The complexities of measuring fine sediment accumulation within gravel‐bed rivers

Fine sediment storage within gravel beds is a key component of catchment sediment budgets and affects the health of benthic and hyporheic habitats. Here, we assess the performance of two substrate infiltration traps for the characterization of fine sediment (<2 mm) accumulation. One design, the vertically extending sediment trap, permits both lateral and vertical exchange in the sediment column, whereas the second type, a more traditional fixed‐area sediment trap with impermeable side walls, permits only vertical exchange. Traps were deployed at three sites on the River Tame, Birmingham (UK), over varying installation periods (14–401 days). Results indicate that the facilitation of multiple pathways of exchange within the vertically extending sediment traps (vertical and lateral) resulted in a significantly greater amount of fine sediment being accumulated than in adjacent fixed‐area sediment traps. This suggests that lateral transport is an important component contributing to fine sediment accumulation. However, there are notable and inherent problems associated with the use of different types of sediment trap and in the way the data should be presented and interpreted. This paper discusses the practical implications of the study findings and reflects on the complexities of undertaking accurate sediment deposition measurements in the field.

sediment deposition storage and the rates of accumulation are, therefore, important factors when considering land use management approaches.
Despite significant advancements in our understanding of sediment dynamic processes at fine resolutions (temporally and spatially) via flume experiments (e.g., Huston & Fox, 2015;Vericat, Batalla, & Gibbins, 2008), field studies of in situ channel bed sediment dynamics remain limited and those which incorporate varying timeframes are even rarer.
Disturbance/resuspension methodologies (Duerdoth et al., 2015;Lambert & Walling, 1988) can be used to assess channel storage of fine sediment at a specific time and place. This technique is useful and has been widely adopted in studies of sediment dynamics but provides little information about the depositional route or timing of the fine sediment.
Freeze coring (Evans & Wilcox, 2014;Walkotten, 1973Walkotten, , 1976 has also been used to determine between-site differences in sedimentation processes with reference to both flow regulation (Petts, 1988) and urbanization (Thoms, 1987a). However, although intensive freeze-core surveys can effectively detect spatial changes in fine sediment storage within riverbeds (Petts et al., 1989), bed fabrics can be disrupted during probe driving (Kondolf, Lisle, & Wolman, 2003) and the destructive nature of the technique tends to limit the approach to annual surveys.
The most common method for monitoring fine sediment deposition rates over a known time interval is bed traps (Carling, 1984;Curran, Waters, & Cannatelli, 2015;Reiser, Ramey, & Lambert, 1985). A large variety of trap designs have been employed, but the accuracy of the calculated fluxes for the different bed trap designs remains poorly understood (Petticrew, Krein, & Walling, 2007).
There are two main types of bed trap, the first of which are empty pit traps consisting of lidless, solid-walled containers set in, or upon, the channel bed (Kondolf et al., 2003;Tipping, Woof, & Clarke, 1993). Pit traps are effective for collecting fine sediment derived from the gravitational sedimentation of suspended sediment onto the surface of the channel bed as well as coarser material transported as bedload and provide useful information at event timescales. However, the artificial conditions they represent (i.e., no trapping of within-bed or intragravel transport) present a challenge for investigations into fine sediment storage and dynamics.
The second type of bed sediment trap is the substrate trap, which comprises a container that is filled with prescreened gravel (removing all material <2 mm; Tipping et al., 1993) and which seeks to represent the natural substratum framework as much as possible. Many studies have typically employed solid-walled containers, which permit vertical transport of fine sediment but limit the lateral intragravel transport of fine sediment, thereby potentially leading to the underestimation of ingress rates (Carling, 1984;Petticrew et al., 2007). In addition, the exclusion of lateral flushing flows may lead to the pore spaces in the upper layers of the framework becoming clogged, which can create a seal and prevent subsequent infiltration of sediment into the trap gravel (Frostick, Lucas, & Reid, 1984). A number of recent studies have incorporated semipermeable walls (e.g., , which are typically utilized in ecological studies and facilitate simultaneous collection of sedimentological and ecological samples; however, these designs are not fully permeable, and thus, the full extent of intragravel and vertical exchange processes in field settings remains limited. Traditional solid-walled or semipermeable designs are also subject to scour events, which may expose the container above the surface of the riverbed, potentially affecting trapping rates associated with turbulent flow alterations (Lisle & Eads, 1991).
Infiltration baskets consisting of a wire mesh basket dug into the riverbed and filled with prescreened gravels allow for the lateral infiltration of fine sediment (Milan & Large, 2014;Sear, 1993;Thoms, 1987b). As with the solid-walled containers, infiltration baskets are vulnerable to removal by scour during flood events (Sear, 1993) and the presence of the wire mesh may affect the particle size distribution of sediment ingress. Infiltration baskets are also likely to be susceptible to infiltrated sediment and water being lost upon retrieval. A modified version of this trap consists of an impermeable bag that is buried within the gravel substrate and pulled up via cables, thereby including all deposited sediment in the trap area (Lisle & Eads, 1991).
To explore the complex nature of fine bed sediment storage and infiltration, we employed two sediment trap designs: one, which permits vertical deposition only, and one with permeable sides, which allows both vertical deposition and lateral movement. This paper (a) describes the two trap designs; (b) compares the results of the two sampling methods over a range of installation periods (14-401 days); and (c) reflects on the relative importance of lateral and vertical movement of fine sediment in gravel substratum and the complexities of undertaking accurate measurements in the field.

