2.2. Data Sets
The acoustic echogram was collected using a scientific split-beam echosounder system (Simrad EY60, Kongsberg, Norway) with an operating frequency of 70 kHz (
Table 1). Because of this system’s lack of long-term stability, the scattering and absorption loss of echoes and the uncertain orientation of fish in the beam during the investigation would affect the reliability of acoustic data. The EY60 system was calibrated in open water (113°35′ E, 22°03′ N; 20 m water depth) on 25 October 2019, based on the standard guidelines [
20]. In addition, to prevent vibrations and reduce noise, the transducer was fixed 1.0 m underwater outside the starboard side of the hull through the deflector. Real-time longitude and latitude information were collected synchronously with acoustic echograms using a Trimble differential GPS (Trimble, Sunnyvale, CA, USA). The average survey speed was 5.0 knots.
Acoustic echograms were analyzed using Echoview 6.0 software (Myriax Pty. Ltd., Hobart, Tasmania), referring to the recommendations outlined in the standard operating procedures for acoustic fishery surveys in the Great Lakes [
19]. Considering the influence near-field effect of acoustic detection and blind zone, the initial water depth for echo integration was 4.0 m underwater, and the terminal water depth was 0.5 m above the seabed. The background noise in the echogram was reduced using the virtual variable module of the Ecoview system [
21]. According to echogram inspection, other non-target signals (such as bubbles) were removed manually. Since there was no reference minimum volume backscattering strength (S
V) threshold for distinguishing fish and plankton echoes in this area, the threshold response module would be used to determine a confidence threshold for fish echo extraction [
22]. According to the response result, −70 dB (decibels referenced to 1 m
−1 [dB re m
−1]) could be a threshold to remove the backscatter of plankton and minimize fish exclusion. The elementary distance sampling unit (EDSU) is the length of the transect along which the acoustic measurements are averaged to give one sample [
23] and was set to 1 km in this study.
A series of physical, chemical, and hydrological factors were selected or measured for all 16 sampling stations (
Figure 1, S1–S16). Samplings were started when the vessel arrived at each station, and the acoustic data in the environmental sampling periods were not used for analysis. At each station, water depth was measured by the EY60 system, and sea surface temperature (SST), dissolved oxygen (DO), salinity, and pH were measured using a handheld YSI multiparameter instrument (professional plus) in situ. Suspended solids (SS) were measured by the weighting method (Chinese national standard for suspended substances) using 100 mL of river water filtered through nitrocellulose and cellulose-diacetate blend membrane (pore size 0.45 μm) [
24]. Chemical oxygen demand (the amount of oxygen consumed during the chemical oxidation of oxidizable substances in the water body, COD) was determined by the potassium dichromate method; if the concentration of COD is less than 0.001 mg/L, it will not be detected (nd) (EPA 410.1).
Chlorophyll a (Chl a) was measured via the fluorescence method, where 500 mL of stream water samples were pre-filtered through a 200 μm mesh sieve to remove large zooplankton and debris. Filtered water was then passed through 0.7 μm GF/F filters (Whatman, England), with the filters then wrapped in aluminum foil and stored at −20 °C in darkness. Chl a on filters was extracted by 10 mL of 90% acetone at 0°C for 20 h in the dark and measured using a Trilogy Fluorometer (Turner Designs, Sunnyvale, CA, USA: Trilogy Module: CHL-A NA, Model #7200–046).
For nutrient analyses (nitrate [NO3−], nitrite [NO2−], ammonium [NH4+) and phosphate [PO4+]), the remaining water samples which had passed through 0.7 μm GF/F filters were analyzed by segmented flow automated colourimetry using the manufacturer’s standard procedures (San + + Automated Wet Chemistry Analyser, Skalar, Breda, The Netherlands).
The fishing vessel towed an otter trawl with a stretch mesh of 4.0 cm throughout the body, 1.0 cm stretch mesh cod-end, and a 2.2 m spread between doors. Four trawl samples were conducted during this survey (S2, S7, S10 and S15,
Figure 1), covering all transects to obtain information on fish biology and species composition in the survey area to analyze the fish community structure and resource assessment. After the vessel arrived at the stations, it started trawling forward along the transects. The sample duration was limited to 30 min at approximately 3 knots. After each trawl sample, the vessel returned to the starting point to continue the environmental samples and acoustic detection. The echograms during the trawl samples were not used for data analysis. All catches were classified and counted on the deck. Since the catch number of all species was less than 50, the length (accurate to 1 mm) and weight (valid to 1 g) of each individual were measured.
