READ ME File For Dataset supporting the publication "Adopting a reductionist approach to advance acoustic deterrents in fish conservation" Dataset DOI: 10.5258/SOTON/D2820 Date that the file was created: Feb, 2024 ------------------- GENERAL INFORMATION ------------------- ReadMe Author: Amelia Holgate, University of Southampton Date of data collection: April 2021 Information about geographic location of data collection: University of Southampton Related projects: Applying appropriate frequency criteria to advance acoustic behavioural guidance systems for fish - Doi:10.1038/s41598-023-33423-5 -------------------------- SHARING/ACCESS INFORMATION -------------------------- Licenses/restrictions placed on the data, or limitations of reuse: No limitations Recommended citation for the data: This dataset supports the publication: AUTHORS: Amelia Holgate, Paul R. White, Timothy G. Leighton and Paul Kemp TITLE: Adopting a reductionist approach to advance acoustic deterrents in fish conservation JOURNAL: Frontiers in Freshwater Science PAPER DOI IF KNOWN: 10.3389/ffwsc.2024.1320582 Links to other publicly accessible locations of the data: NA Links/relationships to ancillary or related data sets: Dataset for Applying appropriate frequency criteria to advance acoustic behavioural guidance systems for fish: DOI: 10.5258/SOTON/D2486 -------------------- DATA & FILE OVERVIEW -------------------- This dataset contains: -Signal_Detection_Theory_V4.xlsx - Analysis ans calculation of parameters for the signal detection theory aspect of tha article. -Trial_Results_Exp2_V2.xlsx - Recording variables from the trials inclusing startles and any external factors. Relationship between files, if important for context: NA Additional related data collected that was not included in the current data package: NA If data was derived from another source, list source: NA If there are there multiple versions of the dataset, list the file updated, when and why update was made: NA -------------------------- METHODOLOGICAL INFORMATION -------------------------- Description of methods used for collection/generation of data : Experimental setup Trials were conducted in a white medium density polyethylene cylindrical arena (modified 100 L Round Water Tank; 55.5 cm diameter × 45 cm depth, 4 mm thick; Direct Water Tanks, Retford, Nottinghamshire, UK) suspended in a large tank (8 m length × 8 m width × 5 m depth). The arena was suspended in water (Fig. 1) using a bespoke metal frame to ameliorate the typical challenges faced by using a small tank in air (Holgate et al., 2023). By doing so, the impedance difference between the water in the tank and the surrounding air was reduced, creating a more homogeneous sound field in terms of SPL. The experimental arena was covered by a black polyethylene mesh (6 mm mesh width) to prevent escape of fish and filled to a depth of 30 cm. Water was replaced (≈20 L water change) after each trial to maintain good water quality. An underwater transducer (Electro-Voice UW-30; maximal output 153 dB re 1 μPa at 1 m for 150 Hz, Lubell Labs, Columbus, USA) was suspended 70 cm below the arena and a hydrophone (8105, manufacturer-calibrated sensitivity -205 dB re: 1 V/ μPa; Brüel & Kjær, Denmark) placed 20 cm from the tank to continuously monitor the sound throughout each trial. Trials were recorded via a webcam (C920; HD 1080p; 24 frames per second; Logitech Pro, Switzerland) installed directly above the experimental arena and footage subsequently analysed. The laboratory was lit by fluorescent lighting which provided sufficient illumination for video recording. Experimental design The study consisted of 20 replicates of 8 treatments (160 trials) based on a combination of a 250 Hz signal played at one of four SPLs (115, 125, 135 or 145 dB re 1 μPa) under one of two background noise conditions (treatment: 105 dB re 1 μPa band-limited 100-2500 Hz random noise; ambient control: ambient conditions). The frequency and SPLs of the signal were selected based on the results of a previous study (Holgate et al., 2023) in which 250 Hz was found to elicit an avoidance response more readily than any other frequency. The background noise was selected in light of the 3Rs (NC3Rs, 2014) such that the SPL wasn’t damaging to the ear but sufficiently above the ambient noise levels. Prior to the start of each trial, conducted between 16 and 30 April 2021, a single fish (total number of individuals, N = 160) was acclimated in the experimental arena for 30 min. During the trials, the fish were randomly assigned a background noise level (ambient or 105 dB re 1 μPa band-limited noise). Following this, each individual experienced a total of four exposures of the same treatment (to determine if there was any tolerance of the signal). Each exposure consisted of a sinusoidal 120 ms tone ramped with a 20 ms Hanning taper (Holgate et al., 2023) and was followed by 4 min (pre-signal period) before the next exposure. Although the activation latency of the Mauthner cells (neurons located in the hindbrain responsible for mediating a rapid escape reflex) in goldfish is 5–10 ms, the tone was played at 120 ms to ensure sufficient time for the frequency