Quantifying the habitat and zoogeomorphic capabilities of spawning European barbel Barbus barbus, a lithophilous cyprinid

Suitable gravel availability is critical for the spawning success of lithophilous fishes, including redd builders. Redd construction during spawning can alter substrate characteristics, thereby influencing hydraulic conditions and sediment transport, highlighting the importance of spawning as a zoogeomorphic activity. Here, interactions between redd‐building fish and their spawning environment were investigated for European barbel Barbus barbus with a comparative approach across three English rivers: Teme (western), Great Ouse (eastern) and Idle (central). Sediment characteristics of spawning habitats were similar across the rivers, including subsurface fine sediment (<2 mm) content (≈20% dry weight), but elevated subsurface silt content and coarser surface sediments were found in the river Teme. Water velocities were similar at spawning sites despite differences in channel width and depth. Redds were characterized by a pit and tailspill, with no differences in surface grain‐size characteristics between these and the surrounding riverbed, but with topographic alteration (dimensions and tailspill amplitude) in line with those of salmonids. Estimates of the fraction of the bed that spawning barbel were capable of moving exceeded 97% in all rivers. Estimated reproductive potential varied significantly between the rivers Idle and Teme (3,098 to 9,715 eggs/m2), which was largely due to differences in barbel lengths affecting fecundity. Larger barbel, capable of producing and depositing more eggs, but in more spatially extensive redds, meaning fewer redds per given surface area of riverbed. Predictions of barbel egg mortality based on sand content were low across both rivers. The effects of silt on barbel egg and larvae development are unknown, but the levels detected here would significantly impact salmon egg mortality. Similarities in fish length to redd area and the size of moveable grains by spawning barbel and salmon suggest they have similar geomorphic effects on sediments, although fine sediment tolerance is highly divergent.

Knowledge of spawning substratum preferences amongst nonsalmonid lithophilous fishes is comparatively poor, with a paucity of information on how river hydrology interacts with sediment composition to influence larval emergence success and recruitment (Mann, 1996, Baši c et al., 2017, Duerregger et al., 2018 Table 1). This is despite the high ecological and recreational importance of many of these fishes (Winfield & Nelson, 1991), such as the European barbel Barbus barbus (barbel hereafter) (Britton & Pegg, 2011). Studies of barbel have focussed on the extent of pre-spawning movements (Baras, 1997; Baras & Cherry, 1990; Baras, Lambert, & Philippart, 1994), with knowledge of spawning requirements limited to studies suggesting preferences for shallow (≈37 cm) and fast-flowing water with loose gravels near overhanging vegetation (Melcher & Schmutz, 2010). Barbel are a lithophilous, aggregative fish, typically inhabiting the middle reaches of European rivers from southeastern England and France in the west to the Black Sea basin in the east (Britton & Pegg, 2011).
The spawning behaviours of salmonids and barbel have some similarities, especially redd construction where similar-sized females  excavate pits in the substrate using rapid, lateral flexions of their bodies, prior to gamete deposition, fertilization and burial (salmonids: Burner, 1951; barbel: Baras, 1994). Despite similarities in spawning mechanisms, there is no quantitative information on the habitat and redd characteristics of spawning barbel. While these knowledge gaps can, at least in part, be informed by the salmonid literature (Table 1), there are important differences between the fishes, including timing of spawning seasons, incubation duration, egg sizes and burial depth. Globally, salmonids can spawn across the year, with timing dependent on latitude (Beechie, Moir, & Pess, 2008); in Britain, salmonid eggs are deposited in redds in winter, with emergence in the following spring. Thus, their eggs are exposed to fines for longer periods than the eggs of fast-incubating spring/summer spawning cyprinids (Baši c et al., 2017). Exception to this longer period of exposure to fines in salmonids includes spring-spawning steelhead trout Oncorhynchus mykiss with a 7 to 10 day incubation period (Goode et al., 2013;Phillips, Lantz, Claire, & Moring, 1975) versus 12 to 20 days in barbel (Baši c, Britton, Rice, & Pledger, 2018) under similar sediment but different temperature conditions (steelhead: 7-11 C, barbel: 16-18 C). Another key difference is the egg diameter, with salmonid eggs generally larger than cyprinid eggs (5-9 mm vs. 0.5-3.0 mm, respectively) (salmon: Aulstad & Gjedrem, 1973;Beacham & Murray, 1993;barbel: Pinder, 2001), indicating higher metabolic demand for salmonid eggs (Rombough, 2007). Salmonid eggs are also often buried deeper (salmonids: 11-30 cm depth; cyprinids 3-17 cm depth; Table 1), with oxygen availability decreasing as sediment depth increases (Freixa et al., 2016). Because of these differences, the effects of fine sediment on spawning success of barbel remains unclear and although Baši c et al. (2018) revealed delayed larval emergence from substratum with high sand content (>30% sand by mass), impacts on egg-to-emergence survival were not detected.
