Tracking anadromous fish over successive freshwater migrations reveals the influence

Abstract


Introduction
There are few rivers that now remain free-flowing along their entire length, particularly in developed regions (Jones et al. 2019;Belletti et al. 2020).Anthropogenic fragmentation of riverine ecosystems occurs primarily through river-regulation structures, such as dams and weirs, which are constructed for a variety of purposes, including power generation and navigation (Grill et al. 2019).A major ecological impact of river fragmentation is its disruption to diadromous fish migrations (Hall et al. 2011;Birnie-Gauvin et al. 2017), which has contributed to their population declines in recent decades (Limburg & Waldman, 2009).
These structures act as physical impediments that prevent or delay access of migrating adults to optimal spawning habitat (Lundqvist et al. 2008;Castro-Santos et al. 2017;Newton et al. 2018), and migration delays incurred at barriers can increase predation risk and have negative energetic consequences, especially when there are multiple passage attempts (Castro-Santos & Letcher, 2010;Nyqvist et al. 2017).Moreover, where rivers contain multiple barriers, the effects of sequential barriers can be cumulative (Keefer et al. 2013;Castro-Santos et al. 2017;Davies et al. 2021).
Barriers to migrating anadromous fish are often semi-permeable, with passage achieved by only a proportion of the upstream or downstream migrants and/or the migrating fish being delayed until conditions enable successful passage (Nyqvist et al. 2017;Newton et al. 2018).
As migration and thus barrier passage are time-limited processes, analyses within telemetry studies often adopt a rates-based approach that enable assessments of the impacts of timevarying and time-constant covariates on passage rates (Castro-Santos & Haro, 2003).These studies have revealed that environmental factors, such as river discharge and water temperature, significantly affect barrier passage rates (Nyqvist et al. 2017;Harbicht et al. 2018).Individual factors, such as body size, shape and condition, can also affect the barrier passage rates of individuals (Keefer et al. 2009;Nau et al. 2017;Goerig et al. 2020).
Iteroparous anadromous fishes that spawn multiple times in their natal river will potentially encounter the same barriers on multiple occasions, although the effect of these previous barrier encounters on passage is poorly understood (Nau et al. 2017).Assessments of passage by the same individuals at the same barriers in different years should thus increase our understanding of how interactions of individual and environmental factors influence passage success (Pess et al. 2014).These assessments could also indicate whether potential biases are incurred in data that are reliant on only newly tagged fish, through comparing passage rates between their year of tagging and their subsequent return (Nau et al. 2017).An example of iteroparous anadromous fish suitable for generating data on their successive annual migrations is the twaite shad Alosa fallax, which is distributed across the north-east Atlantic and Mediterranean (Aprahamian et al. 2003a).Recent declines and extirpations of their populations in European rivers have been attributed to pollution, overfishing and anthropogenic structures that act as barriers to their upstream spawning migration (de Groot, 1990;Aprahamian et al. 2003a;Antognazza et al. 2019).In their northern range, they are highly iteroparous, with previous spawners often representing over 50 % of all migrants (Aprahamian et al. 2003b).Although sensitive to handling and sedation, recent advances in surgical tagging protocols have enabled internal transmitter implantation (Bolland et al. 2019), enabling assessment of successive spawning migrations by the same individual (Davies et al. 2020).
Here, the freshwater spawning migration of twaite shad were assessed over multiple years to test how individual and environmental factors influenced anthropogenic barrier passage in the lower River Severn basin, UK.The use of long-life acoustic tags enabled individuals to be tracked in up to three successive annual spawning migrations.The primary objectives of the study were thus as follows: 1) estimate the impacts of anthropogenic barriers on twaite shad upstream migrations, including the proportion of upstream migrants passing each barrier and the migratory delay incurred by individuals during barrier passage; 2) determine the upstream extent of twaite shad spawning migrations within the basin relative to anthropogenic barriers and major tributaries, and the factors influencing the likelihood of approach to barriers; and 3) determine the individual and environmental factors influencing barrier passage rates by twaite shad, including comparisons of passage rates of newly tagged versus returning individuals, and previously successful versus unsuccessful individuals.