| Sampler design
The study utilized two versions of bed trap design: a basket-type bed trap (vertically extending sediment trap [VEST]) and a more traditional bed trap design (fixed-area sediment trap [FAST]) of identical size and shape ( Figure S1). The VEST is similar to the infiltration bag described by Lisle and Eads (1991), however, one modification was made to address bag slippage problems during recovery of the samples (Petticrew pers. comm), which can lead to problems with calculating the volume of sampled gravels. The modified design here was made from collapsible and impermeable ventilation tubing in place of a bag, which was sprung with stainless steel wire (collapsed length 12 cm; maximum extendable length 30 cm; internal diameter 20 cm; cross section area 314 cm 2 ; maximum volume 9,425 cm 3 ). The trap was reinforced around the top with a stainless-steel rim and attached to a stainless-steel base with two flexible woven webbing handles fastened to the reinforcing ring. The highly flexible material allowed the walls of the trap to collapse down within the gravel bed during the sampling period, thus minimizing the physical effect of the trap on the natural particle exchange processes in the riverbed, while  Table 1.
The water column depth at each installation location was approximately 10-30 cm under low-flow conditions, to ensure traps were permanently inundated and to facilitate safe installation and recovery.
Traps were filled with gravels excavated from the installation pits to provide a site-specific and natural gravel framework with which to assess infiltration rates. Prior to installation, gravels were wet sieved in situ to remove fines (<2 mm), and any extremely large or artificial clasts (e.g., large house bricks that would not fit within the diameter of the trap) were removed. This approach standardized the framework across the three sites as much as possible and avoided significant variation in the available pore space between individual traps.
Traps were removed by inserting a plastic tube (collar) around the top of the trap to a level above the height of the water column. This prevented the flow from removing accumulated fines from the surface and stabilized the hole thereby facilitating the reinstallation process.
All gravels and fine material were emptied into large field-portable sieves, incorporating 4 × 25-L storage tanks, and sieved using river water with fines retained in the tanks. To calibrate for the inclusion of suspended sediment in the water column, a sample of river water was taken just before recovery to enable the total volume of trapped sediment to be adjusted relative to suspended sediment concentrations within the river. A lack of data for a 6-month sample at James Bridge was caused by a loss of the FAST sediment trap due to deep scour during a major flood. The downstream pair of traps at any given site was most frequently sampled in order to minimize disturbance (trampling) of upstream traps.
Only a limited number of samplers were available, and so, traps were reused (removed, cleaned, and reinstalled). Because of the significant effort required for installation and removal of the traps, replication of all sites was not feasible and, therefore, sediment accumulation comparisons are made using paired traps from all three sites, and is accounted for in the statistical model design; see below.
Safety issues associated with access to an extremely flashy urban river system (Lawler, Petts, Foster, & Harper, 2006) led to some irregularity in deployment times. Each trap generated~80-100 L of water for rinsing and cleaning, and this large volume further limited how many traps could be changed in one sampling event because the total volume was transported back to the laboratory for further analysis.