Understanding how spawning fishes influence river geomorphology and hydrology is also important, especially where significant numbers of large-bodied spawning fish are involved in redd construction, as observed in salmonid spawning events (Hassan, Tonina, & Buxton, 2015;Moore et al., 2007). Redd construction is a form of zoogeomorphology (Butler, 1995) whereby animals act as geomorphological agents to move sediments and change morphology, sediments and hydraulics. Female salmonids excavate a pit in the riverbed, during redd construction, that results in localized coarsening of the bed surface within the depression . Sediment displaced during pit creation is transported and deposited downstream, forming the tailspill-a mound of sediment typically coarser than both the pit floor and surrounding bed (DeVries, 1997;Lapointe, Eaton, Driscoll, & Latulippe, 2000;Montgomery et al., 1996;Rennie & Millar, 2000). Such alterations to the physical environment can have important implications for sediment mobility in rivers (Buxton, Buffington, Tonina, Fremier, & Yager, 2015;Montgomery et al., 1996), with the extent of disturbance by some species equivalent to the displacement of gravels caused by flood events (Gottesfeld, Hassan, Tunnicliffe, & Poirier, 2004). Spawning salmonids can also alter hyporheic exchange and add marine-derived nutrients into the riverbed (Buxton, Buffington, Yager, Hassan, & Fremier, 2015). There are, however, some biological controls on redd building, with different species and sizes of fish having different zoogeomorphological effects (Moore et al., 2007;Riebe, Sklar, Overstreet, & Wooster, 2014).
The geomorphic impact of spawning barbel is yet to be quantified, but recent studies have revealed that their foraging behaviour can alter the size distribution and/or structure and so, mobility of fluvial sediments (Pledger, Rice, & Millett, 2014, 2017Rice, Pledger, Toone, & Mathers, 2018), with larger fish having greater effects (Pledger, Rice, & Millett, 2016). For example, large-bodied cyprinids, including adult barbel (>500 mm), disturbed 26% of riffle substrates per day, and displaced clasts up to 90 mm in size . Thus, due to their size and abundance in some rivers, we hypothesize that spawning barbel are effective zoogeomorphic agents T A B L E 1 Summary of spawning habitat preferences of non-salmonid (Cyprinidae and Petromyzontiformes) and salmonid lithophils Using a comparative approach across three English rivers, the Teme (western England), Great Ouse (eastern England) and Idle (central England), objectives were to: (a) assess grain-size characteristics of surface (n = 3 rivers) and subsurface (n = 2 rivers) sediments at barbel spawning sites and use these in conjunction with salmonid spawning models to predict the reproductive potential of barbel spawning sites and estimate egg survival; (b) quantify flow characteristics (water velocity and depth) and oxygen availability at barbel spawning sites; (c) characterize barbel redds and identify how their construction alters surface sediment composition and (d) use published data on barbel spawning and measurements from this study to determine how effective salmon-derived models are at estimating barbel reproductive potential and egg survival. The application of salmonid models to barbel is considered appropriate given the paucity of knowledge on barbel spawning requirements, and the similarities in female lengths and their redd building characteristics. However, we acknowledge that future work is needed to develop barbel-specific models, given the differences in physiology, timing of spawning season, density of spawners, incubation duration and egg survival rates. The salmonid models thus provide initial, first-order estimates of barbel reproductive potential and egg survival.
2 | MATERIALS, METHODS AND DATA ANALYSIS

| Study sites
The study was conducted in the rivers Great Ouse, Teme and Idle, three gravel-bed rivers located in the east, west and centre of England, respectively ( Table 2). All three rivers run through predominantly agricultural landscapes and maintain low gradients (<0.33%; Downs & Thorne, 1998;Neal et al., 2000;Severn Rivers Trust, 2012).
Study reaches were representative of the "Barbel Zone" (Huet, 1959; Figure 1) and hosted translocated, natural or stocked barbel populations (Table 2). Other lithophilous fish occupy the study reaches (Table 2)  Study sites were selected on the Teme, Idle and Great Ouse (n = 13, 20 and 13, respectively; Figure 1) on the basis of direct spawning observations, and/or because they matched existing qualitative descriptions of spawning habitats-riffles with "coarse" substratum and shallow, "moderate to high-velocity flow" (Baras & Cherry, 1990;Hunt & Jones, 1975;Lucas & Batley, 1996).

| Riffle surface and subsurface sediment and analyses
To assess the size distribution of surface sediments at coarse-grained spawning riffles, 300-to 400-count Wolman samples (Rice & Church, 1996) were used at each site on each river. Grains were selected using a systematic grid-sampling scheme, by pacing across the width of each spawning riffle and, at each step, blindly selecting a grain to measure with a gravelometer and then taking a step downstream and returning in the opposite direction.
