The Response of River-resident Fish to Reservoir Freshet Releases of Varying Profiles Intended to Facilitate a Spawning Migration

This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1029/2018WR024196 ©2020 American Geophysical Union. All rights reserved. The Response of River-resident Fish to Reservoir Freshet Releases of Varying Profiles Intended to Facilitate a Spawning Migration


Introduction
River ecosystems are hydrological networks structured by the flow of water, sediment and nutrients, which may facilitate the movement and migration of inhabiting fauna (McCluney et al., 2014). All elements of the natural flow regime (Poff et al., 1997), including the quantity, timing and variability of flows, are considered ecologically important drivers for riverine ecosystem functioning (Enders et al., 2009;Nislow & Armstrong, 2012). Fish populations rely on a variety of flows to migrate between and exploit a diversity of habitats (feeding, spawning and refuge) and complete their life cycles (Cowx & Welcomme, 1998;Nislow & Armstrong, 2012). The majority of rivers in the developed world are impounded in some way, often by water-storage reservoirs for potable supply, flood control or hydropower (Gillespie et al., 2015;Vörösmarty et al., 2010). Impoundments alter the magnitude, timing and duration of ecologically important natural flood flow characteristics, which should typically facilitate spawning migrations (by providing the cue and opportunity) with subsequent impacts on the downstream fish communities (Nilson et al., 2005).
The European Union Water Framework Directive (WFD) states that all heavily modified water bodies, such as those impounded by dams, must reach Good Ecological Potential (GEP) by 2027 (WFD; 2000/60/EEC). GEP is the ecological quality that can be achieved in the affected water bodies without significant adverse impacts to the societal benefits provided. In an attempt to comply with this, the UK Technical Advisory Group (UKTAG, 2013) provided recommendations, using the concept of "environmental flows" or eflows, to identify a number of ecologically important components of river flows. The application of environmental flows is widely used (Acreman et al., 2014) as a mitigation measure in regulated rivers around the world and is defined as "the quantity, timing, duration, frequency and quality of water flows required to sustain freshwater, estuarine and near-shore ecosystems and the human livelihoods Accepted Article ©2020 American Geophysical Union. All rights reserved. and well-being that depend on them" (Acreman and Ferguson, 2010). Regulated rivers receive these flows through artificial freshet releases of appropriate magnitude, duration, timing and frequency. These components are referred to as 'building blocks' (Figure 1) and provide guidance to identify which are likely to be ecologically beneficial in a particular river, at a particular time of year. This approach aims to find the most efficient flow regime that conserves ecosystem functioning whilst preserving water for potable supply.
Research into the impacts of reservoir releases on downstream biota and the importance of sustainable water management is increasing worldwide (Chen & Olden, 2017;Pahl-Wostl et al., 2013;Sabo et al., 2017;Vörösmarty et al., 2010). Research into fish migration during freshets has predominantly focused on anadromous salmonids, which move from the sea into fresh water to spawn (e.g. Aprahamian et al., 1998;Hawkins & Smith, 1986;Hawkins, 1989;Laughton, 1991;Smith et al., 1994;Solomon et al., 1999;Webb & Hawkins, 1989). Few studies have investigated the response of river-resident species, which perform spawning migrations entirely within fresh water, in response to artificial freshet releases from reservoirs.
Further studies are hence required to develop evidence-based mitigation guidance to provide suitable freshet reservoir-release. Notwithstanding, there is currently a dearth of knowledge about the migration of brown trout to their spawning grounds during artificial freshets in regulated rivers, despite explicit recommendations for the timing, frequency, magnitude and duration of the autumn/winter flow elevation building block to reach GEP in UKTAG guidance (Table 1; UKTAG, 2013). Brown trout are often the dominant fish species in upland rivers where reservoirs are prevalent in many regions, making them a suitable study species.