Study duration and area
The study assessed the upstream spawning migrations of twaite shad in the River Severn basin in 2018, 2019 and 2020, which tend to commence in April and conclude in June (Antognazza et al. 2019).The Severn is the longest river in Great Britain, rising in mid-Wales and flowing for 354 km before discharging into the Bristol Channel, and has a drainage area of 11420 km 2 (Durand et al. 2014).The study area in the lower river basin includes confluences with two major tributaries, the River Teme and River Avon, and eight major weirs (four on the main river channel, and two on each of the lower reaches of the River Teme (T1, T2) and River Avon (A1, A2)) (Figure 1, Table 1).The normal tidal limit is at Maisemore (S1a) and Llanthony weirs (S1b) on the western and eastern branches of the river, respectively (Figure 1), although large spring tides can penetrate the river up to Upper Lode Weir (S2).Between the spawning migrations of 2018 and 2019, two weirs on the River Teme (Figure 1) were modified to remediate fish passage.T1 was lowered, and a rock ramp installed to reduce the approach gradient at T2.With the exception of S2, which featured a notch and Larinier fish pass, there were no fish-passage structures on study weirs in the rivers Severn or Avon during the study period (Table 1).Passage of weirs without fish passage structures could thus only be achieved through ascent of the weir face, or during periods when weirs were inundated by high flows (all weirs) or high tides (S1a, S1b, S2).

Fish capture, tagging and release
At the start of their migration season in early-mid May 2018 and 2019, upstream-migrating adult twaite shad (referred to as 'shad' in methods and results) were captured by rod-and-line angling immediately downstream of S1a and S2.In addition, shad were captured at S2 using a trap positioned at the upstream exit of the 'notch' fish pass.Following their anaesthesia (Ethyl 3-aminobenzoate methanesulfonate: MS-222), all fish were weighed (nearest 10g), measured (fork length, nearest mm) and approximately three scales were removed for analysis of spawning history.These scales were analysed subsequently to determine their number of spawning-marks (and so their migration history) using a projecting microscope (x48 magnification) (Baglinière et al. 2001).Following the collection of their biometric data, the shad were surgically tagged with 69 kHz, V9 acoustic transmitters (www.innovasea.com),using the tagging protocol of Bolland et al. (2019), and following ethical review and according to UK Home Office project licence PD6C17B56.A total of 184 shad were tagged over the two years (Table 2, Figure S1), of which 173 were tagged with programmed long-life acoustic transmitters.At the end of June, these transmitters were programmed to switch from a randomized 60-second pulse interval (minimum interval between acoustic pulses 30 seconds, maximum interval 90 seconds) to a 600-second pulse interval until April the following year, when they were programmed to switch back to their randomized 60-second pulse interval.This programming was to increase the battery life of the transmitters to approximately three years, potentially enabling the tracking of three consecutive spawning migrations of tagged individuals.Non-programmed transmitters (11 shad) featured an identical initial pulse interval but did not switch to a 600-second interval, so tracking of these fish was possible in one migration only.
At S1a, all tagged shad were captured downstream of the weir and released upstream of the weir (Figure 1) in order to quantify approach and passage at the next weir (S2) (Table 2).At S2, the majority of tagged shad caught downstream or in the upstream trap were released upstream of the weir (n = 96) to study the extent of their onward migration and the impact of the subsequent weirs in the rivers Severn, Teme and Avon.A small proportion of tagged shad caught by rod and line (n = 10) and in the trap (n = 8) were also released downstream of S2 in 2018 in order to increase the sample size of fish used to assess passage at this weir.

Acoustic array
Prior to the commencement of each spawning migration period, an array of acoustic receivers (VR2-W and VR2-Tx, www.innovasea.com) was installed throughout the study area (Table 1; Figure 1).The furthest downstream receiver in the array (51.8347, -2.2901; Figure 1) was located in the estuary, 8 km downstream of the tidal limit, at the approximate summer limit of saltwater intrusion into the river (Bassindale, 1943).Receivers were deployed upstream and downstream of each weir and in unobstructed reaches between weirs (Table 1; Figure 1).
Although no shad were tagged in 2020 due to Covid-19 restrictions, the receiver array was installed to enable tracking of returning fish tagged in previous years.Receivers were anchored on steel fencing pins driven into the riverbed.In the River Teme, which featured sections of fast-flowing riffle, receivers were deployed in slower-flowing pools to maximise detection distance.In each tracking year, data were downloaded from receivers approximately every two weeks.Most receivers were removed after a two-week period with no further movements were detected within the array since the previous download.The most downstream receiver remained in place to account for any individuals which emigrated after receiver removal, but this did not occur.Range tests revealed that 100 % of test tag transmissions were detected a minimum of 100 m away from receivers in the River Severn, and a minimum of 50 m away from receivers in River Teme.In all cases, detection range was greater than river width at receiver deployment location.Step-by-step detection efficiency values for each receiver in the array was calculated for each study year using the R package actel.Detection efficiency for receivers in the array ranged from 52.5-100%; lower detection efficiencies were associated with receivers in narrow channels and/or high turbidity tidal areas (e.g.downstream Lower Parting annual efficiency: 52.5-93.1%;downstream S1b: 43-91.4%).Detection efficiency of receivers in non-tidal areas in the River Severn was generally high (median: 98.9%).