T A B L E 2 Geomorphological, ecological and biological characteristics of the three study rivers To determine the degree of surface armouring, surface

| Barbel redd characteristics and their impact on surface sediment
The characteristics of six barbel redds at four sites (site 4 at Stanford Bridge: 2 redds; 3 sites at Powick; 10, 11, 12: 4 redds) on the River Teme were measured between May 2015 and June 2017, 2 to 8 days post-spawning. Sample size was constrained by a combination of low numbers of observed redds, accessibility and high turbidity within the river Teme, that reduced visibility and the ability to accurately measure the redds. Three redds measured in 2017 were confirmed as barbel redds by rapid egg assessments that involved disturbing the top 50 mm of tailspill sediment using a pencil in a circular motion, with a hand-held aquaria net placed behind the redd to collect the eggs that were then identified by their colour and morphology (APEM, 2009;. The other three redds were not conclusively identified as barbel redds, but it was considered unlikely that they were constructed by other lithophils present in the Teme, given that minnow Phoxinus phoxinus are too small for their gravel movements to be mistaken for barbel and chub Squalius cephalus are not redd constructors (South East River Trust, 2018; Table 2). Sea lamprey Petromyzon marinus are found in the river Teme but generally spawn later in the year and create a more circular redd with large grains deposited around the edge (Pinder, Hopkins, Scott, & Britton, 2016). Shad Alosa spp. are surface spawners and are thus unlikely to influence bed morphology (Acolas et al., 2004).
Single length and width measurements were made of the two component parts of each redd pit and tailspill (Supporting Information, Figure S1)-the pit (the excavated hole or depression in the bed) and the tailspill (an area where grains, mobilized from the pit during spawning are deposited). Single depth measurements from the water surface to the pit bottom, tailspill top and adjacent river bed (datum) were recorded and used to calculate pit depth and tailspill height by subtracting the depth of adjacent river bed. Length, width and depth measurements were made to the nearest cm (±0.5 cm). Surface sediment samples were collected from each of the pit and tailspill areas to investigate differences in grain-size distributions between these areas and the surrounding riverbed. This was achieved via 30-count Wolman samples. Only 30 clasts were measured in each redd area due to the small redd areas (A PIT = 1.2 ± 0.2 m 2 , A TAILSPILL = 2.3 ± 0.3 m 2 ; mean ± 95% CI) and to minimize redd disturbance. The largest grains visually observed on the tailspill surfaces were measured (mm) to investigate the maximum particle size moved by barbel during redd construction.
The areas (A, m 2 ) of pits and tailspills were calculated with Equation (5) as ellipses as per Ottaway et al. (1981) from measured values of pit and tailspill lengths and widths.
The total area of each redd was calculated by combining pit (A PIT ) and tailspill (A TAILSPILL ) areas. The net volume (V) of each pit and tailspill was then calculated as, assuming that both component parts were half-ellipsoid (McCart, 1969) and the pit depth and tailspill heights were used as vertical parameters in calculations. Volume of sediment displaced during spawning was assumed to be the volume of the tailspill.
Pair-wise Student's t tests in R (R Core Team, 2017 were carried out to compare surface sediment characteristics (D 5 , D 50 , D 95 , mean, sorting, skewness, kurtosis and percent fines) between the pit, tailspill and surrounding riverbed.

| Barbel reproductive potential and geomorphological controls
To assess the reproductive potential of barbel spawning sites, a salmonid-specific model was parameterised, as there was no barbelor cyprinid-specific model available. For salmonids, fork length (FL) is positively correlated with redd area and the maximum size of particles that can be moved by spawning fish (Crisp & Carling, 1989), which can in turn be used to estimate reproductive potential (Riebe et al., 2014). These measurements are taken from dead or dying spawned fish, something not possible for barbel as their post-spawning mortality is very low and they tend to disperse after spawning (Gutmann Roberts, Baši c, & Pledger, 2019;Hancock, Jones, & Shaw, 1976). To provide a consistent set of predictions based on FL, a salmonid relationship from Riebe et al. (2014) was used to estimate barbel redd area (m 2 ; "A REDD "; Equation 7), rather than only using our measured redd dimensions.
An ANOVA was carried out to compare predicted redd area (from the eight fork lengths measured in the Teme) to the six measured redds.
The maximum or threshold particle size (D T, mm) that a barbel of a given length (mm) can displace during redd construction was estimated using Equation (8) (Riebe et al., 2014).