Accepted Article
©2020 American Geophysical Union. All rights reserved.
During this empirical investigation, the influence of the timing, magnitude and duration of autumn/winter freshet releases from two impounding water-storage reservoirs on brown trout movements was assessed over multiple years. The UKTAG recommendations for the autumn/winter flow building block for dispersal and fish migration were used to guide freshet releases ( Figure 1; Table 1). The freshet profiles were iteratively adapted based on findings and knowledge from the previous year(s). To provide control data, movements of fish during freshets were compared to movements of the same fish on days when freshets were not being released, and to fish in reaches not being impacted by the freshet release, i.e. to compare movements to those under 'normal' conditions. By using multiple years of research and a control reach, we uniquely attempted to robustly quantify the autumn/winter flow building block required to stimulate and facilitate river-resident brown trout spawning migrations. Fish movements during freshets, in comparison to the days before and after freshets and relative to the entire tracking period were analysed to identify whether a spawning migration occurred.
Movement for reasons other than a spawning migration inevitably occurred and were also explored to establish if there were unintended benefits of the freshet releases and to help understand the general ecology of river-resident fish in regulated rivers. The findings will inform practical and evidence-based guidance on environmental flows for reservoirs operated by the water industry.

Study area
The impact of reservoir freshet releases, intended to simulate natural high-flow events, on brown trout movements during the spawning migration season was assessed downstream of two potable water supply reservoirs in the River Holme catchment in northern England. The reservoirs and downstream river were deemed a typical heavily modified water body that must Accepted Article ©2020 American Geophysical Union. All rights reserved.
reach GEP according to WFD (2000/60/EEC). The aim was to find the most efficient flow regime that provides conditions required by inhabiting fish in order for them to complete their life cycles whilst conserving water for potable supply. Brownhill and Digley water storage reservoirs on Ramsden Clough and Marsden Clough, respectively, were located approx. 4km southwest of Holmfirth, West Yorkshire ( Figure 2). The study was performed in three reaches in October and November 2012 (a, b and c; Figure 2) and two reaches further downstream were added to the investigation in October to February 2013/14 and 2014/15 (d and e; Figure 2). The experimental design allowed the movements of fish in response to freshets (in impact reaches) to be compared with those in a reach unaffected by the release (control reaches), as only one reservoir released water at any one time; i.e. when a freshet was released from Digley Reservoir, fish in Ramsden Clough were used as controls and Marsden Clough for Brownhill Reservoir ( Figure 2). It was expected that if freshets resulted in upstream migrations and brown trout were seeking areas that are inaccessible due to impassable weirs (indicated in Figure 2), they would congregate downstream in weir pools. Reservoir overtopping was logged on approximately a weekly basis but was ungauged, and did not coincide with any freshet releases.
The study reach was typical brown trout habitat and spawning habitat was identified throughout using qualitative walkover survey, Wolman pebble count (Wolman, 1954) and quantitative assessment of depth, flow and substrate size (Armstrong et al., 2003).

Freshet design
Eleven freshets (09:00 am release start) of contrasting timing, magnitude and duration were investigated during three study years, i.e. 2012, 2013 and 2014. These were employed using an iterative process based on direct observation of river flow and fish movements in previous years

Accepted Article
©2020 American Geophysical Union. All rights reserved.
The profile was comparable to freshets released for ecological reasons since 2004. In 2013, freshets of differing timing and magnitude were released: i.e. freshets were released in October, November and December, alternating between small-and large-magnitude (peak = 122.4 × Qn95; 465.0 ML/d) (Figure 3). In 2014, the magnitude and duration of freshets released were medium-magnitude (peak = 43.1 × Qn95; 163.9 Ml/d) and long-duration (28-hours) in October, November and December, but a small-magnitude and long-duration freshet was released in December ( Figure 3). Small-magnitude and long-duration freshets scheduled for October and November 2014 were cancelled due to natural reservoir overtopping events presenting a downstream flood risk.
Brown trout spawning migrations in unregulated rivers occur in October, November and December (e.g. Ovidio et al., 1998;Armstrong et al., 2003). During the investigation, UKTAG guidance specifying that autumn and winter flow elevations should have a 6 × Qn95 magnitude, 12-hour duration and occur once per week at night in October, November and December "to support the migration… to their spawning grounds" (UKTAG, 2013). Operational constraints and reservoir licensing restrictions prevented a magnitude of exactly 6 × Qn95 being achieved at this site at night during small and medium magnitude, long duration releases in 2014, i.e. 5.0 (20.0 ML/d) and 8.4 (32.1 ML/d) × Qn95 were released, respectively.

