Decision letter for "Assessing methods to improve benthic fish sampling in a stony headwater stream"

HandlingEditor: JohnMurray-Bligh Abstract 1. Electrofishing is a well-established and widely used method for surveying fish populations. Nonetheless, its effectiveness is impacted by numerous factors, including water chemistry, habitat type and fish species. Both physiological and behavioural responsesmake bottom-dwelling ‘benthic’ fish which lack swim bladders (e.g. European bullhead Cottus gobio) particularly difficult to survey by electrofishing. 2. We compare the performance and practicalities of electrofishing for benthic fish at a rocky northern English headwater stream with two sampling methods originally designed for crayfish surveys; the triple drawdownmethodwhich involves repeated dewatering of a site, and the Pritchard Trap method which involves sunken traps filled with natural substrate that samples a small, fixed (0.25m2) area of river bed. 3. Both the Pritchard trapping and triple drawdown methods provided similar highdensity population density estimates for bullhead which were at least 2.5–5 times higher than predicted from electrofishing derived sweep depletion curves. 4. Electrofishing and the triple drawdownmethodareboth resource-intensive, requiring expensive equipment and a team of trained operatives. These approaches also pose a risk to fish and non-target organisms. In contrast, Pritchard Traps provide a cost-effective passive, low risk survey method requiring minimal training and only one operative. Pritchard traps, therefore, show particular promise for benthic fish surveying andmonitoring.

Fish populations can be surveyed using a variety of methods, including netting (e.g. seine netting; Neilson & Johnson, 1983;Pierce et al., 1990), trapping (e.g. minnow traps, Bloom, 1976;Bryant, 2000) acoustic telemetry (Crossin et al., 2017) and electrofishing (Beaumont, 2016;Reynolds, 1996). Electrofishing, widely used in stream biological monitoring, involves applying an electric field in the water to temporarily incapacitate fish, allowing them to be caught (Beaumont, 2016). Many physical factors affect the efficiency of electrofishing, including water clarity, depth and conductivity, substrate type and fish species. Benthic fish are notoriously difficult to capture by electrofishing, owing to their relatively small body size, behaviour and preference for staying close to the riverbed. Some benthic fish show a poor electrotactic response (Beaumont, 2016;Cowx, 1983), with some taxa also lacking a swim bladder (e.g. species in the Cottidae), reducing their buoyancy and thus the effectiveness of the anodes' pull. Further limitations to electrofishing relate to benthic species being associated with structures like cobbles and boulders that partially shield them from electric fields rendering incapacitated animals inaccessible. Whilst electrofishing and other contemporary methods have proven suitable and effective in sampling many species in various freshwater systems, a strong need persists for new methodologies that generate reliable quantitative data on benthic fish populations.
In some instances, benthic invertebrate sampling techniques have been adapted to sample benthic fish, for example Hess samplers and Surber samplers to survey European bullhead Cottus gobio in English chalk streams (Harrison et al., 2005;Woodward et al., 2008). These benthic invertebrate survey methods proved successful at quantitatively sampling bullhead, chiefly due to their sedentary nature. Recent methodological advances in surveying freshwater crayfish also show potential promise for benthic fish survey. The habitat requirements of benthic fish and crayfish often overlap (Bubb et al., 2009;Ruokonen et al., 2014), and methods that successfully survey crayfish within benthic habitats could hence reasonably be expected to also catch benthic fish. In this study, we investigate whether two recently developed crayfish survey techniques are suitable for quantitative benthic fish population assessment. The triple drawdown (TDD) technique introduced by Chadwick et al. (2021) involves repeated draining and re-wetting of an isolated section of a watercourse, with hand-removal of available refuges and river biota. The sequential capture of specimens from target species also allows for depletion analyses (e.g. Carle & Strub, 1978), facilitating estimates of the total population present and of the efficiency of the method (see Chadwick et al., 2021, for further details).
The Pritchard Trap (PT) is a passive sampling trap, comprised of a collapsible mesh bag and quadrat (0.5 m × 0.5 m) set into the available substrate, which provides quantitative estimates of crayfish population demographics upon retrieval (≥4 days deployment time; see Pritchard et al., 2021).
In this study, we assess performance and practicalities of the two aforementioned crayfish survey methods against conventional electrofishing for surveying benthic fish, especially European bullhead C. gobio (hereafter 'bullhead'), in a rocky headwater stream in Northern England. We firstly hypothesize that both TDDs and PTs allow for quantitative assessments of benthic fish population densities. We F I G U R E 1 Site map showing study area, including the three study sites, Confluence, Footbridge and Farm along Long Preston Beck (LPB) further hypothesize that PTs require a minimum deployment time of 4 days to robustly survey benthic fish, based on the 4-day deployment required for crayfish. When comparing the ability of each method to produce robust demographic data on benthic fish populations, we hypothesize that electrofishing underestimates benthic fish population sizes when compared to both TDD and PT survey results.