The reproductive potential of barbel at each of the sites was also calculated using the barbel-specific redd measurements from this study, with the mean redd area (A PIT + A TAILSPILL from Equation 5) used instead of A REDD (Equation 7), and using the threshold grain-size D T from the largest grain measured from barbel redd tailspills. A barbel fecundity-length relationship (F(B), Equation 13b) was also determined from 15 manually stripped females from the river Trent at a hatchery (Calverton fish farm, unpublished) and from data on 33 barbel (length minimum-maximum: 314-740 mm) extracted from Hancock (1979).
This enabled a comparison of reproductive potential predicted from salmonid models (12) using salmonid-vs. barbel-specific input parameters for A REDDS , D T and F. High-density barbel spawning observations (Baras, 1994) were also used to estimate the maximum number of redds per square metre (N REDDS ) across a site in the river Meuse, where the fraction of moveable particles is assumed to be 1, due to high and even coverage of redds and in the absence of any grain-size data or redd measurements. Baras (1994) reported 2 years of data, but only successful spawning observations from 1990 were used as this year had the most successful spawning attempts and therefore provides a maximum estimation of spawning potential. The observations are from a 150 m 2 spawning riffle with 71 "successful" spawning attempts in one season, with data on 19 female barbel, with data from 5 females omitted in the original study due to lack of certainty around observations. Assuming no redd superimposition, the spawning riffle area divided by the number of redds (assumed to be number of successful spawning attempts) provides an estimate of redd area.
Barbel egg survival estimates provide indications of the potential suitability of subsurface spawning substratum in the study rivers; however, the absence of suburface sediment data for the river Idle meant these calculations were only completed for the rivers Great Ouse and Teme. To predict the potential effect of bed sediment condition (specifically fines content) on barbel recruitment at individual sites within the two study rivers, egg survival was estimated using salmonid models (Equations 14 and 15; Peterson & Metcalfe, 1981).
Survival % (S c = 0.5-2 mm) and fine to medium sand (S f = 0.06-0.5 mm). Given the lack of information on the synergistic effects of silt-and sand-sized particles on barbel egg development and emergence, Equations (14) and (15) (15). Successful incubation was defined as at least 50% estimated survival, as per Kondolf (2000).
Comparisons of survival potential between the two rivers were carried out using a one-way ANOVA, following normality and homogeneity tests. Values given throughout the results are means (±95% confidence intervals), unless stated otherwise.

| Barbel spawning site characteristics
Spawning riffle areas varied across rivers, from 37 ± 7 m 2 in the Great Ouse to 108 ± 10 and 500 ± 249 m 2 in the Idle and the Teme, respectively. Riffle surfaces utilized by spawning barbel across the three riv-  F I G U R E 3 Flow characteristics and dissolved oxygen across three rivers: Idle (n = 20), Great Ouse (n = 13, four for oxygen only) and Teme (n = 10, five for oxygen only); wetted width (m), depth (m), near-bed velocity (m s −1 ), velocity at 0.6 depth (m/s) and amount of dissolved oxygen (mgL −1 ). Values represent means (±95% confidence interval) and letters right of bars indicate homogeneous groups rivers was 2.13 ± 0.45%. There was no significant difference between rivers in terms of the proportion of subsurface fines (ANOVA; F 1,21 = 0.92, p = 0.35) or sand (ANOVA; F 1, 21 = 1.15, p = 0.30, Table 3). There was, however, a significant difference in silt content (ANOVA; F 1,21 = 23.78, p < 0.01, Table 3) between rivers, with river Teme subsurface sediments (0.47 ± 0.09%) containing higher proportions of silt than river Great Ouse subsurface sediments (0.17 ± 0.08%). Subsurface sediments at four sites in the river Teme contained >0.5% silt and these were located between Ashford Carbonel (site 2) and Powick (site 11), whereas subsurface sediments at only one river Great Ouse site (2) contained >0.5% silt (Figure 4). There was no significant difference between rivers in terms of the organic content of fines (ANOVA; F 1, 15 = 0.18, p = 0.68, Table 3

| Barbel redd characteristics
In the six measured redds, the tailspill tended to be longer and wider than the pit, with a mean difference of 29 ± 22 and 19 ± 17 cm, respectively (Table 3). Mean tailspill height was 14 ± 5 cm and pit depth ranged between 4 and 11 cm, with a mean difference in height of the tailspill and depth of the pit of 20.0 ± 8.8 cm. Total redd area ranged from 1.37 to 9.11 m 2 at sites 10-12, and 2.23-2.58 m 2 at site 4. The mean total redd area was 3.47 ± 0.42 m 2 . The area of the tailspill was larger than that of the pit for five of the six redds. Pit volumes ranged from 1,900 to 34,819 cm 3 (14,390 ± 5,338 cm 3 ; mean ± 95% CI) and the tailspill volume ranged from 0 to 234,834 cm 3 (74,754 ± 43,284 cm 3 ; mean ± 95% CI). Pit volume was larger than tailspill volume for two of the redds, and for four of the redds, the opposite relationship was found. The largest grain diameter found in barbel tailspills was 110 mm, providing a barbel-specific estimate of the threshold grain size (D T ).
