GUO Chuan-Bo， WANG Rui， ， QU Xiao， ， XIN Wei， CHEN Yu-Shun， and LI Zhong-Jie，

(1. State Key Laboratory of Freshwater Ecology and Biotechnology， Institute of Hydrobiology， Chinese Academy of Sciences，Wuhan 430072， China; 2. University of Chinese Academy of Sciences， Beijing 100049， China)

Abstract: To assess possible bias of different fishing methods is essential to appropriate fisheries management. In the current study， fish assemblage structure of a shallow Yangtze River lake was assessed by combining one international standard sampling gear (multi-mesh gillnet)， and one traditional Chinese gear (the dense-mesh weir). Using Lake Biandantang as a case study， a total of 27 fish species were collected from the two gears combined， including 2 new species that had not been recorded previously in this lake. Results suggested that fish assemblages had changed greatly compared to a previous study conducted in the 1990s. Specifically， differences were found in species composition， abundance， biomass， and length distributions collected from the two gears. Difference in total length (TL) distributions of fishes caught was the most conspicuous change. Fishes collected from the weir ranged from 40—70 mm TL， whereas fishes collected from gillnets ranged from 90—140 mm TL. Multivariate ordinations based on relative abundance and biomass data also indicated fish assemblage structural differences between the two gears. The comparative results showed that the multi-mesh gillnet was effective at assessing fish assemblages in shallow lakes， such as those found along the middle and lower reaches of the Yangtze River. Additionally， assessments using only one gear could have some limitations with respect to interpreting fish assemblage changes over time.

Key words: Fish assemblages; Assessment; Multi-mesh gillnet; Weir; Biandantang Lake

Fish assemblage structure is one of the key biological components of aquatic ecosystems， and combines important assemblage attributes， including species richness， species composition， and relative abundances. These biological assemblage measures are fundamental requirements in order to conduct environmental monitoring and assessment. Fish assemblage metrics are being increasingly used to determine the ecological status of many ecosystems[1—4].However， accurate assessment of fish assemblages is challenging for ecologists largely due to gear selectivity issues. Various gears are available to assess fish assemblages worldwide， including direct underwater observation， pop-nets， cover nets， gillnets， electrofishing， trawls， and purse seines. However， the capture effectiveness in terms of species selection and size selectivity of various fishing gears may differ significantly; thus， high variance may result when characterizing fish assemblages， regardless of the gear.Therefore， there is a great need to evaluate differences of various fishing gears (i.e. gillnet， weir) when assessing fish assemblages in aquatic ecosystems[5—9].Gillnet and weir were two main protocols in fishing both in scientific research and fishery industry in freshwaters worldwide[6，8].

Multi-mesh gillnet， a standard fish sampling gear developed by the Nordic Freshwater Fish Group(NOFF) in 1990s， has been used worldwide for fish sampling in rivers and lakes[10]. To evaluate the efficiency and selectivity of multi-mesh gillnets， nume-rous studies comparing multi-mesh gillnets with electro-fishing， pop nets， and trawls have been conducted，and thus， provided a systematic assessment of multimesh gillnets[11—13]. Multi-mesh gillnets are passive gears， and as such， their catch effectiveness depends largely upon fish activity; thus， variability with respect to species， morphology， and sizes can be significant[11]. On the other hand， weir fishing is a traditional fishing method used throughout China， and also an important component of the traditional “unified fishing method”.In general， gillnets and weirs have been widely used for fish assemblage sampling in rivers， lakes， reservoirs， and other water bodies due to their convenience[14—19].

Currently， there are no studies that have compared the effectiveness of multi-mesh gillnetting and fish weirs with respect to assessing fish assemblage attributes， despite that multi-mesh gillnets have been employed previously for fish sampling in shallow lakes[14，15，17]. In this study， our goal was to evaluate the effectiveness of multi-mesh gillnets and densemesh weirs during the course of annual fish sampling begin conducted in Biandantang Lake –a typical shallow Yangtze River lake. Specifically， our objectives were to: (1) compare current and historical fish assemblages in Biandantang Lake; and (2) assess the feasibility and effectiveness of multi-mesh gillnetting and fish weirs for sampling fish assemblages in shallow lakes. We further hoped to infer which sampling method might be more effective for use in shallow lakes that are common along the middle and lower reaches of the Yangtze River.