Site description
The study was conducted at the upland headwater stream Long Pre-

Survey design
The study comprises two main components.  (Carle & Strub, 1978) in the FSA package (Ogle, 2018) in R Studio version 1.1.463 (RStudio Team, 2020). Total population estimates were generated, which allowed method efficiency to be calculated as the total number of fish caught as a percentage of the total estimated population. Density estimates were then calculated as the number of fish caught over the site area -and the expected density using the estimated total population over the site area.

Triple drawdowns
TDDs were undertaken at each site immediately after the electrofishing surveys on each isolated stretch of LPB (summer 2018; Table 1).
Given the large size of the dewatered river sections for the TDD of 45-50 m 2 , two Honda Trash pumps (2 and 3 inches), four sweeps and 6-10 operatives were required at each site. All other aspects of the TDD approach were consistent with Chadwick et al. (2021).
Multiple sweeps at each site allowed depletion analyses to be calculated using the same method as described above. This allowed for the generation of total population estimates, method efficiency and, in combination of site area measurements, fish density estimates for the TDD.

Pritchard Traps
Specifications and general operation of PTs followed the approach

Comparison of methods (electrofishing, TDD and PT)
The fish data generated through each method were compared to determine differences in estimated community species structure. Density estimates and population size structure of bullhead as the dominant benthic fish species in the system were also explored across all three methods. Additional PT sampling was undertaken in 2018 with a low sampling effort (n = 4) for density estimates at Footbridge and Farm F I G U R E 3 Photographs of a PT: (a) ready to be set into outlined area and filled with naturally occurring substrate collected into a bucket and (b) set into the riverbed prior to secondary sampling by electrofishing and TDD. Repeat PT sampling was also undertaken in summer 2019 (Table 1) to increase sample size (n = 30; 7.5 m 2 ) in order to enable robust comparisons of population demographics. In all these sampling events, PTs were deployed for a minimum of 4 days. An estimate of true fish density for the three sites was generated through summing all fish physically removed via electrofishing prior to the TDD and the total TDD-derived population estimate (Carle-Strub) for each site. Comparative analyses of the methods were carried out in SPSS (version 27) for statistical descriptors on demographic data. We used the ggplot 2 package (Wickham, 2016) in R (version 3.5.1) for the graphical representation of the population structures.
Capture efficiency was estimated at 84%, resulting in a total popula- Fish numbers generally stabilized after 2 days, but with high fluctuations at Confluence where the overall lowest density of fish was recorded. Furthermore, while a high density was recorded at Footbridge after only 1 day, subsequent data showed an increase in observed numbers with time, and peak density estimates for every site were only reached at the maximum exposure time interval of 10 days.