Sampling, tagging and tracking procedure
The largest available brown trout, i.e. the adults in the population, were caught by electric fishing, anaesthetised using buffered tricaine methanesulphonate (MS-222, 0.08 g L -1 ), weighed (g) and measured (fork length, mm) ( Table 3). The size of tagged brown trout did not differ significantly between batches (permutation tests for independence on length; Z = -0.3688, n = 145, P = 0.712 and weight; Z = -1.280, n = 145, P = 0.200). Prior to surgery, the unique frequency (between 173.000 and 173.999 MHz, with a nominal spacing of 10 kHz) of each tag was verified and logged using a hand-operated receiver. Radio transmitters were sterilised with diluted iodine solution and rinsed with distilled water prior to use. An 8-10 mm long, ventro-lateral incision was made anterior to the muscle bed of the pelvic fins and the whip antenna was run via the incision in the body cavity to the exterior, posterior to the pelvic fins using a shielded needle. The transmitter was then inserted into the body cavity and the incision closed with an absorbable suture. Gills were irrigated with a diluted dose (0.04 g/L -1 ) of anaesthetic throughout the tagging procedure, which lasted between 3-4 min. Each fish was released at the approximate site of capture when fully recovered from the anaesthetic. Radio tagged brown trout were located manually using a hand-operated receiver (Sika model, Biotrack, Wareham, UK) and a three-element Yagi antenna daily in 2012 and weekly in 2013 and 2014 as the study period was longer and over a larger area. In 2013 and 2014, fish we also located daily for three days before and two days after freshet releases. During freshets in 2012, 2013 and 2014, fish were located every 30-minutes, 1-hour and 4-hours, respectively. When a fish changed location, the longitudinal distance moved was measured (to the nearest metre) using a tape measure in order to record short distances accurately. Temperature was recorded at 15-min intervals on a Tinytalk logger in each study year (Gemini Data Loggers; www.geminidataloggers.com).

Brown trout movement data analysis
Fish move for a large spectrum of fundamental behavioural and ecological reasons, including spawning migrations, habitat exploration, for feeding and/or predator avoidance. The intention of freshet releases during this investigation was to initiate and facilitate a spawning migration, therefore the analysis performed was tailored to identify whether such a migration occurred. In doing so, non-spawning migration movements were inevitably reported and hence explored, although this was not the primary focus of the investigation. There is no universally accepted definition of migration, but fish movement during a spawning migration is thought to be persistent, undistracted and straightened-out (Dingle, 1996;2006), and between separate habitats (Northcote, 1984), discrete sites (Baras & Lucas, 2001) or localities (Shaw & Couzin, 2012). Migrations are also thought to involve a substantial proportion of the population (Northcote, 1984;Shaw & Couzin, 2012) moving with predictability or synchronicity in time (Baras & Lucas, 2001;Brönmark et al, 2013;Shaw & Couzin, 2012). Therefore, if a freshet release facilitated a spawning migration during this investigation it was anticipated that a large proportion of radio tagged brown trout would have performed a unidirectional movement to a new location, i.e. a discrete patch of spawning habitat.
The spawning location fish are migrating towards will, in theory, be a different finite distance from the starting point of each individual and thus migration distance will vary between individuals. Therefore, both the pattern and extent of movements were analysed to deduce whether a spawning migration occurred to avoid concluding fish that moved a relatively short, unidirectional distance during a freshet did not perform a spawning migration. It was assumed a fish would stop moving when it reached a spawning location, and thus be at the extremities of its range during the freshet when the freshet ended. Likewise, if a freshet ended prior to a fish completing a spawning migration it would also be at the extremities of its range during the freshet.