1 Materials and Methods

1.1 Study area

Biandantang Lake (37°16′N， 114°41′E) is located in Hubei Province， central China， along the middle reach of the Yangtze River. It was formerly part of Bao’an Lake， but was separated following construction of a stone dam in the 1970s. Biandantang Lake is a typical of many shallow lakes in this reach of the Yangtze River. It has a surface area of 3.3 km2[20]，an average depth of 1.9 m and an annual mean water temperature of 19.4℃. As recently as the 1990s， the lake was dominated by macrophytes， with biomass of submerged macrophytes as high as 2309 g/m2. A more recent survey conducted in June 2010 suggested that submerged macrophytes have almost disappeared from the lake due to ineffective protection and inappropriate fish stocking management. At present， a small population ofTrapa bispinosaRoxb. exists in the northwest portion of the lake (unpublished data).

1.2 Sampling gears and procedures

In this study， multi-mesh gillnets and fish weirs were deployed simultaneously to sample fish assemblages in Biandantang Lake. Multi-mesh gillnets，were modified from Appelberg[10]and composed of 12 panels knot to knot， with mesh sizes ranging from 5 to 55 mm (i.e.， an exponential rate of approximately 1.25). The dimensions of the multi-mesh gillnets were 1.5 m high and 30 m long， which equaled 45 m2in total net area. Dense-mesh weirs were hand-made by a local fisherman， with each weir consisting of two long block nets and two fyke nets. The height of the weir was 2 m， with a mesh size of approximately 5 mm.

Annual fish sampling was conducted bi-monthly at three sampling sites (S1， S2 and S3) in Biandantang Lake between September 2010 and January 2011. Sites selected were located at different offshore distances and assumed to represent distinct habitat types across Biandantang Lake. During the entire 5-month sampling period， 4—8 multi-mesh gillnets and 2—4 weirs (each weir have two pair of block nets and fyke nets) were deployed at each sampling site; overall， the total sampling effort expended for multi-mesh gillnets was 89 net-nights， whereas 18 net-nights of effort was expended using fish weirs. Preliminary experiments had indicated that sampling durations from 17:00 p.m to 06:00 a.m (approximately 13 hours of fishing time) were optimal for multi-mesh gillnets because low capture efficiencies due to excessive accumulation of captured fishes were avoided (personal observations， unpublished data). Similarly， the sampling time for fish weirs was consistent with multi-mesh gillnets. Specific locations of all sampling sites and their offshore distances are contained in Tab. 1.

1.3 Data Collection and Analysis

For each sampling site， offshore distance， water depth， and water transparency were measured during each sampling month. Immediately following sampling， all fishes were identified to species， counted， and individually measured for total length (TL，in mm) and weighed (in g). Fish assemblage structure was reflected by calculation of three common indices below:

Shannon Wiener Diversity Index:

Tab. 1 Distribution of the sampling sites in the Biandantang Lake

Chi-square tests of independence were used to assess differences between fish species composition andTLdistributions from fish samples collected from the two fishing gears (i.e.， multi-mesh gillnets and fish weirs). Non-metric multidimensional scaling(NMDS) was used to assess differences between fish assemblages depicted by the two gears based on the relative abundances and relative biomasses. A Bray-Curtis similarity index was calculated to illustrate the differences between assemblage compositions. All data were analyzed using SPSS22.0， V.8.5 (SPSS，Inc.， Chicago， Illinois， USA) and PC-ORD， V.5.0(MjM Software， Gleneden Beach， Oregon， USA).

2 Results

2.1 Assemblage structure of fish in Biandantang Lake

A total of 27 species representing 7 families of fishes were identified from Biandantang Lake. Of the 27 total species， 21 (78% of the total) were from the family Cyprinidae. There are 21 and 21 fish species captured by the multi-mesh gillnet and weir respectively， with 15 species co-occurring in both gears.Species composition differed between the two gears.All individuals ofCyprinidae carpio，Megalobrama amblycephala，Pelteobagrus fulvidraco，Siniperca chuatsi，Hypophthalmichthys molitrix， andAristichthys nobiliswere caught only by multi-mesh gillnets.Conversely， weirs were the only gear to catchSarcocheilichthys nigripinnis，Acheilognathus tonkinensis，Rhodeus ocellatus，Rhodeus fangi，Hemirhamphus intermedius， andNeosalanx taihuensis(Tab. 2).Chi-square test results indicated that the species composition differed between gears=131.16，P＜0.001).