Sediments within the pit and tailspill were coarse (mean grain size = 22.2 ± 9.5 and 28.2 ± 3.7 mm, respectively) and moderately well sorted (sorting = 0.58 ± 0.07 and 0.62 ± 0.05 mm, respectively) (Table 4). While fine sediments appeared more prevalent at the surface of the pit than at the surface of the tailspill (Table 4), the difference was not significant (p > 0.05; Table 5). Surface grain-size parameters of the pit and tailspill sediments were similar (t tests, p > 0.05 in all cases; Table 5). Generally, levels of fine sediment found on the redd surfaces were low, but surficial sediments of the pit at site 10 were fines-rich (26.7% fines by area) and finer than the surrounding bed (6.4% fines by area). There were no significant differences between the surface grain-size characteristics of the pit and tailspill areas compared with their surrounding riffle for D 5 , D 50 , D 95 , mean, sorting, skewness, kurtotsis and percentage of fine sediment (Table 5).

| Barbel reproductive potential and geomorphological controls
As mean barbel FLs were 464, 616 and 651 mm for the rivers Idle, Great Ouse and Teme respectively, the maximum particle size that barbel could move (D T , Equation 8) was estimated to be 98, 117 and 121 mm, respectively (Table 6). This salmon-derived estimate suggested an 11 mm overestimation for the Teme, relative to the larg- Extracting data from the Meuse gave a barbel-specific value for N REDDS of 0.47 redds m −2 , which was higher than predicted values across the rivers Teme and Ouse, but lower than the Idle predicted value (Table 6), but as the redds were predicted to be smaller in the Meuse, this could account for the difference, especially as fish sizes in the Meuse and Idle were most similar.
Barbel eggs display a higher tolerance to sand than salmon, with barbel eggs remaining unaffected by SI levels as high as 3.5, which would cause 100% mortality of salmon eggs (Figure 4) Note: Mean barbel length (mm ± 95% confidence intervals) was measured at each study site, threshold particle size (D T , mm; largest grain that can be moved by spawning fish) was predicted (8) for salmonid-specific parameters and measured for barbel-specific parameters, the fraction of moveable surface particles (F M ) was predicted (9), redd area (A REDD , m 2 ) was predicted for salmonid specific parameters and measured for barbel-specific parameters, spawning capacity (number of redds m −2 , N REDDS ) was predicted (11), fecundity was predicted (13a [salmonid], 13b [barbel]), and reproductive potential (N EGGS , eggs m −2 ) was predicted (12). Teme (site-specific) refers to the four sites where redds were measured. N/A; not applicable. a Assumed whole site to be moveable due to high coverage of redds, no sediment data available. When not accounting for silt, one more site in the Teme showed survival above 50% leading to 40% of sites indicating "successful" survival to emergence, but the number of sites did not change for the Great Ouse. The barbel predictions derived from the equation from Across the rivers, surface sediments were relatively coarse and well-sorted gravels, with most substratum metrics similar between the Great Ouse and the Idle, but with coarser sediments in the Teme.
Despite this, the fraction of particles estimated to be moveable by barbel was high (minimum 97%; Table 6) and similar across all three rivers; this was due to the length of barbel relative to the size of gravels. Thus, the presence of large clasts is not expected to be a limitation to barbel redd construction in these three rivers, either when using the maximum grain size estimated by the salmonid model or the largest grain from the tailspill. This has implications for relatively high potential for displacement by barbel, given that the three rivers generally had low D 50 values, all below 35 mm. By comparison, in South Prairie Creek (North America), pink salmon (Oncorhynchus gorbuscha) were capable of displacing grains ≤95 mm, yet the D 50 sizes were 39 to 118 mm (Riebe et al., 2014). Consequently, at spawning sites, pink salmon were only able to move 59% of the bed (Riebe et al., 2014). Nevertheless, coarser sediments can be associated with larger egg sizes within and between populations (Quinn, Hendry, & Wetzel, 1995), which might have positive implications for post-emergence and overwinter survival. For example, larger eggs tend to produce larger larvae that are less vulnerable to predation and displacement, with higher lipid reserves and greater competitive abilities (Nunn, Harvey, & Cowx, 2007). Thus, the benefits of the majority of spawning gravels being made up of particles barbel can move, might be offset by surviving eggs being smaller and potentially leading to increased risk of egg and/or larval entrainment.