F I G U R E 5
Density of benthic fish (m −2 ) generated from PTs (2019, n = 3, 0.75 m 2 ) after deployment times of 1, 2, 4, 7 and 10 days. Error bars show deviation from average of minimum and maximum catch densities. Benthic fish detection rates (as the percentage of PTs containing any benthic fish specimens) are presented above each bar

Population structure and density
Three benthic fish species were sampled at the sites: bullhead, stone loach and European eel (hereafter eel). All sampling methods consistently found bullhead to be the most abundant species at all study sites.
Of the electrofishing-derived population, bullhead comprised 82.4-94.2% of the population and stone loach 5.8-17.6%. Electrofishing caught one eel at Farm (0.4% of the population). Of the individuals cap-tured during the TDDs, 81.8-97.5% were bullhead, 2.2-16.5% stone loach and 0.3-1.7% eel. Eel were sampled at all sites through the TDDs (four to six individuals/site). In the PT-derived benthic fish population based on repeat sampling in 2019, 76.3-86.1% of individuals represented bullhead, with the remaining 13.9-23.7% representing stone loach. No eels were captured by the PTs.
All methods detected a much lower density of fish at Confluence relative to Footbridge and Farm (Table 2). Overall, electrofishing generated density estimates that represented only ∼20% of the estimates generated by both TDDs and PTs at each respective site in the same year ( Table 2)

Performance of the methods
The TDD and PT techniques were both originally developed to generate quantitative data on crayfish populations. Evaluation of these methods in a rocky headwater stream nonetheless clearly shows their ability to generate valuable information on benthic fish population density and structure. Whilst the assumed capture efficiency generated through electrofishing data ranged from 45% to 84%, and the assumptions of the depletion analyses were satisfied, these two alternative techniques also confirmed that the total number of benthic fish actually available to be caught was severely underestimated by the electrofishing depletion curves. Indeed, subsequent TDDs revealed total population estimates to have 3.2-5.3 times more fish than were estimated to be present based on the electrofishing surveys. This would suggest that the reduction in fish captured during electrofishing sweeps was not due to a true reduction in fish present in the site via removal, but that the fish present became progressively less catchable. Fish catchability could be affected by a number of factors, for example behavioural responses to repeated electric shock, or the physical disturbance of prior sweeps causing fish to seek shelter. The fishable area of the site TA B L E 2 Benthic fish densities recorded from electrofishing, triple drawdowns (TDDs) and Pritchard Traps (PTs; n = 4 in 2018; n = 30 in 2019). Estimated totals result from adding the specimens caught by electrofishing to the estimates resulting from the TDDs on the same, isolated stretch of river

F I G U R E 6
Bean plot (i.e. probability density of the catch data) of bullhead size class distribution (mm TL) captured through electrofishing (2018), TDDs (2018) and repeated PTs (2019) across study sites. The area sampled using each method (m 2 ) and the number of bullhead captured (n) is also denoted (m 2 ) and electrofishing setup, such as the number of operatives or anode devices, could also influence the effectiveness. The TDD, however, systematically removes available refugia from the channel, leaving no place for fish to hide, resulting in a much higher catchability. While the ability to produce total population estimates through sweep depletion analyses is a valuable tool in fish stock assessments for both monitoring and managing wild and stocked fisheries (Cowx, 1983;Vehanen et al., 2013), our results highlight methodological constraints potentially limiting the reliability of such assessments.
The PTs produced density values generally congruent with the TDDs and sampled a wide range of size classes even after a minimum deployment time of just one day. However, the observed population structure of bullhead in PTs showed an even size class distribution -which differed strongly from both other methods that indicated a strong juvenile dominance. These pronounced observed differences can be related to a number of potential causes. When compared with the 'single point in time' samples generated by our electrofishing and TDD surveys, these differences could, for example, represent real changes in the population structure over the PT sampling season that included the summer months when bullhead growth rates are highest. It has been noted that, in productive systems, bullhead can attain lengths of 50 mm within their first year (Mills & Mann, 1983). As such, we can therefore expect much smoother population structures due to growth effects in the PT samples.
Furthermore, antagonistic interactions and competition between bullhead and invasive signal crayfish could influence PT samples. Signal crayfish have been shown to be dominant over bullhead and exclude bullhead from refugia (Bubb et al., 2009). Although Bubb et al. (2009) found no evidence of changes in the response of bullheads to different sized crayfish, it is possible that juvenile bullhead are more sensitive to crayfish presence, especially given crayfish can predate on small bullhead (Guan & Wiles, 1997, 1998 (Smyly, 1957). Bullhead also exhibit a strong 'homing instinct' and will repeatedly return to a particular stone (Smyly, 1957

