Predator, prey, and substrate interactions: the role of faunal activity and substrate characteristics

. Many taxa possess a range of strategies to reduce the risk of predation, including actively seeking suitable refuge habitats; however, the global spread of invasive species may disrupt these behavioral responses. In lotic ecosystems, interstitial spaces in the substrate are important refugia for small organisms. Some predators are ecosystem engineers that exhibit zoogeomorphic agency — the ability to modify the geomorphology of their environment. It is therefore possible that direct ecological effects of predators on prey may be realized through modi ﬁ cations to the prey ’ s habitat, including the availability of refugia, by predators that are zoogeomorphic agents or via external stressors such as ﬁ ne sediment loading. This study examined three research questions in a mesocosm study across a gradient of sediment-stress treatments: (1) What affects do predators ( Pacifastacus leniusculus , invasive cray ﬁ sh) and prey ( Gammarus pulex , amphipods) have on the ingress of ﬁ ne sediment into gravel substrates and therefore on available interstitial refugia? (2) Do prey taxa seek refuge from (invasive) predators in the form of vertical movement into subsurface sediments? and (3) How does ﬁ ne sediment ingress in ﬂ uence predator – prey interactions and prey survival through predator avoidance behavior. Here, we provide direct evidence demonstrating that ﬁ ne sediment ingress into gravel river beds can be facilitated by zoogeomorphic activity with P. leniusculus increasing the in ﬁ ltration of ﬁ ne sand particles (but not coarse sand) during foraging activities. Predator – prey interactions were found to be a primary factor mediating zoogeomorphic activity, with the isolation of cray ﬁ sh from prey ( G. pulex ) leading to increased ﬁ ne sand ingress. When present with signal cray ﬁ sh, G. pulex displayed vertical avoidance behavior, entering subsurface substrates to evade predation by P. leniusculus . Coarse sand treatments resulted in higher predation rates of G. pulex , most likely due to clogging of interstitial pore spaces between gravels limiting the effectiveness of the prey ’ s vertical avoidance behavior strategy. A new conceptual model that captures the interactions between predator, prey, zoogeomorphic processes and habitat availability is presented. This model highlights how predator – prey interactions can be strongly mediated by dynamic bi-directional interactions between organisms and the physical environment they inhabit as ecological and geomorphological processes are intrinsically linked. presence of Pacifastacus leniusculus , sediment treatments ( n = 5), and the interaction between these factors.


INTRODUCTION
Predation and the associated behavior of both predator and prey are important drivers of evolutionary change (Vermeij 1982), with predator-prey interactions receiving considerable research attention (e.g., Heck and Crowder 1991, Chivers and Smith 1998, Schmitz et al. 2017). Many species display a repertoire of strategies and defense mechanisms to reduce the risks associated with predation that reflect adaptations developed over millennia (Strauss et al. 2006, Belgrad and Griffen 2016, Alberti et al. 2017. For instance, prey may respond to the presence of a predator by reducing activity levels and by increasing or timing their use of safe microhabitats to reduce the risk of predation (Lima andDill 1990, Sih andMcCarthy 2002). However, the rapid spread of non-native species across the globe (Early et al. 2016) may disrupt the behavioral response of indigenous native taxa. The overall effect of non-native predators may therefore be enhanced if they possess predation strategies that prey do not recognize or are illequipped to counter (Sol andMaspons 2016, Segev et al. 2017). Understanding the behavioral adaptions that taxa employ to successfully evade and survive predation pressures by native and non-native taxa is therefore important in the context of managing contemporary ecosystems.
Crayfish (order Decapoda) are one of the most widely translocated groups of aquatic organisms (Kouba et al. 2014). Where crayfish invade they typically reduce biodiversity and biomass of the wider faunal community (Twardochleb et al. 2013. The most widely recorded taxa affected by invasive crayfish colonization are Mollusca, with reductions in species richness, abundances, and biomass (e.g., Dorn 2013, Ruokonen et al. 2014. Despite a large number of studies centered on crayfish-prey interactions (Crowl andCovich 1990, Alexander andCovich 1991), studies examining the behavioral response of ubiquitous prey taxa within crayfish invaded ecosystems are limited (but see Haddaway et al. 2014). In response to predation by non-native crayfish, several natural avoidance strategies have been observed in macroinvertebrate species, including rapid locomotion, increased voluntary drift, and vertical migration into the subsurface sediments (Alexander and Covich 1991, Nystr€ om 1999, Haddaway et al. 2014. However, utilization of subsurface sediments as a refuge is dependent on the availability of open, interstitial pore spaces. River bed gravels typically form a structural framework within which finer sand (<2.0 mm) and silt (<0.063 mm) particles are stored (Church et al. 1987). The amount of available pore space depends on (1) the grain size distribution of the framework gravels and their packing density (Frostick et al. 1984, Cui andParker 1998), and (2) the amount and size of fine sediment delivered to and stored within the framework. The relative size of the framework and fine sediment is a key control on bed porosity because it directly affects the propensity of fine sediment to infiltrate into or block (clog) gravel interstices (Frings et al. 2008, Wooster et al. 2008). If interstitial space is sufficient, particles infiltrate and infill from the base of the substrate in a process called unimpeded static percolation (Lunt and Bridge 2007). In contrast, where interstitial spaces are small, large sand grains may block pore throats, impeding the infiltration of fine sediments in a process known as bridging (Gibson et al. 2009a). The availability of interstitial refugia for prey species (e.g., freshwater amphipods) in riverine systems is therefore linked to the bed sedimentology and the fine sediment regime (Gayraud andPhilippe 2003, Jones et al. 2012).