| Laboratory methods and data analysis
In the laboratory, sediment samples were allowed to settle for a minimum of 24 hrs. The supernatant was then decanted and fines <2 mm oven dried for a further 24 hr before dry-sieving to obtain the following particle size classes: 600-2,000, 125-600, 63-125, and <63 μm. Total sediment accumulated (kg) was determined for pairs of VEST and FAST traps installed for different residence times (range of 14-401 days; Table 1) over a 13-month period (n = 27). As traps from differing sites were used in the comparison of the mass of fine sediment accumulating FIGURE 3 Schematic diagram illustrating the design, installation, and sampling techniques for the vertically extending sediment trap (VEST) and fixed-area sediment trap (FAST) sediment traps. VEST traps are lowered into an excavated hole protected by a plastic tube extending above river water level (a) with straps extending above the gravel bed (b). The hole is backfilled with cleaned gravel before removing the plastic tube to leave the gravel open to lateral and surface sediment ingress (c). The plastic tube is carefully placed over the trap before the straps are used to lift the VEST vertically through the gravels (d) in order to recover the sample. FAST installation (e-g) also uses the plastic tube to protect the trap during installation and removal (e-g), but the solid walls prevent lateral movement of fine sediment into the cleaned gravels over the installation period (f) in VEST and FAST traps and preliminary analysis indicated that sediment accumulation (kg) varied by site (all p < .05), mixed effects models were employed in subsequent analyses (Venables & Ripley, 2002).
Application of mixed effects models enables the incorporation of random factors, variables that account for excess variability in the dependent variable. The mass of sediment accumulated was tested via a linear mixed effects (LME) model with the fixed terms of "trap type × particle size × residence time (days)" and site specified as a random factor. Models were fitted using the "nlme" package (Pinheiro et al., 2013) with the restricted maximum likelihood estimation function in R version 3.3.2 (R Development Core Team, 2013). Both marginal R 2 (proportion of variance explained by the fixed factors alone) and conditional R 2 (proportion of variance explained by the fixed and random factors) values were extracted using the "MuMIn" package (Bartoń, 2016).
Significant categorical interactions (i.e., particle size and trap type) were investigated further via pairwise comparisons to enable examination of where statistical differences occurred. An LME was also employed to assess whether the rate of sediment accumulation varied as a function of trap residence time (days installed) with site specified as a random factor. Accumulation rates were calculated as the mass of fine sediment filtrating substrates via the surface of the sediment traps, standardized to m 2 /day as per other studies (Collins & Walling, 2007;Frostick et al., 1984). To assess whether the proportion of fine sediment in each grain size varied as a function of trap type, a binomial general linear model (GLM) was fitted using a logit error distribution via the "glm" function.
Preliminarily analysis via a generalized LME model (GLMM), with site fitted as a random term, indicated that the random effect accounted for little variation, and so, the simplified GLM was employed. Table 1 summarizes the period of trap installation, the mass of sediment accumulated per trap, and the calculated rate of sediment accumulation in kg/m 2 /day for the complete paired dataset. Sediment accumulation varied significantly as a function of trap type with VEST traps collecting 46.16% more fine sediment than the FAST traps (Figure 4). Statistical differences in the mass of fine sediment accumulated were also evident as a function of the independent effect of residence time and particle size and the interaction of the two factors (Table 2). There was also significant interaction between trap type and particle size, indicating that the trap design influenced the total accumulation of certain particle sizes (see below). Examination of the R 2 values indicated that the model accounted for a good proportion of the variability, with fixed factors explaining 49.33% of variance and the random factor of site accounting for an additional 4.7% of the total variance (total model 54%). A summary of model outputs (F, p, and R 2 values) is shown in Table 2. When only vertical surface area was considered, the rate of fine sediment accumulation declined significantly as trap residence time increased (F 1,47 = 4.9 p = .032; Figure 6); however, R 2 values indicated a poor fit of the model predictors (marginal R 2 = 8.62%, conditional R 2 = 11.37%).