Fish assemblages depicted by multi-mesh gillnets were dominated numerically byToxabramis swinhonis(49% relative abundance)，Hemiculter leucisculus(12%)，Culter dabryi(10%) andCarassius auratus(10%) (Tab. 2). In terms of assemblage relative biomass，Hypophthalmichthys molitrix(37%)，Carassius auratus(29%)，Toxabramis swinhonis(9%)， andCulter dabryi(5%) accounted for the greatest proportions (Tab. 2). In contrast， fish assemblages depicted by weirs were dominated byNeosalanx taihuensis(71.6% relative abundance)， and followed byToxabramis swinhonis(12%)， andCulter dabryi(11%) (Tab. 2. In terms of assemblages relative biomass，Culter dabryi(38%)，Toxabramis swin-honis(34%)， andNeosalanx taihuensis(7%) were dominant in weir catches. Overall， estimates of mean number per unit effort (NPUE) for weirs and multimesh gillnets were (410±194) ind./net and (30±12) ind./net (mean±SD)， respectively.

Tab. 2 Comparisons of the relative biomass and abundance of fish collected by two gears

Fish diversity in Biandantang Lake as reflected by the Shannon-Wiener， Simpson and Pielou index were 1.80， 0.28， and 0.59， respectively for multi-mesh gillnets; and 1.04， 0.53 and 0.34 for dense-mesh weirs. Results suggested that multi-mesh gillnets might reflect significantly greater diversity levels than weirs， at least in terms of the Shannon-Wiener and Pielou indices.

The relative abundance and relative biomass of the two gears showed thatNeosalanx taihuensis,Toxabramis swinhonisandCulter dabryiwere the primary dominant species in Biandantang lake， followed byCarassius auratus，Hemiculter leucisculus，Cultrichthys erythropterusandPseudorasbora parvaand so on (Tab. 2).

2.2 Differences between fish assemblages in gillnet and weir catches

NMDS analyses based on relative abundances and biomasses was restricted to include only species that comprised ＞1% of the total abundance or total biomass of the assemblage. In cases where species comprised ＜1% of the total abundance or total biomass， these species were merged into one category termed “rare” species. The results shown in Fig. 1 illustrated significant differences between fish assemblages assessed by multi-mesh gillnetting and weir fishing. Differences were also found between the two sampling methods in fish diversity.

2.3 Differences between length distributions in gillnet and weir catches

Fig. 1 Ordination map by Non-metric Multi-dimensional Scaling(NMDS) of three sampling sites based on relative abundance(a)and relative biomass(b)of fish assemblages collected by gillnet(C1-C3) and weir (C4-C6)

Fig. 2 Length frequency distribution of whole fish collected by two gears

Fig. 3 Length frequency distribution of T. swinhonis in two gears

Fig. 4 Length frequency distribution of C. dabryi in two gears

3 Discussion

3.1 Current and historical status of fish assemblages in Biandantang Lake

The 27 fish species collected in the present study in Biandantang Lake represented a 40% decrease from a previous study that recorded 45 species in 2003[16]， mostly due to the previous study collected fish samples from multiple fishing methods， include gillnetting， weir fishing， electro-fishing， and other methods[16]. However， when Xieet al.[21]and Zhang[16]studied fish assemblage of Lake Biandantang in 1995 and 2002 respectively with single pop-net， finally，only 13 species and 18 species of fish were caught.This is because of the rareness of some species and limited investigations， moreover， again suggested that fish captured by single gear would result in large bias in assessing fish assemblages. However， two new recorded species，X. daviaiBleeker andCoilia ecteneswere recorded in this study. The occurrence of both species was probably related to artificial stockings of fry that have occurred in recent years in the Yangtze River[15].

By comparing with the fish assemblage investigation in 1994 and 2003[22，16]， the analysis of weir catches in the current study foundN. taihuensis，T.swinhonisandC. dabryiwere the most dominant species in Biandantang Lake， with relative abundances of 71%， 13% and 11% respectively， having replacedP.parva，R. ocellatusandR. giurinus. The disappearance of submerged-macrophytes might be a reason for the change of dominant species. A previous research indicated that submerged macrophytes can provide shelter， breeding and foraging habitat for small fishes likeP. parva，R. ocellatusandR. giurinus， therefore， relative abundance of these fish species would be expected to much greater in macrophytic environment[23]. Furthermore， the distribution of small fishes were found to be strongly related with the abundance of submerged macrophytes[24，25]. Submerged macrophytes in Biandantang Lake have decreased sharply since 2005， which restricted our study to sampling to largely grassless habitats that were previously vegetated. Thus， the massive reduction of submerged macrophytes might be a leading cause of the change in fish assemblages in Biandantang Lake.