2.3. Fish Density Estimation
The acoustically-derived density was estimated using the catch apportionment method [
18]. The trawl catch composition was used as the primary basis for distributing the acoustic integral values
Sa (m
2/nmile
2, mean acoustic area backscattering coefficient of each EDSU). The number
ρ(i,a) (inds/km
2) and biomass densities
ρ(i,b) (kg/km
2) of the
i-th species in each EDSU were calculated as follows:
where
ci (%) is the number percentage of the
i-th species in each trawl,
(m
2) is the mean acoustic backscattering cross-section of the considered acoustically-detectable species in each transect, and
(g) is the mean body weight of the
i-th species.
where
n is the number of considered fish species in each trawl,
TSi (dB re 1 m
2) is the target strength of the
i-th species, and
Li is the mean length of the
i-th species. The slope of the TS-L relation was assumed to be 20 [
25], and the
b20 of the considered fish species in
Table 2 were cited from relevant acoustic surveys [
16,
26,
27,
28,
29].
When the densities of each EDSU were obtained, the mean
(inds/km
2) and
(kg/km
2) for the
i-th species in the OWF area were calculated as:
where
k = 1 to M indicates the number of EDSUs of the OWF area.
The precision of the acoustic survey detection was represented by the coefficient of variation (CV, %), and the formula is:
where D is the total length of the transects, A is the size of the survey area, and a and b are formula coefficients [
18].
To better analyze the distribution patterns of fishery resources, we divided the study area into two sub-areas, the northern part (S1–S8) and the southern part (S9–S16).
2.4. Fish Community Parameters
The Index of Relative Importance (
IRI) was used to classify the dominant species in the fish groups [
30]. The formula is as follows:
where
i represents individual fish species,
is the catch biomass percentage (%) of a species,
is the catch abundance percentage (%), and
is the frequency of occurrence of the species in the trawl (%). Species with
IRI ≥ 1000 were considered dominant species, those with 100 ≤
IRI < 1000 common species, 10 ≤
IRI < 100 general species, and those with
IRI < 10 rare species [
31].
The Shannon–Weiner diversity index (
H′) represents the species diversity of the fish community.
where
H′ < 1: low diversity, 1<
H′ < 3: moderate diversity, and
H′ > 3: high diversity [
32].
Margalef richness index (
D):
Pielou uniformity index (
J′):
where
Pi is the catch number percentage of the
i-th species in each net,
N is the total catch number for each net, and
S is the number of fish species in each net [
33].
The ABC curve method analyzes the characteristics of the community under different disturbance conditions by calculating the distribution of species abundance and biomass using the
W-statistic as a statistic of the ABC curve method [
34]:
where
Ai and
Bi represent the cumulative percentage of number and biomass of species
i in the ABC curve, respectively, and
S is the total number of fish species. When the biomass curve lies above the abundance curve, the community is generally undisturbed, and k-selected species (slow-growing, large, late maturing) would be dominant. With increasing disturbance, the biomass and abundance curves may cross or overlap, and the community is moderately disturbed. In heavily disturbed instances, the entire abundance curve lies above the biomass curve, and r-selected species (fast-growing, small, opportunistic) would be dominant. The W value represents the degree of disturbance and ranges from −1 to 1. A positive value suggests an undisturbed community, and a negative value indicates a disturbed community [
34]. The ABC curve and the
W index were calculated using Primer 6.0 software (Plymouth Marine Laboratory, Plymouth, UK).
2.5. Canonical Correspondence Analysis
A detrended correspondence analysis (DCA) was applied to the fish abundance matrix of environmental stations (station × abundance of each fish species of the EDSU where the station is located) to analyze the fish assemblages under the influence of environmental stresses [
35]. Based on the value of the “lengths of gradient” item in the DCA, a unimodal model (canonical correspondence analysis, CCA) can be used if the value is greater than 4; a linear model (redundancy analysis, RDA) can be used if the value is less than 3; both models can be used if the value is between 3 and 4 [
36]. According to the result, this study conducted a CCA model using R software (version 3.3.2 and ‘vegan’ package 2.4).
Fish abundance data were Hellinger-transformed [
37], and environmental data were logarithmically transformed before CCA to eliminate the influence of extreme values on the ordination. Monte Carlo permutations (
p < 0.05) were used to select the environmental factors that significantly affected fish distribution, and the permutation cycle number is 999. The environmental factors with high partial correlation coefficients (
p < 0.05, |r| > 0.5) and variance inflation factors >20 were omitted from the final CCA [
38].