To our knowledge this is the first documentation of barbel redd dimensions. It is reasonable to assume the characteristics and persistence of redds will vary between and potentially within systems, as they are functions of spawning barbel lengths, surface sediment conditions, flow characteristics and bed mobility. No significant difference was detected when comparing measured and modelled Teme barbel redd areas, despite a mean 0.54 m 2 overestimation by the salmonid model (Table 4). This suggests that the length to redd relationship between salmonids and barbel is comparable, based on available evidence. The reproductive potential across the specific sites on the river Teme was, however, significantly different when using barbel values compared to using salmon-derived redd area, threshold grain size that a fish could move and fecundity. These led to an underestimate of over 10,391 eggs m −2 site −1 . This was largely due to the barbelspecific fecundity values that are over 10 times higher than the salmon-derived values because of, for example, smaller egg sizes (Table 6). Using published data (Baras, 1994), we were able to determine the number of barbel redds by area as 0.47 redds m −2 in the river Meuse, Belgium. We could not readily compare the Meuse barbel redds by area to the salmon-specific estimates from these study rivers, because barbel from Baras (1994) study were substantially smaller and thus predicted redd size was also smaller. The redd size estimated from the smaller fish in the river Meuse was 2.11 m 2 but this area should be used with caution as it assumed redds were not overlapping, yet Melcher and Schmutz (2010) found that at highdensity spawning sites, barbel redd superimposition was common, as with salmonids (van den Berghe & Gross, 1984). As well as redd superimposition, it is common for salmonids to lay multiple egg clutches within one redd (Tonina & Buffington, 2009a), although the spawning description from Baras (1994) suggests that is not the case with barbel. The data from Baras (1994) was also based on field observations of spawning, yet experimental work from Gougnard, Poncin, Ruwet, and Philippart (1987) showed that not all female redd digging involved the release of ovules. Therefore, spawning observations could lead to an overestimation of successful spawning attempts.
Spawning abandonment can be higher when there are large numbers of males present (Gougnard et al., 1987;Hancock et al., 1976), as in Baras (1994), which is believed to be an evolutionary response that protects eggs from abrasion-related mortality, caused when males mechanically aggravate the redd (Hancock et al., 1976). These high spawning densities were generally not observed across the three study sites here. Barbel inter-spawning periods can be 8 days long, so it would be difficult to establish if all redds constructed by a given female had eggs without disturbing spawning gravels.
Regarding subsurface fines, there was little variation in sand and organic matter content between the rivers Great Ouse and Teme, but Teme sediments contained higher concentrations of silts (Table 3).
Mean silt and clay concentrations were below 0.5%, which is low compared with other lowland English rivers (4-45%; Clarke & Wharton, 2001) and Lapointe et al. (2004) classification of "low fines content" (1.5%). However, presence of silt and clays in spawning gravels at levels as low as 0.03-0.41% can adversely impact Atlantic salmon, S. salar egg incubation and embryo survival (Julien & Bergeron, 2006), with maximum silt levels reaching 0.62 and 0.69% in the Great Ouse and Teme, respectively. The mean sand concentrations of the Teme and Great Ouse (22 and 18%, respectively) were higher than the tolerance threshold of salmonids (14%; Bryce et al., 2010). However, experimental work on barbel suggests survival to emergence is not affected by sand concentrations up to 40%, even though emergence timing can be accelerated by 8 days (Baši c et al., 2018). These early emerging fish were smaller, blind and had larger yolk sacks, suggesting their post-emergence survival could be low relative to more developed later-emerging larvae (Baši c et al., 2018), given their reduced ability to navigate flows and increased predation risk (Krupka, 1988;Vilizzi et al., 2013).
The tolerances of larvae to silt and sand cannot, however, be considered in isolation, as the salmonid literature suggests it is the interactions of the different fine sediment size fractions that can be more important for egg and embryo development (e.g., Lapointe et al., 2004). Indeed, Levasseur, Bergeron, Lapointe, and Bérubé (2006) revealed that it was the proportion of subsurface bed material <0.125 mm in S. salar redds that explained 83% of annual variation in embryo survival; in their study, there was a threshold of 0.2% concentration below which embryo survival dropped sharply below 50%. We found that different size fractions of fines in the rivers Teme and Great Ouse did not result in differences in the mean predicted egg and embryo survival between the rivers. However, there was an inverse relationship between predicted survival and the sand index for the salmonid model ( Figure 5), with the levels of silt reducing survival by 2 and 4% in the rivers Great Ouse and Teme, respectively. Only 30 and 31% of sites were predicted to have egg survival rates over 50% in the Teme and Great Ouse, respectively.