spectrum to be well defined (Eaton, 1977; Zeddies and Fay, 2005). Fish behaviour was continuously recorded during the trial and each fish was used once only. Methods for processing the data: Recordings of fish behaviour obtained during each trial were reviewed blind and in a random order. Presence or absence of a startle response, defined as a change in body tortuosity with erratic swimming such as a sudden increase in speed or a change in direction (Kastelein et al., 2008), were recorded for each trial. This did not include any freezing behaviour and the startle responses were not categorised as C or S startles due to limitations with the recording devices. Whilst there are a number of avoidance behaviours such as C-starts, S-starts, single bends and double bends (Domenici and Hale, 2019), this study focuses on the startle response in the broader context, i.e. using it as a proxy for an avoidance response. All statistical analyses were performed in R (version 3.6.3: https://rstudio.com/). Logistic regression was performed using a generalised linear mixed model (GLMM) with a binomial error structure and a “logit” link function. Akaike’s Information Criterion (AIC) was used to support each model by giving a parsimonious quantification of model fit (Spake et al., 2015). To determine whether external factors may have confounded the results by influencing the probability of startling, a reductive model was developed. Factors included in the model were: tank days (minimum number of days in the husbandry tank); time of day (relative the beginning of the trial to the nearest hour); difference between holding and experimental tank temperature (°C); and mass (g). The initial GLMM contained all predictor variables with exposure (order of stimulus exposure), trial, and “exposure:trial” included as random effects. Manual backwards selection using variable significance (significance at p < 0.05) was undertaken as model simplification (Table 2). Trial was considered to be a random effect since the model produced the lowest AIC. Exposure (the nth stimulus played, 1 – 4) was included in a GLMM as a fixed effect alongside the other external factors, however, none were recognised to predict startle responses and the null model had the optimum AIC (Table 3). Since there was no effect of tolerance (exposure 1 – 4), the data were combined and treated as a single set. Software- or Instrument-specific information needed to interpret the data, including software and hardware version numbers: All statistical analyses were performed in R (version 3.6.3: https://rstudio.com/). Standards and calibration information, if appropriate:NA Environmental/experimental conditions:NA Describe any quality-assurance procedures performed on the data:NA People involved with sample collection, processing, analysis and/or submission:Amelia Holgate -------------------------- DATA-SPECIFIC INFORMATION -------------------------- Trial_Results_Exp2_V2.xlsx Number of variables: 31 Number of cases/rows: 641 Variable list, defining any abbreviations, units of measure, codes or symbols used: Trial Exposure (Exposure 1-4) Assigned (Trial code) Noise? (was there background noise?) SPL (Sound pressure level: 115, 125, 135, 145 dB re 1uPa) Date Min_Tank_Days (Days on the facility) Time_Acc (Time) Discrete Time (Time to the nearest hour) Temp_Hold (Temperature of the holding tank in 'C) Temp_Treat (Temperature of the treatment tank in 'C) Temp_Diff (Temperature difference between the holding tank and the treatment tank in 'C) Startle (Was there a startle? 1 = yes, 0 = no) Startle after first startle? Was there a startle after the first startle in subsequent exposures 1 = yes, 0 = no) Video time in min and seconds Length_SL (standard length in cm) Mass (mass in g) Depth Rough depth (an estimation) length10 Length multiplied by 10 to get the answer in mm RecordedL Recorded sound level at that depth (estimated and rough) in dB NL Recorded noise level at that depth (estimated and rough) in dB SNR Estimated signal to noise level in dB (rough - not reliable) ReceivedL Estimated received level in dB (rough - not reliable) NoiseRL Estimated received noise level in dB (rough - not reliable) Startles in 4 min before presentation Number of startles in the 4 min before signal presentaion External Noise Number of external noises in the 4 min before signal presentation Unknown Unknown signal Miss Missed signals (1 every 0.5 s where there isn't a startle) No Response Number of 'none responses' to a stimulus Startle Check Double checking Correct? Was the double check consistent with the previous results (to determine percentage accuracy) Missing data codes:NA Specialized formats or other abbreviations used: Signal_Detection_Theory_V4.xlsx Rows = 641 Trial Trial number Exposure Exposure 1-4 Assigned Assigned trial name Response NA, False alarm or Hit False Alarm NUmber of false alarms Hit Number of hits No Response Number of no response Miss Number of misses FAR False alarm rate (false alarm / (false alarm + no response)) HR Hit rate (hit/ (hit + miss)) Lambda (NORM.S.INV(Hit rate)+NORM.S.INV(False alarm rate))/2 d' NORM.S.INV(Hit rate)-NORM.S.INV(False alarm rate)