Accepted Article
©2020 American Geophysical Union. All rights reserved.

Movements during freshet releases, in comparison to control reaches
Three metrics were used to quantify the extent of fish movement during a freshet, namely range, total distance moved and beeline distance. The range during a freshet was calculated as the longitudinal distance between the furthest upstream and downstream position recorded for each individual. The total distance moved during a freshet was the sum of the longitudinal distance moved by an individual between each occasion it was located. The beeline distance was distance between locations occupied immediately before and after a freshet (Bovet & Benhamou, 1988).
Two indices were used to quantify if the pattern of fish movement during a freshet was characteristic of a spawning migration, namely directionality and relocation. The directionality index, also referred to as straightness or tortuosity by others (Fritz et al., 2002;Morales & Ellner, 2002), was the ratio between the beeline distance and the total distance moved (beeline distance / total distance moved) during the freshet, ranging between 0 and 1, with a higher value indicating fish performed more unidirectional movements during freshets, as expected during a spawning migration. The relocation index was the ratio between beeline distance and the range (beeline distance / range) during a freshet, ranging between 0 and 1, with a higher value indicating the fish relocated closer to the extremity of the range occupied during a freshet release, as expected during a spawning migration. Total distance moved was plotted with directionality, and range with relocation, including a key to aid interpretation of fish movements; data in the top right sector of both plots would be indicative of a spawning migration.

Movement before, during and after freshet releases
It was assumed that fish movements during days immediately before and after freshets were normal for that month and prevailing environmental conditions. Hence movements by the same Accepted Article ©2020 American Geophysical Union. All rights reserved.
individuals would be larger on the day of a freshet if it facilitated a spawning migration. Daily distance moved by individual fish were calculated as the longitudinal distance between locations occupied on consecutive days for fish in the impact reach in the three days before and two days after a freshet, thus enabling comparison to distance moved during the day a freshet was released. Box plots of daily distance moved in the impact reach before (days 1 -3), during (day 4) and after (days 5 and 6) each freshet were plotted to enable extent and direction of movements to be visualised; a box in the upper half of the plot (i.e. upstream movement) on day 4 would be indicative of a spawning migration.

Movements during freshets relative to the entire tracking period
A fundamental assumption of performing a freshet release to facilitate a spawning migration is that river-resident fish do not/cannot perform long distance movements in their absence. If this was true, the distance moved by an individual during a freshet would represent a large proportion of the distance moved by that individual during autumn/winter. In an attempt to quantify this, the range during a freshet was calculated as a percentage of the home range occupied during the whole tracking period for each individual, and was referred to as relative range during freshet. Home range was determined by the longitudinal distance between the furthest upstream and downstream position recorded for each individual during the entire tracking period (Hojesjo et al., 2007). Range per day tracked was also calculated by dividing the home range by the number of days over which the individual was tracked, which describes the extent of river used, standardised for the period of tracking (Ovidio et al., 2002). In addition, the extent and timing of the largest unidirectional movements performed at times other than during a freshet were also reported.

Statistical analysis
As there were no pairwise significant differences in home range between sites, data for fish inhabiting all reaches impacted by a freshet release were pooled for analysis. Data were tested for normality of variance using Shapiro-Wilk Normality tests. Data were non-normally distributed during comparisons between two groups (between control and impact reaches for range, total distance moved and beeline distance as well as directionality and relocation indices, and relative range during each release) and there were many tied values in the data.
Therefore, in order to obtain exact P-values, permutation tests for independence were conducted using the 'coin' r package (referred to as permutation-test) (Hothorn et al., 2006).
Permutation tests were also used to compare daily distance moved between the three days before, day during and two days after each freshet release. A Wilcoxon Rank Sum test was performed on daily distance moved between these groups (before, during and after) where necessary to identify in which days the movement was significantly different to each other. For other comparisons between multiple groups for non-parametric data that were not tied (i.e. overall range per day tracked between all three study years), Kruskal-Wallis tests (referred to as KW-test) were used with a Dunn non-parametric pairwise multiple comparisons post-hoc test (referred to as post-hoc test) using the r package 'Dunn.test ' (Dinno, 2017). Fish length and total distance moved, beeline distance, range, directionality and relocation during freshets and range per day tracked in each study year were tested for correlations using Pearson product-moment correlation (referred to as cor-test). Median and interquartile ranges were extracted from data in tables using the r package 'purrr' (Henry & Wickham, 2018). All statistics were carried out in R studio v 3.3.0.