Anthropogenic modifications have altered the quantity and composition of fine sediment delivered to rivers globally (Walling and Collins 2016), with sediment yields of many rivers currently exceeding background levels (Owens et al. 2005, Collins andZhang 2016). Consequent filling (or colmation) of interstitial spaces within the river bed can lead to the disconnection of river bed (benthic) and subsurface (hyporheic) habitats (Descloux et al. 2013, Mathers et al. 2014. Where this occurs, taxa that possess predator avoidance adaptations that rely on vertical movement into the river bed may be subject to reduced refuge availability and therefore enhanced predation pressure. However, the classic view of fine sediment dynamics at the water-substrate interface has been conceived entirely around geophysical principles in which bed shear and fluid turbulence drive entrainment, transport, and deposition of fine sediment (Beschta and Jackson 1979, Diplas and Parker 1992, Kuhnle et al. 2016). There is a growing body of literature demonstrating that animals including fish and macroinvertebrates can also alter the accumulation and distribution of fine sediment (Statzner et al. 1996, Zanetell and Peckarsky 1996, Nogaro et al. 2006, Pledger et al. 2017) via the expenditure of biotic energy . For example, macroinvertebrate prey may winnow fine sediment from interstitial spaces and thereby maintain and/or re-establish vertical connectivity and migration pathways within the river bed (Visoni and Moulton 2003, Mermillod-Blondin et al. 2004, Mermillod-Blondin and Rosenberg 2006, Nogaro et al. 2006, Stumpp and Hose 2017. This is an example of fluvial zoogeomorphology (Butler 1995) the process through which animals alter the geomorphology and sedimentology of their environment (Rice et al. 2012, Statzner 2012, Albertson and Allen 2015, Vu and Pennings 2017. Crayfish are widely recognized zoogeomorphic agents that manipulate coarse sediments and bedload flux , recruit fine sediment to rivers via their burrowing activities (Faller et al. 2016, entrain detritus into the flow when foraging and fighting (Usio andTownsend 2004, Harvey et al. 2011), and increase the suspended sediment load in rivers . It is therefore likely that crayfish may also have a direct influence on the infiltration of sediment into subsurface habitats, a process which has yet to be tested for any biota.
This research therefore sought to examine the engineering of a shared physical environment by a predator and its prey, and the effect of their geomorphological work on predator-prey interactions. These interactions are situated within a broader framework in which ecosystems are subject to multiple external stressors (Sih et al. 2004, Strayer 2010, Jackson et al. 2016) that may drive and/or mediate the interactions between predator, prey, and the physical environment, in this case biological invasions and increased fine sediment loading. Predator-prey interactions are a recurrent theme in ecology, but the typical assumption (conceptualized in Fig. 1A) is that the physical environment is a fixed template in which predation occurs. The growing recognition that organisms can engineer their environments (Jones et al. 1997, Moore 2006, Wright and Jones 2006, in this case via zoogeomorphological processes (Rice et al. 2012, Statzner 2012, means that both predator and prey may modify their shared habitats (Fig. 1B).
In this study, we examine the effect that predators (invasive crayfish) and prey (freshwater amphipods) have on the ingress of contrasting surface fine sediment treatments into experimental substrates and how changes in the physical environment influence predator-prey interactions. Amphipods are diverse and keystone organisms in marine and freshwater systems with over 2000 freshwater species recorded to date (V€ ain€ ol€ a et al. 2008). Gammarus pulex (L.) (Amphipoda: Crustacea) is the most widely distributed and abundant amphipod species in the UK (Gledhill et al. 1993), often dominating macroinvertebrate communities by biomass and abundance (MacNeil et al. 1997). G. pulex are a highly mobile taxon, capable of burrowing into fine sediment to find trophic resources and habitat (Vadher et al. Fig. 1. Conceptual framework of the interactions between predator-prey relations, the physical environment, and the addition of an external stressor. Panel (A) represents the traditional view of predator-prey interactions; panel (B) characterizes the additional processes and interactions between biotic and abiotic factors which take place in the natural environment. Red arrows represent the additional mediating factors considered in this study for the first time. 2015) and are therefore an ideal model prey organism. The North American signal crayfish (Pacifastacus leniusculus; Dana) is one of the most prevalent non-native decapod species globally (Kouba et al. 2015), including in UK freshwaters (Holdich et al. 2014), and was therefore chosen as the model predator organism.