Summary metrics
All statistical analyses were conducted using R statistical software (version 4.0.2,R Core Team, 2020).Initially, emigration and return rates were calculated for shad released in each tracking year, as well as for returning shad in each subsequent year.Shad were classed as having emigrated from the river if their final detection location was the most downstream receiver in the array (Figure 1) and they were classed as returning if they were detected moving upstream into the array in subsequent years.
To understand the relative impacts of weirs on upstream-migrating shad, the following key approach and passage summary metrics were calculated for each weir in the study area: n available, n approached, percent approach, n passed, percent passage and passage time (Table 3).These metrics were calculated separately for each of the study years, and for newly tagged versus returning individuals.To understand the overall impact of weir on the upstream migration of tagged individuals, the following summary metrics were calculated for each individual in each year: upstream extent, total passage time and delay proportion (Table 3).
To further contextualise weir impacts on upstream movement, the upstream passage times of acoustic tagged individuals through a representative obstructed reach (downstream S2 to upstream S2) and unobstructed reach (upstream S1 to downstream S2) were calculated and compared using Kruskal-Wallis rank sum.Upstream passage times were calculated as the difference in time between the first detection on downstream and upstream receivers, and standardised by the river distance between upstream and downstream receivers in each reach (unobstructed reach; ~17 km; obstructed reach: ~1km).

Factors affecting approach of weirs
The individual factors affecting weir approach by newly tagged and returning shad were tested using binomial generalised linear mixed models (GLMMs) in the R package lme4, and generalised linear models (GLMs) in base R. Individuals that were available to approach S2 and/or S3/T1 were categorised as either approaching (1) or non-approaching (0).Two sets of models were constructed to test the effects of individual covariates on approach likelihood.
The first model set tested whether tagging status (newly tagged versus returner) affected the likelihood of weir approach, using GLMMs.These models included the approach classification (0/1) for fish that provided two years of approach data at a weir.Additional individual covariates were body length and spawning history (number of previous spawning events indicated by scale analysis).A fixed effect of weir was also included to test whether approach likelihood of individuals that were available to approach S2 differed from approach likelihood of those available to approach S3/T1.A random effect of individual fish i.d. was included in the models, to account for repeated measures from the same individuals across different years.
The second model set tested whether approach of S3 and/or T1 in the previous year affected the subsequent likelihood of approach of either weir for returning fish, using GLMs.These models included the approach classification (0/1) of returning individuals with known approach classifications in the previous year.Additional individual covariates were body length and spawning history.Approach of S2 was not included in this model, due to high approach rates by returning individuals at this weir.
Candidate model sets containing all possible combinations of covariates (body length, spawning history, river section, tagging status) without interactions, excluding pairs of covariates that were strongly tied (previous spawning and body size), were tested and ranked according to AICc.Models within 2 AICc of the top-ranked model were considered to have strong support (Burnham & Anderson 2002), unless they were a more complex version of a nested model with lower AICc (Richards, et al. 2011).We considered the risk of obtaining spurious results due to an 'all possible models' approach was low, due to the low number of covariates tested (<6); indeed, including all covariates counters the risks of confirmation bias and minimises the risk of excluding unanticipated results (Alcott et al. 2021).