Here, we revealed that barbel have a much higher tolerance to sands than S. salar when using the sand index (Figure 4). This leads the salmonid model to overestimate mortality by a mean of 34% across the sites in the two rivers when silt is not accounted for (Figure 5b Kondolf's (2000) 50% threshold. Fines vary spatially in rivers (Julien & Bergeron, 2006), due to fluctuations in sediment supply, flow velocity, channel profile, erosion, deposition and lithology (Middelkoop & Asselman, 1998). Spatial variation across sites in barbel egg survival rates were not predicted from the barbel-derived calculations, despite salmonid egg survival rates varying from 0 to 63% and in-situ survival in studies showing 0 to 100% (Malcolm, Youngson, & Soulsby, 2003).
The barbel egg survival estimates from the parameterised salmonid models should be used cautiously, given the divergence in the egg survival rates from the salmonid model and the ex-situ egg survival rates of Baši c et al. (2018). Reasons for this are likely to include the higher oxygen requirements of salmonid eggs, although there is a paucity of information on the oxygen requirements of cyprinid eggs (Elshout, Dionisio Pires, Leuven, Wendelaar Bonga, & Hendriks, 2013). This is important, because the detrimental impacts of fine sediments on salmonid eggs and larvae relate to how they inhibit oxygen delivery to eggs and embryos by entombment (Sear et al., 2016) and by decomposition of organic matter (Collins et al., 2013;Sear et al., 2017). Compared to organic content found in other UK lowland rivers (2-21%; Clarke & Wharton, 2001), the organic matter found in these rivers was low (2%), although the implications of this are not clear given the knowledge gap in how organic materials influence cyprinid development (Collins et al., 2013). Our predictions of barbel egg survival did not consider the effect of pore water velocity in redds, although near-bed water velocities were high (Teme: 0.36 ms -1 , Great Ouse: 0.38 ms -1 ). These near-bed water velocities might have facilitated increased oxygen delivery and metabolite removal at depth, increasing egg survival (Lapointe et al., 2004). Also, differences in burial depths might influence egg survival rates between species. For example, barbel eggs in this study tend to be in the top 5 cm of sediment, whereas similar sized salmonid eggs tend to be buried at greater depths (e.g., 15-30 cm for S. salar; Moir et al., 2002). Thus, relative to salmonids, barbel eggs may be exposed to greater hyporheic flows, higher oxygen exchange and increased flushing of metabolic waste, especially where redd topography is similar (Tonina and Buffington, 2009a & b).
The differences in incubation duration and egg size between barbel and salmonids also have implications for their survival (Springate & Bromage, 1985). Longer incubations of salmonid eggs might increase risk of exposure to suboptimal conditions within the egg pocket, such as reductions of oxygen concentrations to critical levels (Sear et al., 2017), especially as salmonid eggs are comparatively large and likely to have higher metabolic rates and oxygen requirements than barbel eggs (Rombough, 2007). Thus, the relatively small barbel eggs that are incubated for relatively short periods are likely to have lower oxygen requirements than salmonids, with a greater capacity for tolerating reduced oxygen levels. However, a recent study on a similar rheophilic cyprinid, nase Chondrostoma nasus, shows that even fine sediment (<0.085 mm) of 10 and 20% content can cause mortality and physically block larvae from emerging (Nagel, Pander, Mueller, & Geist, 2020). No sites were found to have the levels of sand above those that have been experimentally tested (40%) to impact barbel mortality (Baši c et al., 2018).
Spawning barbel altered the topography of the riverbed by displacing volumes of sediment during redd construction. However, the surface grain-size distributions between the pit and tailspill were not different (Table 5), as has been found for other redd-building species (Chapman, 1988;Everest et al., 1987;Kondolf, Sale, & Wolman, 1993;Young, Hubert, & Wesche, 1989). A potential explanation for this is that the sediment excavated from the pit into the tailspill was not limited by grain size for barbel as it has been for some salmonid studies, where unmoveable coarse grains are left in the pit . Tailspill amplitude measured for barbel is comparable to those for Chinook, chum and sockeye salmon (13 and 15 cm; Buxton, 2018) This suggests shear stresses applied to the tailspills of barbel redds may be elevated, as has been found for salmonids. Spawning salmon have been found to change the surface sediment characteristics, although there have been contrasting effects, with sockeye salmon Oncorhynchus nerka reducing the coarseness and sorting of sediments in the redd compared to the surrounding riffles , whereas chum salmon Oncorhynchus keta increase coarseness and sorting (Montgomery et al., 1996). However, altered surface sediment characteristics were not evident from barbel redds in comparison to surrounding riffles. Spawning salmon have also been shown to alter the composition of subsurface sediments, reducing fines content of the egg pocket (Young et al., 1989). Although not tested here, observations of plumes of fine sediment from barbel spawning were made.