We address the following research questions: (1) Does the presence of G. pulex and P. leniusculus (alone and in combination) modify fine sediment infiltration rates? (2) Does G. pulex display predator avoidance behavior in the presence of P. leniusculus in the form of vertical movement into subsurface sediments? (3) Does elevated sediment ingress into subsurface habitats result in reduced survivorship of G. pulex due to predation? Each of these questions were examined across a gradient of infiltration scenarios, as defined by sediment loading and grain size, thereby providing different environmental conditions in which to examine the interaction of predator-prey relationships and zoogeomorphic activity. Fine sand treatments were used to assess the role of pore infilling and therefore reduction in refuge volume, while coarse sand allowed the assessment of surface-pore bridging and therefore a reduction in refuge access. Differentiation between these processes is ecologically important as pore space may still be available but inaccessible should coarse sand bridge pathways, prohibiting the transfer of resources and organisms below the level of clog development. In contrast, a reduction in pore volume renders the interstitial space unusable regardless of the grain size present (Mathers et al. 2019).

Experimental setup
Experiments were undertaken in three identical sediment columns (following Mathers et al. 2014) comprising two interlocking substrate sections representing surface (benthic) and subsurface (hyporheic) substrates (Fig. 2). Each section was 32 cm in diameter and contained a 50 mm thickness of coarse fluvial sediment (gravel particles 20-64 mm in diameter). This size distribution is consistent with natural coarse river framework gravels and included particles that Pacifastacus leniusculus are known to displace (up to 38 mm in diameter; Johnson et al. 2010). The two sections were stacked vertically to provide a total substrate depth of 100 mm. Six mm holes (at a density of 0.06 cm À1 ) allowed water, fine sediment and Gammarus pulex to move between the sections. The bottom section (50-100 mm) was perforated with smaller holes (2 mm diameter at a density of 0.09 cm À1 ) to prevent emigration of G. pulex and limit fine sediment loss from the base of the column, while also permitting vertical hydrological exchange. In addition, 0.25mm netting was secured around the base of the column and a 5-mm rubber seal created around the base of the top section. The column was covered to prevent movement of crayfish out of the column. Experiments were conducted under natural ambient light conditions as G. pulex are phototactic and display negative migration behavior in response to light (movement away from the light source; MacNeil et al. 1999) and P. leniusculus activity is strongly diurnal with most activity occurring at night (Guan andWiles 1998, Nystr€ om 2005).
Downwelling flow conditions were employed during all experiments. Previous experiments using the mesocosm facility have documented the affinity of G. pulex for surface substrates under downwelling hydrological exchange (Mathers et al. 2014). Application of downwelling flow conditions therefore provided a baseline distribution of G. pulex for this set of experiments and facilitated the detection of avoidance behavior as increased occupancy of subsurface substrates. The sediment columns were placed inside separate cylindrical water containers (97 9 57 cm, volume = 210 L). External pumps delivered flowing water to the columns which passed through the column under gravity. A sprinkler rosette was attached to the end of the pump outlet to disperse water (2.7-2.8 L/min). Preliminary observations indicated that this flow of water was sufficient to maintain low interstitial flow through the sediments but was not great enough to initiate sediment transport. Consequently, any vertical movement of fine sediment during the experimental period was primarily a function of gravity or the direct activity of G. pulex and/or P. leniusculus.
The experimental containers were aerated throughout the experiments using an aquarium pump and temperature was held constant (15 AE 0.4°C) via an external water cooler (Aqua Medic, Titan 150, Bissendorf, Germany). The ❖ www.esajournals.org temperature selected for the experiments corresponds with peaks in crayfish activity under field conditions in the summer months within the UK (Bubb et al. 2004) and is comparable to that employed during other laboratory studies (Basil and Sandeman 2000). Experiments were undertaken during late spring and summer (May-July) to coincide with the main period of crayfish activity in the Northern Hemisphere.