Factors influencing passage rates of weirs
The factors influencing passage rates of newly tagged and returning shad were tested using time-to-event analysis (Castro-Santos & Haro, 2003;Goerig et al. 2020).This analysis measured the relative effects of individual and time-varying covariates on passage rates at S2 (Figure 1), as this weir had the largest sample size of approach and passage over the three tracking years.Shad entered the 'risk set' (the set of individuals to pass) when they were detected on the receiver immediately downstream of S2 during an upstream approach (Figure 1).Individuals remained in the risk set until their retreat downstream (confirmed by detection on receiver approximately 1 km downstream of S2 (Figure 1)) or their passage over the weir.This approach ensured that fish were only considered to be candidates to passage (and subject to covariate conditions) while they were actually present.Mixed effects Cox models of passage rate, incorporating individual and environmental fixed effects and a random effect (fish i.d.), were constructed using the package coxme in R (R Core Team, 2020;Therneau, 2020).The random effect accounted for statistical dependence among repeated passage from the same fish in different years (Therneau et al. 2003).
During data preparation, raw detection data for each shad were converted into 15-min observations of location, defined as the location of last detection, and observations of movements between receivers.Approach observations occurring at the receiver immediately downstream of S2, and passage observations (first detection upstream), were selected.These observations were then associated with individual metadata (body length, spawning history, previous success) and environmental data.Environmental covariates were downstream river level (m), water temperature ( o C) and diel period (as day/night, based on time of sunset and sunrise at weir S2, using the maptools package (Bivand & Lewin-Koh, 2019)).Individual body length (cm), spawning history (n previous spawning events, grouped into 0, 1+) were also included as covariates.Shad that passed the weir were censored from the model dataset at the time of passage, and non-passing individuals following their final upstream approach.
Following data preparation, two model datasets were created to test specific factors relating to the tagging status and previous experience of individual tagged shad on passage rates at S2. Dataset 1 enabled testing of tagging status (newly tagged versus returning shad) on passage rates, and so contained approach and passage events for acoustic-tagged shad released downstream of S2 in 2018 and 2019 that also returned to the weir following year, i.e. 2019 and 2020.Dataset 2 enabled testing of the impact of previous success at passing weir S2 during the first year at liberty (2018 and 2019) on subsequent passage rates in the return year (2019 and 2020, respectively), so contained approach and passage events for returning acoustic-tagged shad with known passage (successful or unsuccessful) during their first year at liberty.Body length was excluded as a covariate from testing on Dataset 2 due to the unknown body length of returning individuals.
To analyse these two datasets, initial data exploration assessed collinearity between covariates (Zuur, et al. 2010).Model selection was then conducted as per the GLMMs.The assumption of proportional hazards in the top-ranked Cox models was assessed by visual inspection of Schoenfeld residuals to confirm a zero slope for each covariate (Schoenfeld, 1982).Covariate effects from the top-ranked model were presented as hazard ratios (HR), which represent the effect on passage rates of increasing the value of continuous covariates by one unit (e.g. by 1 m for river level) or by changing the value of a categorical covariate.
Survival curves for categorical predictive variables, and representative levels of continuous predictive variables, were plotted using the R package survminer.

Summary of emigration and return
Of the 173 shad tagged with long-life acoustic transmitters in 2018 and 2019, 125 (72 %) emigrated from the river (Table 4).Of these emigrating fish, 71 (57 %) were subsequently detected returning to the River Severn for a second year, and of these 53 (75 %) emigrated for a second time.Emigration rates were similar between newly tagged fish and returning fish in each year, and return rates were the same (57%) for newly tagged fish that emigrated in 2018 and 2019 (Table 4).Of the 73 fish tagged in 2018, 7 (10 %) returned for a third year in 2020, all of which had also returned in 2019.

Weir approach, passage and passage time
The percentage of shad that approached and passed weirs in the River Severn basin varied spatially (between weirs), temporally (between years), and also between newly tagged and returning fish (Table 5).At S1a/b, the first weirs encountered by upstream-migrating shad, the combined percent approach and passage of returning individuals at these structures were very high (98-100 %) in 2019 and 2020 (Table 5).Of those that moved upstream of S1a/b, the percent approaching the next weir S2 was high in each tracking year, particularly for returning individuals (98-100%) relative to newly-tagged individuals (91-93%) (Table 5).
Of the movements recorded upstream of S1a/b (n individuals = 114; n upstream movements = 152), 94 % resulted in an approach of S2, with the others reached their upstream extent between 1 and 4 river km (rkm) downstream of S2 (Figure 3a).Of the upstream movements recorded upstream of S2 (n individuals = 127; n upstream movements = 164), 63 % approached S3 and/or T1, and upstream extents for non-approaching fish were concentrated around the lower River Teme and its confluence with the Severn (19 %, Figure 2b), with a further 19 % reaching an upstream extent within the 24 rkm section of the River Severn between S2 and the River Teme confluence (Figure 3b).Of the 11 migrations tracked upstream of T1 by 9 individuals, there were 3 approaches of T2, with the remaining 8 reaching upstream extents between 7 and 13 km downstream of T2 (Figure 3b).Overall, weirs formed the upstream extent for 64% of migrations tracked upstream from S1a/b, and 41% of migrations tracked upstream from S2.

