by the red dots. The dimensions of the baffles are given at the top. The two photos show the inside view of the culvert from downstream and its downstream entrance. The locations of the 3 PIT-detection flatbed antennas (River A, Entrance A and Exit A) are given in the right-hand box.
In the text

	Fig. 2 Water level variation upstream of the culvert and periods with all PIT-detection antennas operating (red line) during the study.
In the text

	Fig. 3 Timing of the first fish presentation downstream of the culvert after tagging, for a) translocated fish and b) fish in natural migration, with indication of fishing/tagging days.
In the text

	Fig. 4 Proportions of the first individual detections (%) at the different antennas, according to the period of day, for translocated fish (left) and fish in natural migration (right).
In the text