The fine sediments used in the experiment consisted of two prewashed fluvial size fractions: fine sand (0.125 lm-1 mm) and coarse sand (1-4 mm). Finer fractions (<0.125 lm) were removed by wet sieving to ensure that turbidity did not vary between experimental trials. Prior to each experimental run, fine sediment was applied evenly to the surface of wet gravel in the top section of the column using a 1 or 4 mm sieve, respectively. Preliminary tests indicated that the application of 5 kg/m 2 filled all interstices (100% of pore framework volume) under the fine sediment treatment and covered the surface of all gravel particles. In addition to this heavy sediment loading, a moderate sediment loading of 3 kg/m 2 was used. The two size fractions were chosen to include grains with a low propensity to clog interstitial spaces (0.125 lm-1 mm) and grains with a high propensity to bridge between framework clasts and thus limit further infiltration (1-4 mm). Appropriate grain sizes were determined using calculations based on studies by Gibson et al. (2009b) and Frings et al. (2008), who provide ratios to discriminate between pore filling loads and bed structure (framework) clasts. For each experimental trial, a mixture of both fine sand fractions (equivalent to Fig. 2. Conceptual diagram of the experimental setup consisting of a coarse surface layer and finer subsurface layer. The red arrow represents the migration pathways available to Gammarus pulex, and the black arrow represents the flow of downwelling water and direction of sediment transport. Photographs illustrate the grain size matrix prior to sediment addition. 2 kg/m 2 of each size fraction) was mixed thoroughly with the gravel matrix and placed in the bottom section. This poorly sorted grain size mixture acted as sediment trap but did not reduce interstitial space sufficiently to preclude G. pulex from migrating into the subsurface layer.
Five sediment treatments were examined: (1) an open gravel framework without application of fine sediment; (2) 3 kg/m 2 fine sand sedimentation in the surface section; (3) 5 kg/m 2 fine sand sedimentation in the surface section; (4) 3 kg/m 2 coarse sand sedimentation in the surface section and; (5) 5 kg/m 2 coarse sand sedimentation in the surface section. These treatments represented a gradient of fine sediment loading (Appendix S1). Each treatment was undertaken for four different scenarios: (1) no organisms present; (2) 75 G. pulex; (3) one P. leniusculus; and (4) one P. leniusculus and 75 G. pulex. The sediment treatments (n = 5) and organism presence/absence (n = 4) were combined in a full factorial design giving 20 treatment combinations. Each combination was replicated five times to give a total of 100 experimental runs. Treatments were randomly allocated to an experimental trial.

Animals used in experiments
All crayfish were collected from a local stream (Wood Brook, Loughborough, UK; 52°75 0 69″ N., À1°22 0 74″ W.) using baited traps and immediately transported to the laboratory. To limit variability that might be associated with differing size and age, only medium-sized individuals with a carapace length of 40 AE 5 mm were selected. Selected individuals did not display any obvious injury (such as damaged carapace or loss of chelae, legs, or antennae), or regenerating chelae which might have affected their foraging behavior (Basil andSandeman 2000, Koch et al. 2006). The sex was recorded, although males and females have been documented as exhibiting no significant differences in behavior (Guan 1994), so this influence was not considered in experiments. Only intermolt individuals were used in the experiments (Kuhlmann et al. 2008) because activity and feeding behavior is known to be modified during ecdysis (molting; Reynolds 2002).
Each crayfish was housed individually between experiments and lettuce was provided ad libitum and supplemented with crayfish pellets (Tetra: TetraCrusa Menu) every other day when not involved in experiments. Preliminary tests indicated that when P. leniusculus were not fed in the days prior to experiments, insufficient numbers of G. pulex survived the experimental trials to enable analysis. All G. pulex specimens were collected from a local stream (Burleigh Brook, Loughborough, UK; 52°76 0 09″ N., À1°24 0 58″ W) where they occurred at high abundances (>100 individuals per m 2 ) using a standard pond net (mesh size, 1 mm) prior to each experimental trial. Individuals used in the experiments consisted of mixed size classes; 1-10 mm length.

Experimental procedure
For experiments in which no organisms were present (control application experiments), sediment was applied to the surface of the top section and left for 24 h. Experiments with P. leniusculus present were initiated in the same manner but with the addition of one crayfish on the surface section immediately after the application of the sediment treatment. For experimental trials which included only G. pulex, seventy-five individuals were released onto the top section of the columns and left for 24 h to redistribute themselves. Preliminary experiments indicated that this was a sufficient number for appropriate survival rates at the termination of the experiments to enable the detection of an avoidance behavior if present. In experimental trials where P. leniusculus and G. pulex were present at the same time, all G. pulex individuals were placed in the experimental facility and left to acclimatize for an hour prior to the addition of crayfish. A single pre-conditioned (soaked in water) horse chestnut (Aesculus hippocastanum) leaf was placed in both sections of the sediment columns as a food source for the G. pulex to reduce intraspecific predation (Dick 1995). A slice of carrot (~10 g) and crayfish pellets were provided as alternative food sources for P. leniusculus (following Bubb et al. 2002, Kuhlmann et al. 2008, to avoid excessive predation due to the absence of an alternate sedentary food source (Dorn 2013, Ruokonen et al. 2014. Shelter, in the form of an open-ended cylinder (110 9 100 mm), was provided for the crayfish in order to reduce pit digging behavior triggered by the absence of refuge (Johnson et al. 2010). This ❖ www.esajournals.org allowed the direct effect that foraging activities had on the ingress of fine sediment to be identified. New G. pulex specimens were used for each experimental trial. One crayfish was used per experimental trial, and each individual was used once for each treatment.