Individual factors influencing approach of weirs
There were 16 GLMMs that tested the factors influencing approach of S2 and S3/T1 by all fish (Table S1).The best-fitting model retained weir as a predictor of weir approach (ΔAIC from null model = 12.5), indicating that shad available to approach S3/T1 were less likely to approach these weirs than those available to approach S2 (Table 5, Figure 4a).Body length was also retained in the model but its effect was non-significant (P = 0.15; Table 6, Figure 4b), and a simpler model containing weir as the only predictor of approach also received good support (ΔAIC from best-fitting model: 0.18).There were seven GLMs that tested the likelihood of weir approach by returning fish at S3/T1 (Table S1).The best fitting model (ΔAIC from null model = 1.3) retained the previous approach of S3/T1 as the sole predictor, with the model indicated a marginally significant positive effect of previous approach on approach likelihood (P= 0.06; Table 6, Figure 4c).There were no less complex models within 2 AIC of the best-fitting model.

Individual and environmental factors influencing passage rates of weir S2
Across the three study years, tagged shad approached and passed weir S2 between mid-April and early June, with a peak in May (Figure 5).Among approaching shad in their return year (n = 69), 66 (96%) had approached the weir the previous year.There were 32 mixed effects Cox models testing the individual and environmental factors influencing passage rates of weir S2 by newly tagged and returning fish (Dataset 1) (Table S2).The best fitting model (ΔAIC from null model = 28.5;Akaike weight = 0.15) revealed that returning fish passed S2 at a significantly higher rate than newly tagged fish (p < 0.01; hazard ratio (HR) = 6.04 (2.11-17.27)),Table 7a, Figure 6).Shad passed S2 at a significantly greater rate during higher river level conditions and at higher water temperatures, although there was no significant difference between passage rates at early and mid-season temperatures (Table 7a, Figure 7).
Diel period (higher passage during the day versus at night) and body length (positive effect of body size on passage rates) were also included in the best-fitting model, although these effects were non-significant (Table 7a).
A further 64 mixed effects Cox models tested factors influencing passage rates of weir S2 by returning fish (Dataset 2; Table S3).The best fitting model (ΔAIC from null model = 21.0;total Akaike weight = 0.17) revealed that previous passage success significantly increased passage rates for returning fish relative to previously unsuccessful fish (p = 0.04; HR = 3.58 (1.15-11.6),Table 7b, Figure 6).Diel period, river level and water temperature were also included as predictors (Table 7b, Figure 7); hazard ratios for other covariates were of the same direction as in Dataset 1, although their magnitude varied (Table 7b).Previous spawning history and body length were not included as predictors in the top-ranked models of passage rates by newly tagged or returning shad, providing no support for an effect of these passage rates of acoustic tagged individuals.There were no less complex models within 2 AIC of the best-fitting models for Datasets 1 or 2.

Discussion
Weirs in the lower Severn basin impacted the upstream migration of threatened twaite shad, and passage rates and temporal delays to migration varied among weirs.Environmental conditions influenced passage rates, where episodes of elevated river levels and temperatures were important for facilitating passage.For returning tagged fish, there was evidence for a significant positive effect of previous success on passage rates, potentially suggesting a conserved ability and/or motivation to pass barriers between years.Returning fish also passed at higher rates than newly tagged fish.