| RESULTS
Pairwise comparisons of the individual particle sizes by trap type indicated that significantly more large particles in the size fractions of 600-2,000 and 125-600 μm were collected in the VEST traps in comparison to the FAST traps ( Figure 5; Table 3). Smaller particles (63-125 and <63 μm) demonstrated no statistically significant differences associated with trap type (p > .5; Table 3). This is a reflection of the larger mass of sediment collected by VESTs as no significant differences were determined for the proportion of mass within each size fraction as a function of trap type (p > .05; GLM), with the model accounting for only a small amount of variability (R 2 = 10.85%; Figure S2).  Petticrew et al., 2007). This finding supports a number of other studies that have compared various trap designs-those that permit just vertical exchange and those that permit both lateral and vertical exchange. Within a flume study, Carling (1984), for example, determined that solid-walled traps reduced the trapping efficiency of fines by up to 31%. Sear (1993) and Mathers and Wood (2016) also reported similar findings within field studies with reductions in trapping efficiency of 20-25% and 29%, respectively.
Observations made when excavating the traps showed that the upper layer (~5 cm) of gravel in the FASTs quickly became blocked with fine sediment forming a seal (e.g., Herzig, Leclerc, & Goff, 1970;Huston & Fox, 2015), and fine sediment penetrating less deeply within the FAST traps (to a depth of~15 cm) compared to the VESTs (~20 cm). This suggests that lateral flushing flows were not possible within the FASTs and this may in part cause the build-up of sediment (and formation of a seal/colmation) in the surface layers of the FASTs.
This prevents further sediment accumulation and possibly leads to overestimates of any net lateral sediment accumulation in this study.
This observation highlights the importance of vertical and lateral connectivity in maintaining healthy hyporheic habitats. Rivers, which are dominated by the process of vertical fine sediment ingress, may suffer more readily from the formation of fine sediment surface clogs.
Colmation can significantly reduce vertical hydrological exchange, limiting the transfer of oxygen and nutrients, and effectively disconnects surface substrates from subsubsurface habitats, thereby potentially reducing stream biodiversity (Bo, Fenoglio, Malacarne, Pessino, & Sgariboldi, 2007;Mathers, Millett, Robertson, Stubbington, & Wood, 2014;Simpson & Meixner, 2012). VEST traps collected a greater mass of larger size fractions, which may be a function of vertical hydrological exchange being maintained, allowing ingress of coarser fines deeper into the substrates. However, there were no differences in the proportional composition of the fine sediment matrix with both trap types representing comparable particle size distributions.

| Complexities of sediment accumulation measurements associated with time
Within this study, both the vertically extending (permeable) and fixed area (impermeable) traps exhibited rapid filling in the short term ( Figure 6). This may be explained by the presence of large void spaces between the cleaned river gravels, which provided maximum connectivity and a high storage volume in which fine sediment could accumulate (Xu, Wang, Pan, & Na, 2012). Hoyal et al. (1997) found that the greatest rates of deposition were in clean gravel beds and that a reduction in deposition occurred long before the bed was filled with sediment, due to an infilling of the surface void space. The rapid trap filling observed in this study is unlikely to be representative of natural sediment deposition rates in undisturbed river channels but may be an artefact of the experimental method used; during trap installation, a significant loss of Note. FAST = fixed-area sediment trap; VEST = vertically extending sediment trap.
FIGURE 5 Average mass of fine sediment (+1 SE) for grains 600-2,000, 125-600, 63-125, and <63 μm by trap design for all traps installed on the River Tame. Black = vertically extending sediment traps and white = fixedarea sediment traps bed framework structure, including any surface armouring, will occur.
Moreover, screening of gravel for use within sediment traps involves removing all material <2-mm diameter, which represents an unnatural matrix in gravel frameworks and can create a suck or draw for fine sediments. However, this problem is inherent with these methods as it is necessary to start with clean gravel (i.e., zero fine sediment) in order to quantify the mass of fine sediment, which has accumulated during the installation period.
Traps that were removed, sieved, and refilled with clean gravel more frequently (i.e., the traps installed for shorter time periods) collected more fine sediment than traps, which were resident for several months. This may be associated with renewed connectivity of substrates associated with vertical hydrological pathways, an increase in initial storage capacity and the removal of benthic algae (Potamogeton spp and Cladophora spp) that may inhibit sediment accumulation and ingress (Bo et al., 2007;Fox, Ford, Strom, Villarini, & Meehan, 2014;Papanicolaou, Diplas, Evaggelopoulos, & Fotopoulos, 2002). Hydraulic conductivity, and associated sediment transport, is also strongly linked to the time since the last streambed disturbance (Boano et al., 2007;Stewardson et al., 2016) and, as such, longer trap residence times are able to capture this natural decay in infiltration rates that may be a function of bed turnover (or in this instance trap removal). The results of the study, therefore, suggest that substrate traps installed for short periods may overestimate natural long-term rates of sediment accumulation within gravel frameworks and caution is needed when interpreting the data.

| Complexities of understanding the processes captured by sediment trap designs
Finally, we consider the accuracy of the two commonly employed sediment traps in the measurement of fine sediment accumulation as a function of the processes they quantify. Solid-walled traps permit only vertical exchange of fine sediment, and therefore, once fine material has infiltrated substrates within the trap, the only direction of transport is vertically through the top of the trap, which can occur through turbulent flows (Detert & Parker, 2010;Kuhnle, Wren, & Langendoen, 2016;) and, in particular, during large flood flows, which are capable of scour events (Harris, Whitehouse, & Moxon, 2016  Note. FAST = fixed-area sediment trap; VEST = vertically extending sediment trap.