At the end of each experimental run (24 h), G. pulex individuals were collected and counted from each of the two substrate sections by washing the contents of each section through 4-mm sieves. All fine sediment was removed from the column and subsequently oven dried at 60°C until a constant weight was recorded to determine mass per section. For the subsurface section, the mass of fine sediment initially placed in the section was subtracted from the total fine sediment mass to calculate the fraction which had infiltrated down into the subsurface. These masses were converted to infiltration rates (kgÁm À2 Áday À1 ) as a measure of the amount of sediment mobilized.

Statistical analysis
Differences in infiltration rates between organism combinations for each sediment treatment were examined via a linear model using the function lm in the stats package in R version 3.12 (R Development Core Team 2013). Differences between all experimental treatments were tested using a Tukey post-hoc test via the glht function in the multcomp package (Hothorn et al. 2012).
Differences in the abundance of G. pulex in the subsurface as a function of sediment treatment, P. leniusculus presence, and the interaction of the two factors were examined via a linear model and tested using a Tukey post-hoc test. To assess differences in the vertical distribution of G. pulex (associated with avoidance behavior) within each sediment treatment and each organism combination (P. leniusculus presence or absence), a linear mixed effects model was employed with treatment specified as a fixed factor and substrate section nested within the experimental replicate (column) as a random factor (reflecting the fact that sections within individual columns were not independent). Post-hoc tests were conducted using a Tukey post-hoc test to determine the effect of the different sediment loadings on G. pulex movement patterns with and without P. leniusculus. Survivorship of G. pulex (number of individuals retrieved at the termination of the experimental trial) in the presence and absence of P. leniusculus and as a function of the sediment treatment was examined within a general linear model using the glm function in the stats package. A Poisson error distribution and log link structure were fitted to account for non-normal residuals.

Sediment infiltration rates
3 kg/m 2 fine sand sedimentation.-Infiltration rates of fine sand into the subsurface layer were greatest in experiments with P. leniusculus present (mean 1.93 kg/m 2 AE SEM 0.10, 63.3% of the sediment initially applied on the surface of the upper layer). These rates were significantly greater (P = 0.026 Tukey) than under control conditions with no organisms present (1.61 kg/m 2 AE SEM 0.06, 53.6%; Fig. 3A).
5 kg/m 2 fine sand sedimentation.-Fine sand sedimentation experiments with P. leniusculus present had the greatest infiltration rates (mean 3.06 kg/m 2 AE SEM 0.37, 61% of initial sediment, Fig. 3B); however, none of the pairwise comparisons were statistically different (Table 1).
3 kg/m 2 coarse sand sedimentation.-No significant differences between infiltration rates with or without organisms were determined (Table 2). All organism treatments displayed similar low sediment infiltration rates with surface clogging being evident (range 0.095-0.111 kg/m 2 , 3.1-3.7% of initial sediment application Fig. 3C). 5 kg/m 2 coarse sand sedimentation.-Infiltration of sand into the subsurface was greatest when both P. leniusculus and G. pulex were present, and the rate was statistically greater than for control conditions (P < 0.001 Tukey), G. pulex (P < 0.001 Tukey), and for an individual crayfish (P = 0.009 Tukey; Table 1). However, infiltration rates for all organism treatments were low when compared to fine sand treatments (range 0.082-0.286 kg/m 2 , representing only 1.6-5.7% of the initial sediment application; Fig. 3D).
Vertical migration of G. pulex in response to P. leniusculus presence and fine sediment treatment The distribution of G. pulex between layers was dependent on the presence of P. leniusculus (P < 0.001) and fine sediment treatment ❖ www.esajournals.org (P < 0.001) but did not vary as a function of the interaction of these factors (P = 0.247; Table 2; LMM).
Clean gravel. -In the absence of P. leniusculus, the majority of G. pulex remained in the surface layer, but in the presence of P. leniusculus, G. pulex were more equally distributed between the surface and subsurface layers (Fig. 4). Statistical differences were apparent in the number of G. pulex recorded in the surface or subsurface layers for treatments when P. leniusculus were absent with a greater number of G. pulex recorded in the surface layer (t 1,8 = À6.770, P = <0.001; Table 3). The presence of P. leniusculus resulted in no differences between the layers.
3 kg/m 2 fine sand sedimentation.-The presence of P. leniusculus resulted in a greater number of G. pulex migrating into the subsurface (Fig. 4). In the absence of P. leniusculus, 25% of individuals were located in the subsurface layer at the end of the experiment, but when crayfish were present, this proportion increased significantly to 50%. During experiments without P. leniusculus, there were significant differences in the vertical distribution of G. pulex (t 1,8 = À6.856, P = <0.001), but no differences were evident when P. leniusculus were present (P > 0.05; Table 3). 5 kg/m 2 fine sand sedimentation.-There were no significant differences in surface and subsurface abundances of G. pulex when crayfish were present (P > 0.05), but when P. leniusculus were Fig. 3. Mean infiltration rates (kgÁm À2 Áday À1 AE 1 SE) for each biotic treatment: (A) 3 kg/m 2 fine sedimentation; (B) 5 kg/m 2 fine sedimentation; (C) 3 kg/m 2 coarse sedimentation; and (D) 5 kg/m 2 coarse sedimentation. Treatments where infiltration rates were not significantly different are denoted with the same letter (Tukey post-hoc test P < 0.05). Note the order of magnitude reduction in vertical (infiltration rate) scale between fine (upper row) and coarse (lower row) sediment treatments. absent the surface layer contained a greater number of individuals (t 1,8 = À5.815, P = <0.001; Table 3; Fig. 4).