Impact of weirs on shad migration
The proportion of fish that passed each weir was variable, being generally high for the tidal weirs in the lower river basin but as low as 0% (in some study years) for weirs further upstream.These results suggest that once shad had moved into freshwater, a substantial percentage were prevented access to upstream spawning habitatThe prevention of access to spawning groundshas been heavily implicated in the decline of populations of anadromous shads in the River Severn and elsewhere (e.g.Aprahamian et al. 2003;Limburg & Waldman, 2009;Buffery, 2018).Weirs also imposed considerable migration delays on the fish (Table 5, Figure 3), with such migration delays known to have negative consequences on the reproductive success and survival of anadromous fish generally (Castro-Santos & Letcher, 2010), with delays also potentially subjecting migrants to elevated predation risk (Schmitt et al. 2017;Alcott, et al. 2020).Weirs formed the upstream limit of migration for the majority of acoustic-tagged shad (Figure 3), suggesting that weirs act to constrain the spawning distribution of shad in the Severn basin.
The results presented here emphasise the need for passage remediation work in the lower River Severn basin, supporting the work that has been continuing on the river in this respect (www.unlockingthesevern.co.uk).Facilitating shad passage at these structures could include barrier removal, or the retro-fitting of fish passes that take into account the specific knowledge base on passage requirements for alosines (Haro & Castro-Santos, 2012;Pess et al. 2014;Mulligan et al. 2019); for example, over 26,000 upstream migrating allis (Alosa alosa) and twaite shad were observed using such a fish pass on the River Mondego, Portugal, across five spawning migrations (Belo et al. 2021).Indeed, the preliminary results presented here indicated that modifying weir T1 did increase shad passage rates at this structure, increasing from 0 % pre-modification to 50-67% post-modification, albeit these involved relatively low numbers of tagged individuals (Table 5).The results here provide a vital baseline for future monitoring of passage improvement work in the basin, and emphasise the importance of river restoration projects to collate data on fish movements in the pre-and post-construction periods to enable more robust and rigorous evaluation of outcome (Roscoe & Hinch, 2010;Noonan, et al. 2012).

Factors affecting approach of weirs
Although barriers formed the upstream limit of migration for the majority of the tagged shad, a subset of individuals within each impounded section did not approach weirs, particularly in the reaches of river upstream of S2 and T1 (Figure 3).This potentially indicates the availability of apparently high-quality spawning habitat in the lower River Teme, which is characterised by shallow (0.75 -2 m), fast-flowing riffle and run habitat (Antognazza et al. 2019).Twaite shad that reached their upstream extent further downstream may have spawned in considerably deeper (> 3 m) and slower-flowing habitat, which is consistent with studies suggesting the species spawns in the upper and middle reaches of estuaries (e.g.Magath & Thiel, 2013).There was also evidence that the likelihood of barrier approach was repeatable across years, with shad that approached S3 and/or T1 in the year of tagging more likely to approach the same weir(s) upon their return.This tentatively suggests these individuals had a conserved motivation to approach and pass barriers, and/or displayed some fidelity to their areas of previous spawning, which has relevance to river reconnection efforts as it suggests that not all upstream migrants may be motivated to exploit habitat upstream of a barrier following passage remediation (Pess et al. 2014).

Individual factors affecting weir passage rates
Returning twaite shad had significantly higher passage rates at weirs than newly tagged individuals (Figure 6, Table 7), with this potentially being a negative consequence of their capture and/or tagging.A confounding factor here is that the shad will have grown between tagging and subsequent return, but there was no evidence for an effect of body size on passage rates.Likewise, analysis of spawning marks on scales enabled the effect of previous spawning experience to be tested, but there was no evidence that previous spawning experience affected passage rates.Thus, it is likely sublethal capture/tagging effects may have manifested as a reduced ability and/or motivation to pass weirs in the immediate posttagging period.Tagging effects can be a pernicious feature of telemetry studies in alosines (Frank et al. 2009;Eakin, 2017) with, for example, PIT-tagged alewife Alosa pseudohaerengus returnees having higher passage rates over weirs than newly tagged fish (Nau et al. 2017;Gahagan & Bailey, 2020).Thus, in passage studies of iteroparous anadromous species, returning fish could be the most reliable indicators of weir passage rates, but not all tagged fish will return in subsequent years, and thus higher costs may be incurred generating a reliable sample size (Raabe et al. 2019).
In this study, significantly higher passage rates were recorded in individual returning twaite shad that successfully passed a weir in the previous year when compared with previously unsuccessful fish (Figure 6, Table 7).Inherent phenotypic traits (body size, body shape) (Goerig et al. 2020) may enable certain individuals to be more successful at passing barriers, but there was little evidence for phenotypic traits being a predictor of passage success in this study.Another potential explanation relates to variation in migratory motivation linked to spatial fidelity or natal homing.A widely reported feature of shad spawning distributions in fragmented river basins is that spawning often occurs in areas immediately downstream of weirs (Acolas et al. 2006;López et al. 2007).This was also observed here and might lead to imprinting of juveniles to areas downstream of barriers, resulting in a reduced motivation to progress upstream upon their return.Further, there may also be learned spatial preferences in repeat-spawning adults, whereby they display preferences to using spawning areas that were used in previous years (Pess et al. 2014).Hatchery-reared American shad (Alosa sapidissima) have demonstrated that imprinting is likely to occur at the tributary level (Hendricks et al. 2002), although the mechanism of imprinting, and precision natal homing and spatial fidelity in alosines is generally poorly understood (Pess et al. 2014).