The surface sediment showed elevated fines in the pit compared with the tailspill, which could be due to the topography of the pit causing flow recirculation and fine sediment deposition . Barbel eggs were detected 50 mm below the sediment surface in the tailspill, where previous sampling has found eggs within depths of 200 mm (Pinder et al., 2009). In cyprinid species such as nase, egg burial depths can vary between 0 and 300 mm, with those deposited at the surface (less than 5%) being most susceptible to mortality (Duerregger et al., 2018). Barbel eggs that are not retained in the bed can be consumed by other fish, such as grayling, Thymallus thymallus (Perks, 2019), suggesting higher egg mortality for shallowly deposited eggs or in highly mobile sediments. By comparison, salmonids deposit eggs between 0 and 500 mm (Table 1)  The impact of redd construction on river habitats transcends changes in channel morphology and substrata condition. In line with the geomorphic capabilities of some fishes, it is reasonable to assume bed disturbance by spawning barbel may also influence near-bed hydraulics and/or sediment transport processes in at least three ways.
First, and as with barbel foraging (Pledger et al., 2014(Pledger et al., , 2016, it is reasonable to assume particles displaced during spawning, particularly those deposited in elevated positions as part of the tailspill, may sit proud on the sediment surface and in positions of relative instability. Thus, displaced grains may be more susceptible to entrainment, particularly during subsequent high flows. Second, bioturbation of surficial sediments might result in the loss of bed sediment structure and adjustments in packing (Buxton, 2018;Montgomery et al., 1996).
Reworking of sediments in this manner, generating structureless or loosely packed fabrics, may promote bed load transport by reducing critical shear values required for particle entrainment (Buxton, 2018).
Third, redd creation has a significant and varied impact on bed topography, which can influence near-bed hydraulics (Montgomery et al., 1996) and so, potential for particle entrainment. For example, topographic change through barbel redd construction may reduce basal shear stresses  by increasing bed form drag, which may be mitigated by increased shear stresses on tailspill structures (Buxton, 2018;Buxton, Buffington, Tonina, et al., 2015), loss of stabilizing surface structures and/or substrate loosening, increasing particle entrainment probabilities. Further work is required to investigate the impacts of barbel and other cyprinid redd construction on near-bed hydraulics and sediment transport processes.
The spawning riffles studied here were not utilized by high densities of barbel, whereas in the river Ourthe (Belgium), up to 600 barbel utilize riffles of 150 m 2 to spawn (Baras, 1994). Most will be male, but up to 19 females can use the same riffle in a month (Baras, 1994). A single female can have up to 17 successful spawnings in a year, with less than 4 successful attempts being more typical (Baras, 1994). Thus, not all eggs will be deposited in a single redd, as per most salmonid species (Elliott, 1989). Multiple redd creation by barbel thus complicates the estimates of reproductive potential when salmonid model parameters are used. This is also compounded by the high density of female barbel during the spawning season compared to some salmonids, such as S. trutta (e.g., barbel: 0.16 females/m 2 , Baras, 1994; S. trutta: 0.03 females/m 2 , Elliott, 1989). Where spawning habitats are limited, salmonids may utilize suboptimal habitats, and this can facilitate future successful spawning here, the adaptability of barbel to use suboptimal habitats is unknown, but Baras (1994) showed that limited spawning success happened when there was not sufficient spawning habitat, such as specific depth and flow velocity. Barbel spawn in much greater densities than salmonids (salmonids = 0.01 to 0.43; Riebe et al., 2014); barbel = 0.24-0.49 redds m −2 (Table 6) which could lead to more intensive bed disturbance. Consequently, these differences between barbel and salmonids suggest that future work predicting barbel spawning success, such as egg production and survival, is needed to develop more specific models based on their spawning biology (e.g., multiple redd creation).

| CONCLUSION
The present study has started to fill the considerable knowledge gap in the two-way interactions between non-salmonid redd builders and their spawning environment, using barbel as a model species. Barbel were not limited by the size distribution of sediment available for redd construction on riffles in the study rivers, but recruitment may be impacted by fine sediment concentrations within spawning gravels.
The study also indicates a need for further research to determine the optimal conditions for cyprinid egg and larval development and therefore, survival, and how these early-life survival rates influence subsequent recruitment success of cohorts. The influence of these conditions can be tested empirically using both in-and ex-situ scenarios as has been done with salmonids (Cocchiglia, Curran, Hannigan, Purcell, & Kelly-Quinn, 2012), coupled with standardized monitoring of river sediment, water velocity and oxygen conditions. Until this knowledge is developed, the ability of river managers to enhance the spawning success of non-salmonid fishes, such as barbel, will remain highly constrained. Moreover, these studies may enable investigation of how anthropogenically altered rivers are impacting non-salmonid fish communities, and how restoration efforts can ensure sustainable populations in the face of continued environmental change.