3 kg/m 2 coarse sand sedimentation.-Unlike the control and fine sand treatments, when coarse sand was applied the majority of amphipods remained in the surface layer whether or not P. leniusculus was present (Fig. 4). The number of G. pulex in surface and subsurface layers was significantly different in experiments with (t 1,8 = À11.109, P = 0.010) and without P. leniusculus (t 1,8 = 4.197, P = <0.001; Table 3). 5 kg/m 2 coarse sand sedimentation.-The greater loading of coarse sand was associated with the highest counts of G. pulex in the surface layer irrespective of P. leniusculus presence ( Fig. 4; Table 3). The surface layer contained significantly greater numbers of G. pulex than the subsurface layer with (t 1,8 = À4.587, P = <0.001) and without crayfish (t 1,8 = À4.587, P = <0.001; Table 3).
Between sediment treatment comparisons.-Overall, there were significant differences in the number of G. pulex in subsurface substrates when comparisons between sediment treatments, with or without P. leniusculus (Table 4), were considered. In the absence of P. leniusculus, only the addition of the greatest loading of coarse sand (5 kg/m 2 ) resulted in significant differences in the distribution of G. pulex compared to 3 kg/m 2 fine sand (P = 0.022) and 5 kg/m 2 fine sand (P = 0.005; Table 4). In contrast, in the presence of P. leniusculus, the addition of 3 and 5 kg/m 2 coarse sand treatments resulted in reduced numbers of G. pulex in the subsurface layer compared to clean gravel (P = 0.030 and P < 0.001, respectively), 3 kg/m 2 fine sand (P = 0.006 and P < 0.001, respectively), and 5 kg/m 2 fine sand (both instances P = <0.001; Table 4).

Survivorship to crayfish presence as a function of sediment load
Survivorship rates of G. pulex averaged 80% for all experiments but were highly variable as a function of crayfish presence/absence (Fig. 5). Experiments conducted in the absence of P. leniusculus had a mean survivorship of   Fig. 5). The addition of 3 kg/m 2 of coarse sand and 3 kg/m 2 fine sand in the presence of P. leniusculus resulted in significantly lower survivorship than both coarse sand treatments in the absence of crayfish (all P < 0.05; Table 6, Fig. 5).

DISCUSSION
This research sought to examine the behavioral strategies prey may use to evade predation, the engineering of a shared physical environment by a predator and its prey, and the impact that modifications to the prey's habitat as a result of biotic engineering and anthropogenic sediment loading may have for predator-prey interactions. In this study, freshwater amphipods utilized vertical avoidance behavior by actively moving into subsurface substrates to evade crayfish predation. However, the deposition of coarse fine sand clogged the surface layers of the substrate and reduced the effectiveness of this behavioral strategy resulting in increased predation (Fig. 6). When considering the zoogeomorphic potential of taxa, crayfish influenced fine sand ingress but predator-prey interactions themselves were found to be a primary mediating factor, with the availability of prey resources controlling the foraging activity of crayfish and subsequently their direct effect on the sedimentology of the physical environment (Fig. 6). The extent of this effect was, however, dependent on differences in the amount and grain size of the sediment loads applied. In   Tables 3 and 4. the current experiments, large pore spaces prevented crayfish-initiated infiltration of sand from reducing refuge habitat availability to the point that prey capture was more likely than for clean gravels. However, within a mixed, poorly sorted framework it is likely that crayfish-augmented infiltration may have the potential to reduce interstitial refuge availability and as a result prey avoidance behavior (Fig. 6). These ideas are captured in Fig. 6, which develops predator-prey-environment interactions (Fig. 1) into a more complex model in which the fuller sets of interactions are recognized (Fig. 6). This conceptual model emphasizes that predatorprey interactions should not be studied in isolation from the physical environment, and that the role of zoogeomorphic activity should be considered in the context of resource availability and other biotic drivers. The results also highlight the tantalizing prospect that predators may increase foraging success by inadvertently engineering changes in the environment (crayfish may increase fine sediment infiltration sufficiently to reduce available pore space), which requires further investigation in future studies. Each of the research questions we sought to address will be considered in the following sections where we examine the complex and threeway relationship between predator-prey and the physical environment and the critical role that external factors may play.