Environmental factors affecting weir passage rates
The successful passage of barriers, such as weirs, by fish can be influenced by swimming capacity and attempt rate, which in turn can be influenced by environmental variables, such as water temperature and discharge, as well as barrier characteristics, including head height and the presence of fish passage structures (Castro-Santos, 2004;Bunt et al. 2012).In this study, increasing water temperature positively affected passage rates at weir S2 (Figure 7, Table 7).In upstream-migrants, changes in water temperature may invoke physiological and behavioural changes linked to maturation of reproductive organs, factors which then increase its motivation to ascend and pass a barrier (Lubejko et al. 2017).Higher temperatures reduced the failure rates of alewife attempting to use fishways (Franklin et al. 2012) and increased the attempt rates but reduced swimming endurance of American shad attempting to pass velocity barriers, indicating that the relationship between abiotic factors and barrier passage will be dynamic across the alosine spawning migration (Bayse et al. 2019).Other studies have reported increased passage rates within the range of temperatures at which spawning occurs, and attributed this to increased motivation to move upstream and spawn (Raabe et al. 2019).
Increasing river levels downstream of S2 significantly increased passage rates over this weir (Figure 6, Table 7).Downstream river levels at S2 are affected by both tides and river discharge, and thus the relative effects of discharge and tide on passage are challenging to decouple.Nonetheless, the results suggest that prevailing hydraulic conditions at the weir are an important influence on passage by twaite shad.There are several mechanisms by which hydraulic conditions can influence passage of barriers.Water depth at the entrance to fish passes can increase passage rates in American shad (Mulligan et al. 2019), a finding linked to reduced flow velocities at higher water depths.Passage of alosines may also be negatively affected by noise and entrained air and turbulence, all of which may be influenced by downstream river levels (Haro & Castro-Santos, 2012).There was also some evidence that the passage rates of S2 were greater during the day than at night.Shads tend to prefer daylight hours to migrate upstream (Haro & Castro-Santos, 2012;Raabe et al. 2019), while twaite shad spawning is highly nocturnal (López et al. 2011).The lower passage rate at night may thus reflect differences in motivation between day/night approaches, with weir approaches during the day being passage attempts and nocturnal approaches being upstream movements associated with spawning (Acolas et al. 2004;López et al. 2011).In anadromous shads, spawning activity immediately downstream of barriers has been attributed to 'forced' spawning of unsuccessful individuals, as well as the presence of relatively high quality habitat immediately downstream of weirs (Acolas et al. 2004;Acolas et al. 2006;López et al. 2011).Further work is required to understand potential spatial differences in nocturnal versus diurnal approaches to weirs by shad, which will improve current understandings of characteristics such as spatial fidelity and motivation.

Future research
The research presented here was a coarse-scale assessment of the factors affecting weir approach and passage.In future, a more precise spatial and temporal understanding of weir approach and rejection rates, incorporating rates-based analyses, in relation to temperature and river level could be obtained by performing finer-scale telemetry studies immediately downstream of certain weirs, e.g.radio telemetry or high-frequency acoustic telemetry.
However, such technology would not be compatible with that employed to investigate the spatial ecology of the same fish during marine life-phases (Davies et al. 2020), although this could be mitigated by deploying marine receivers that function over a range of frequencies.
Further work could also seek to provide a mechanistic understanding of reduced passage rates in newly tagged fish; experimental studies could elucidate and separate potential effects of capture, sedation and tagging on key predictors of passage ability such as motivation, orientation and swimming performance (Cooke et al. 2011).