Extent of movement
The extent of movement during freshets was generally small for the majority of fish and thus was not considered representative of a spawning migration. Total distance moved was less than 20 m (impact = 68.1% and control = 86.8%), beeline distance was less than 10 m upstream or downstream (83.3% and 98.9%) and range was smaller than 20 m (84.3% and 90.1%) ( Figure   4). The total distance moved and range were significantly larger in impact than control reaches during all but one short-duration freshet studied (see Table 4 for statistics). Total distance moved and range were all larger during/after long rather than short-duration freshet releases in impact reaches, but were always comparable to control reaches (Table 4). Beeline distance from the location occupied prior to the release was statistically comparable between impact and control reaches after all freshets. There was also no correlation between fish length and total distance moved, beeline distance or range in impact reaches during freshets (Pearson's product-moment correlation, P > 0.05).

Accepted Article
©2020 American Geophysical Union. All rights reserved.

Pattern of movement
The pattern of movement in impact reaches was not statistically larger than in control reaches during any freshets (see table 5 for metrics and statistics), and thus were not considered representative of a spawning migration. Indeed, only a small proportion of fish performed exclusively directional movements indicative of a spawning migration during freshets (impact = 7.1% and control = 25.3%) or relocated at the extremities of their range after freshets (31.0% and 47.3%), i.e. directionality / relocation index = 1 (Figure 4). Instead, a substantial proportion of the population performed highly tortuous movements during freshets or were located close to the location they occupied prior to the release after the freshet, as indicated by the large proportion of data in the lower half of plots in but not during any other freshets.

Movement before, during and after freshet releases
There was no significant difference in the daily distance moved (longitudinal distance between the fish locations each day) in the impact reach before (days 1 -3), during (day 4) and after (days 5 and 6) for any of the freshets (Permutation-tests; P > 0.05; Figure 5), these movements were hence not considered representative of a spawning migration. Daily distance moved was significantly different during the small-magnitude, short-duration freshet in December 2013 (Z = 2.097, P = 0.036*; Figure 5), but differences were between days before and after the freshet (Wilcox-test; W = 3413.5, n = 196, P = 0.001). The largest distance moved by an individual fish in a day did not coincide with a freshet release; this was 554 m upstream three days before the medium-magnitude, long-duration freshet release in November 2014 ( Figure 5).