Impact of organisms on fine sediment infiltration rates and the role of predator-prey relationships
Many studies have examined the ingress of fine sands into gravel beds (e.g., Beschta and Jackson 1979, Frostick et al. 1984, Wooster et al. 2008, Franssen et al. 2014), but to date none have     Fig. 6. Conceptualization of the work conducted in this study examining the interactions between predatorprey relations, the physical environment, and the addition of an external stressor building on the traditional concepts presented in Fig. 1. Specific examples in relation to fine sediment dynamics, Pacifastacus leniusculus and Gammarus pulex, are provided. Dashed arrows indicate processes which require further investigation to fully understand the implications for predator-prey-environment dynamics. examined how this process is affected by the zoogeomorphic activity of a single organism or the interaction(s) among multiple organisms. This study provides direct evidence that the presence of P. leniusculus alone and of G. pulex and P. leniusculus together can increase infiltration rates depending on the characteristics of the sediment involved. For fine sand treatments (grain size 0.125 lm-1 mm), the greatest geomorphic effect was associated with crayfish foraging and resulted in~10% increase in fine sediment infiltration into the bed compared to control conditions (no organisms present) or when 75 G. pulex were present. The significant effect of a single crayfish suggests that under field conditions, and where interstitial spaces in the subsurface exist, sediment infiltration may occur at greater rates if crayfish densities are high. Given that P. leniusculus can reach densities up to 15 m À2 (Guan and Wiles 1996), the presence of this organism may significantly modify fine sediment movements between river bed surface and subsurface layers. No significant difference in infiltration rates was detected for G. pulex when in isolation, most likely associated with their relatively small body size (Moore 2006).
When both predator and prey were present (P. leniusculus and 75 G. pulex), sediment infiltration rates were on average 5% higher than control conditions or when only 75 G. pulex were present. It is likely that infiltration rates were reduced during these experiments because of direct preypredator interactions that affected the zoogeomorphic behavior of prey and predator. First, for experiments without crayfish, the majority of G. pulex were in the surface layer and therefore had a greater opportunity to cause ingress of fine sediment over the course of the experiment. In contrast, in the presence of P. leniusculus, on average 25% more G. pulex migrated into subsurface habitats, and consequently, the sum influence on ingress was reduced. Second, in experiments where P. leniusculus were the only organism present, a large proportion of energy was expended in foraging for food. In a number of experiments, P. leniusculus exhibited bulldozing behavior by piling and moving the substrate (Helms andCreed 2005, Johnson et al. 2010). This foraging behavior and the disturbance of surface sediments significantly affected the vertical movement of finegrained sediments into the bed. The addition of prey (G. pulex) resulted in a reduction in foraging activity due to the greater availability and number of encounters with food resources. It is assumed that during the experiments foraging activity would gradually decline over the 24-h period as prey were consumed (Haddaway et al. 2014) and time between foraging activity would have increased. The overall effect was a reduction in bed disturbance caused by foraging, and therefore, a reduction in fine sediment ingress compared with that when P. leniusculus were alone. Prey availability may therefore be a key driver of zoogeomorphic activity. Reductions in prey availability may increase foraging behavior and interspecies competition (especially for P. leniusculus which exhibits a high degree of intraspecific aggression; Pintor et al. 2008); this in turn may enhance fine sediment mobilization (suspension and ingress). These results support findings of other studies on predaceous stoneflies, which surmise that prey scarcity increases predator movement and thus the stability of fine sediments (Statzner et al. 1996, Zanetell andPeckarsky 1996).
In contrast to fine sand deposition, the greater propensity of coarse sand to form a bridge between clasts and thus prevent further infiltration meant that biota had little effect on overall infiltration rates. Under 3 kg/m 2 loadings, no significant differences in infiltration rates were recorded for any of the organism combinations, with only 3-4% (0.1 kg/m 2 ) of the initial sediment application infiltrating into the subsurface. Application of 5 kg/m 2 coarse sand resulted in limited differences in infiltration rates when animals were present, and although significant, the total mass of sediment was low compared to fine sand treatments (mean 0.14 kg/m 2 ingress). For coarse sand sediment treatments, P. leniusculus generated significantly greater infiltration rates (an extra 0.1 g/m 2 on average) compared to control or 75 G. pulex treatments. However, in contrast to fine sand applications, the combination of both G. pulex and P. leniusculus resulted in the greatest infiltration rates. With clogging of the surface layers of the substrate by coarse sands (Mathers et al. 2019), G. pulex were unable to migrate into the subsurface, and therefore, around 95% of individuals remained in the surface layer. This had a significant effect on infiltration rates in the coarse sediment experiment but at an order of magnitude lower than for the fine sand treatments.