Summary
This study quantified the impact of weirs on upstream migrating twaite shad.While returning individuals to their spawning rivers are a rare feature of telemetry-based assessments of barrier passage, their use in this study, enabled by advancements in telemetry technology and tagging protocols, was crucial in their use as 'controls' for understanding potential tagging bias and for understanding the effect previous experience on passage ability.The results revealed that even with previous weir passage experience, migrating fish could still be delayed or not pass at all, with elevated river levels and water temperatures important for passage.Taken together, these results are important contributions to contemporary understandings of anadromous fish migration in fragmented river basins.
The authors have no competing interests to declare.Tables Table 1: Locations and characteristics of study weirs in the River Severn basin, UK (Figure 1) during the study period, which were used to assess the impacts of weirs and factors affecting approach and passage during the upstream migration of acoustic-tagged twaite shad Alosa fallax.The number of fish detected moving upstream with an unobstructed upstream route to a weir

Each weir n approached
The number of upstream-moving fish that were detected on the receiver immediately downstream of a weir

Each weir
Per cent approach, % The proportion of n available fish that approached a weir Each weir n passed The number of fish approaching a weir that were subsequently detected on an upstream receiver        The effect of river level recorded on passage rates.B: The effect of temperature on passage rates.C: The effect of diel period on passage rates.For continuous covariates, survival distributions are displayed for representative data categories (Goerig et al. 2020).
Curves represent % of shad that are yet to pass the weir at each time point.Covariates effects presented are environmental covariates shown to have a significant effect on passage rates in the top ranked mixed-effects Cox models.

Funding
PD was supported by a match-funded PhD grant from the 'Unlocking the Severn' project (Heritage Lottery Fund, Grant/Award Number: HG/15/04573, LIFE Nature Programme Grant/Award Number: LIFE15/NAT/UK/000219), and Bournemouth University.The authors are grateful for funding for the purchase of acoustic tags from the UK Department of Food and Rural Affairs (DEFRA).
as a proportion of the time taken to reach the upstream extent of migration from immediately downstream of the first migration barrier Each individual

Figure 2 :
Figure 2: Upstream passage times of acoustic-tagged twaite shad Alosa fallax through unobstructed versus obstructed reaches of the River Severn, UK.The obstructed reach was downstream S2 to upstream S2 (1 km) and the unobstructed reach was upstream S1 to downstream S2 (17 km) (see figure 1).Passage times were standardised to represent upstream passage times through one km of river reach.

Figure 3 :
Figure3: Numbers of acoustic-tagged twaite shad Alosa fallax detected and their upstream migratory extent in the River Severn basin, UK (Figure1), tracked during spawning migrations in 2018-2020.The percentage of shad reaching each receiver, and the percentage of shad reaching their upstream extent of migration at each receiver, are represented by the size and colour intensity of the circles, respectively.Data are pooled for newly-tagged and returning fish.The weir codes are as in Table1.A: Upstream extent of shad migrations recorded upstream of weir S1 (n migrations = 152).B: Upstream extent of shad recorded upstream of weir S2 (n migrations = 164).

Figure
Figure 4: Summary of covariates from the bestfitting models of weir approach likelihood in twaite shad.A: Number of approaching/nonapproaching individuals by weir for newly tagged and returning individuals.B: Body length of approaching/nonapproaching individuals by weir for newly tagged and returning individuals.C: Number of approaching/nonapproaching individuals at weirs S3/T1 by previous approach, for returning individuals.

Figure 5 :
Figure 5: Distribution of first arrival times of newly tagged (red bars) and returning (grey bars) acoustic-tagged twaite shad at weir S2 during April and May across the three study years.Mean daily water temperatures are displayed as a red line.

Figure 6 :
Figure 6: Kaplan-Meir depletion curves for passage of weir S2 by acoustic-tagged twaite shad.A: The effect of tagging status (newly-tagged versus returning) on passage rates.B: The effect of previous success on passage rates by returning individuals.Curves represent % of shad that are yet to pass the weir at each time point.Covariates effects presented are from individual covariates shown to have a significant effect on passage rates in the top ranked mixed-effects Cox model

Table 2 :
Summary metrics for acoustic tagged twaite shad Alosa fallax captured over two years in the River Severn, UK (Figure1)

Table 3 :
Definition of metrics used to quantify approach and passage of weirs in River Severn basin, UK, by acoustic-tagged twaite shad Alosa fallax, and the impacts of weirs on individual migration

Table 4 :
Summary of emigration and return rates by twaite shad Alosa fallax tagged with 3year acoustic transmitters in the River Severn, UK, in 2018 and 2019