Movements during freshets relative to the entire study
The home range during the tracking period in 2012 was 67.3 ± 99.2 (5.8 -525.3 m); in 2013 was 99.6 ± 77.2 (19.2 -300.1 m), and in 2014 was 99.1 ± 128.2 (11.7 -1090.5 m). Range per day tracked in 2012 (1.7 ± 2.5 (0.1 -13.1 m)) and 2013 (1.8 ± 1.5 (0.3 -8.3 m)) were significantly smaller than in 2014 (2.6 ± 6.7 (0.4 -57.4 m)) (KW-tests; X 2 = 7.8865, df = 2, P = 0.019; 2012 posthoc P = 0.013 and 2013 posthoc P = 0.012). There was no significant correlation between brown trout length and range per day tracked in any of the three study years (cor-test, P > 0.05). The relative range during freshet releases was small in comparison to the entire tracking period, but it was always larger in impact than control reaches (Table 6) significantly so during small-magnitude, short-duration releases in October and December 2013 ( Table 6). competition and trophic interactions (predation risk) affect habitat selection while Johnsson & Forser (2002) found increased residence duration with increased perceived territory value in brown trout, which may also help explain lack of movement in the present study. From an applied perspective, the current guidance (i.e. UKTAG, 2013) applies to all water storage reservoirs, regardless of distribution and quantity of spawning habitat in the downstream reach.
That said, the downstream river and prevailing fish population were considered typical of a heavily modified water body thus raising doubts over the efficacy of artificial freshets for riverresident fish to perform a spawning migration.
In this study, brown were more active (i.e. total distance moved) and had a greater extent of movement (i.e. range during freshet) relative to fish in control reaches during short duration freshets and relative range was significantly larger for two short-duration freshets (both smallmagnitude in October and December 2013). Brown trout are known to occupy flow refugia during severe spates and/or extreme floods to minimise energy expenditure and avoid displacement or mortality (Lobón-Cerviá, 1996), but the magnitude of freshets studied were relatively small and extent of movement was not indicative of fish occupying the least metabolically costly habitat. Instead, while not the objective of the freshets released, short duration freshets may have benefitted individual fish by providing a short opportunity to search to gain knowledge (Gowan & Fausch, 2002) or find superior habitat (Crook, 2004;Gowan et al., 1994;Smithson & Johnston, 1999). Some individuals homed to the location previously ecological response, with 38% reporting an increase, 25% had no change and 21% had a decrease (Gillespie et al., 2015), with magnitude having little influence during this study.
The largest unidirectional movements in the present study occurred when a freshet was not being released from a reservoir, though such movements rarely coincided with that of another fish and thus were probably not a spawning migration. Movements also did not seem to be in response to a particular flow or temperature change, which is in contrast to studies from unregulated rivers with Ovidio et al., (1998) who found spawning migrations occur when water temperature was 10-12 °C and Jonsson and Jonsson (2002) who reported both rising temperature and increased flow were important in stimulating the upstream movement of brown trout. More importantly, in the context of freshet releases, some fish performed longdistance movements during periods of low flow, and thus were not reliant on freshets for the opportunity to perform a long-distance spawning migration. That said, long-distance movements mostly occurred during periods of elevated river level due to rainfall and reservoir overtopping events. Such observations have been found previously for brown trout (Bunnell et al., 1998;Heggenes et al, 2007;Ovidio et al., 2002). Ovidio et al., (1998) also reported large movements occurred in response to reservoir overtopping events rather than freshets of comparable timing. This could be related to the predictability of such events due to preceding rainfall (not present prior to freshet releases), difference in discharge profile and/ or water quality when compared to artificial freshet releases, and is something that could be investigated further. Based on these findings, measures to promote overtopping such as storing water during the spawning season for target species may be pursued by water companies.
Despite this study finding little evidence to support the use of freshets for river-resident fish to perform spawning migrations they may have had unquantified benefits, such as mobilising food items (Gibbins et al., 2007) and releasing and redistributing (also known as flushing) fine

Accepted Article
Copyright © 2020 John Wiley & Sons, Ltd. sediment from spawning gravels beds (Petticrew et al., 2007), as well as structuring invertebrate communities (Lagarrigue et al., 2002;McKinney et al., 2001). Further, it is widely accepted that understanding the natural flow regime has played a large part in the development of environmental flows science and application in the past two decades (Poff, 2017) Africa (O'Brien et al., 2017).
During this study, the reservoirs, downstream river and prevailing fish population were considered typical of a heavily modified water body that must reach GEP according to WFD (2000/60/EEC). The study reach was bounded by weirs in both the upstream and downstream direction but the extent and pattern of fish movement during freshet releases was not restricted by their presence given fish did not congregate downstream of weirs, movements were comparable in control reaches/days and far larger movements occurred on days without freshet releases. The principle aim of this study was to investigate the influence of autumn/winter Table 1. UKTAG recommendations for autumn and winter flow elevations to support brown trout in rivers to their spawning grounds and the migration of adult salmon, sea trout, river and sea lamprey, in order to reach good ecological potential (UKTAG, 2013     Permutation test significance indicated by indicated by * = P < 0.05, ** = P < 0.02 and *** = P < 0.01.