Predator avoidance and the interaction with the physical environment
The application of fine sediment (grain size 0.125 lm-1 mm) under both loadings (3 and 5 kg/m 2 ) resulted in high infiltration rates under control conditions (gravity and downwelling flow) and all organism combinations. However, vertical connectivity was maintained due to the large interstitial spaces between gravel particles relative to the sand grain size (Xu et al. 2012, Mathers et al. 2014. As a result, around 50% of G. pulex were able to migrate vertically into the subsurface when predatory crayfish were present. In the equivalent treatments without the predator, the majority of prey remained in the surface layer, probably reflecting the rheophilic preferences of G. pulex (Gledhill et al. 1993). The results therefore indicate that the presence of the crayfish predator significantly modified the vertical distribution of G. pulex and provides evidence of predator avoidance behavior.
In contrast, the addition of coarse sand sediments (1-4 mm) resulted in bridging of interstitial spaces within the surface layer of the substrates and clogging that disconnected the surface and subsurface layers regardless of zoogeomorphic activity. Infiltration rates were significantly lower for coarse sand than for fine sand treatments, with the majority of sand particles being retained in the surface layer for both organism combinations. As a result, no significant difference in the vertical distribution of G. pulex (with or without P. leniusculus) was observed. The formation of surface-layer clogs probably restricted the ability of G. pulex to migrate into subsurface sediments (sensu Mathers et al. 2019) and influenced predatorprey relationships, in this instance through survivorship of G. pulex (Figs. 1B, 6). In experiments where G. pulex could easily migrate, survivorship averaged 77% (AE SEM 4.5) with no significant differences evident for any of the fine sand sediment or control treatments. The addition of coarser sand clogged the substrate interstices and prevented vertical avoidance behavior resulting in lower survivorship rates. 3 kg/m 2 coarse sand resulted in moderately lower survivorship (70% AE SEM 5.1) and 5 kg/m 2 reduced G. pulex survival to 56% (AE SEM 4.5); a reduction of 28% compared to the control treatment (open gravel framework).
Despite P. leniusculus being a significant zoogeomorphic agent within fine sand treatments, in this set of experiments only the coarse sand treatment affected the vertical avoidance behavior and subsequent predation rates. This result is predominately a function of the open gravel framework used in the experiments (Xu et al. 2012) and we therefore hypothesize that under natural conditions, where poorly sorted mixed gravel frameworks exist, ingress of sand augmented by P. leniusculus may influence pore space sufficiently to affect avoidance behavior in many instances. If this is the case, P. leniusculus may inadvertently improve their predatory success by increasing fine sediment content within the river bed, an example of extended phenotype engineering, whereby the organism creates structures or effects that directly influence their fitness and survival (Jones et al. 1994, Wright andJones 2006). This hypothesis requires further detailed investigation to fully understand the feedbacks between predator-prey and the physical environment (sensu Fig. 6).

Mediation of predator-prey interaction by sediment characteristics
This research illustrates how the structure of the physical environment potentially influences predator-prey interactions (Figs. 1B,6). In each experiment, the nature of the physical environment was a key control on these interactions and feedback, with different outcomes dependant on the amount and grain size of the fine sediment applied. The deposition of coarser sand grains has the potential to reduce the effectiveness of a prey's avoidance behavior and render them susceptible to enhanced predation. This suggests that in nature, spatial and temporal variations in fine sediment dynamics will partly regulate these interactions and further field experimentation is required to test this more rigorously. Sediment dynamics in rivers vary as a function of the abiotic hydro-geomorphological regime, but anthropogenic activities (agriculture, urbanization, forestry, construction, mining) increase sediment loading (Owens et al. 2005) and, in the UK, sediment recruitment to rivers may be further increased in the presence of P. leniusculus through burrowing activities (Faller et al. 2016. It is therefore possible that anthropogenic sedimentation has facilitated the success of invasive P. leniusculus by reducing the availability of hyporheic refugia for prey species.

CONCLUSIONS
The results of this study demonstrate that fine sediment ingress into gravel river beds can be caused by expenditure of biological energy under certain environmental conditions; that prey utilize avoidance strategies to evade predation; that predator-prey interactions themselves mediate the zoogeomorphic effectiveness of an organism and that these interactions and feedbacks are dependent on the environmental context. We present a new conceptual model that captures the interactions between predator, prey, zoogeomorphic processes, and habitat availability (Fig. 6) highlighting that interactions between ecosystem processes may be strongly mediated by dynamic bi-directional interactions between organisms and the physical environment they inhabit. Ecological (predation, avoidance) and geomorphological processes (sediment infiltration) are intrinsically linked and should not be studied in isolation.

ACKNOWLEDGMENTS
KLM acknowledges the support of a Glendonbrook doctoral studentship and co-funding from the Environment Agency to undertake this study. Thanks go to Stuart Ashby and Richard Harland for providing technical and laboratory support and to Nigel Wilby and Nicholas Clifford for useful discussions pertaining to the study outcomes. The comments of two anonymous reviewers have significantly improved the clarity of this study. abundance, behavior, and hydraulic energy shape