Practicalities of the methods
Further to the performance, consideration of practical requirements, resources and risks associated with survey methods strongly influence their suitability. Electrofishing is a well-established method, but requires a suite of expensive specialist equipment and a team of 2-3+ trained operatives (Evans et al., 2017). Physical site characteristics could also limit effectiveness, such as water chemistry or turbidity.
Applying an electric field into a watercourse furthermore poses a risk to operatives and biota, as for example the electric shock can cause fish to suffer burns and fractures, and it can cause crayfish to lose chelipeds (Alonso, 2001).
Survey effort of TDDs is comparable to electrofishing surveys in that they are resource intensive and expensive to undertake, requiring specialist equipment, trained operatives and multiple sweeps . TDDs are also limited to sites with good access such that watercourses can be easily over-pumped and dewatered . Repeated dewatering could also pose a risk to both target and non-target organisms, and thus TDDs need to be undertaken with utmost care and consideration.
PTs, in contrast, require a lower sampling effort (one operative required for ∼15 min to deploy and ∼15 min to extract one trap and process the catch) and hence may present a more cost-effective approach. PTs are relatively cheap to build and easy to transport, so are suitable for remote survey locations. The passive nature of the trap reduces animal welfare considerations, with minimal impact to bycatch.
Whilst PT sampling necessitates two site visits for deployment and subsequent collection, working hours are still considerably lower than for TDDs and electrofishing. PTs are, however, potentially limited to substrate type and with high specificity to microhabitats. As of yet, PTs have only been tested in rocky headwaters where cobbles are the dominant substrate type. Further work is needed to assess their effectiveness in other aquatic systems.

Implications for conservation and management
Bullhead are the only freshwater cottid found in the United Kingdom (Tomlinson & Perrow, 2003). They are a protected species listed on Annex II of the European Commission Habitats Directive (Boon & Lee, 2005;Knaepkens et al., 2005). The ability to monitor their populations is crucial to understand population trends and assess conservation status in the face of various stressors, including invasive species like the signal crayfish (Guan & Wiles, 1997). Methodological constraints and poor catchability have limited effective population assessments of benthic fish, and the importance of such species within ecosystems has likely been underestimated (Harrison et al., 2005). Benthic fish such as bullhead may be considered keystone species in some systems where they attain high abundances and have an intermediate trophic position (Harrison et al., 2005;Woodward et al., 2008). Bullhead have strong associations with substrate type, often preferring coarse gravel and cobble (Welton et al., 1983) which can vary between the seasons (Harrison et al., 2005). As a result, the ability to accurately record their densities within small habitat is paramount to better understand their behaviour and ecology. While the shock of electrofishing can cause fish to rapidly dart between habitats (Harrison et al., 2005), the PTs now offer a passive method to explore specific habitat preferences, recruitment patterns and response to stressors.

CONCLUSIONS
Electrofishing severely underestimates the abundance of benthic fish throughout the rocky headwater study system -even where depletion curves were used to extrapolate 'true' population densities. Fish behaviour and site characteristics such as substrate type may have limited its effectiveness. TDDs and PTs prove effective at surveying benthic fish and generating detailed demographic data. However, site characteristics and resources may limit the wider application of TDDs for benthic fish. The PT presents a cost-effective approach with a low sampling effort and low risk to non-target organisms. Further research is required to assess the effectiveness of PTs in other aquatic systems, but the method shows great promise for the assessment of benthic fish populations.