3.2 Fish assemblages differences between multimesh gillnets and weirs

Several studies have indicated that multi-mesh gillnets commonly yield fewer species than other gears under the same circumstances[8，13，26]. However，21 common species were collected by gillnets and weirs in this study， which was consistent with Tapio，et al.[12]A total of 27 species was collected using these two gears. Fish assemblage composition in terms of biomass and abundance differed between the two gears， which suggested that each gear is highly selective for species. In particular， the multi-mesh gillnets caught more pelagic fishes of greater sizes compared to weirs. Commercial fishes likeC. carpio，M. amblycephala，P. fulvidraco，S. chuatsi，H. molitrixandA.nobiliswere captured by multi-mesh gillnets exclusively， which was similar to results reported by Goffaux，et al.[9]and Eros，et al.[11]In contrast， weirs caught species occupying the middle and lower water strata， such asS. nigripinnis，A. tonkinensis，R. ocellatus，R. fangi，H.intermedius， andN. taihuensis.These species were less likely to be captured by multi-mesh gillnets due to their slow swimming speeds and specialized body shapes (e.g.，H. intermediusandN. taihuensis) that would preclude capture by stationary nets.

Although between-gear differences were detected in length distributions when species were combined into a composite sample， no significant differences were detected in length distributions for any individual species. Thus， multi-mesh gillnets and weirs were equally effective and selective for many species. This was especially true for common species such asT. swinhonisandC. dabryi. Thus， both gears appeared suitable to indexT. swinhonisandC. dabryipopulations in shallow lakes， though weirs may potentially catch more small sizes ofC. dabryi， whereas catches from multi-mesh gillnets were more even. In terms of， sampling efficiencies， weirs were much greater than multi-mesh gillnets， with mean NPUE from weirs more than 13-fold greater than multi-mesh gillnets. However， both gears exhibited large errors as evidenced by high standard deviations， which suggested that catches might be heavily influenced by factors such as habitat heterogeneity， lake depth， and water transparency[24，25]. A possible alternative explanation might be that differences in catch rates between gears might be due to the behavioral traits of certain cyprinid fishes[11].

3.3 Sampling methods of fish assemblage study in shallow lakes

Results of this study suggested that multi-mesh gillnets and dense-mesh weirs were inefficient in capturing fishes with total lengths greater than 200 mm.Thus， in these systems， common commercial fishes such as black carp (Mylopharyngodon piceus)， grass carp (Ctenopharyngodon idellus)， silver carp (Hypophthalmichthys molitrix)， and bighead carp (Aristichthys nobilis) would be selected against by these gears，and thus， likely underrepresented in samples. Selectivity of fishing gears on sizes and body shapes of individual species can influence estimates of species richness and total catches. Therefore， it may become problematic to use these two gears alone to assess fish diversity and species richness in shallow lakes， as large underestimations may likely result. In order to generate more accurate estimates (e.g.， NPUE， H′，etc.) and decrease sample variances with fish assemblage data， it is suggested that more studies of sampling gear selectivity are needed. In particular，studies designed to better assess deficiencies (i.e.， lists of species selected for and against) of each gear are needed. These findings would help investigators design better fish sampling programs that would improve monitoring and detection of environmental impacts on fishes.

Use of multiple fish sampling gears simultaneously is essential for enhanced environmental monitoring.However， studies relative to gear selectivity are still in a preliminary stage in China. A few studies were conducted on the selectivity of gillnets and trawls in 1980s and 1990s[27，28]， but very few in-depth studies are available for comparison. Thus， further research should focus on the selectivity of different fish sampling gears so that the appropriate methods can choose that minimize sampling bias.

Other sampling gears would also be useful in shallow lakes， such as this found in the middle and lower Yangtze River. In general， fish assemblage sampling methods can be divided into four categories，which are poisoning， electric fishing， gillnetting， and underwater observation[29]. Internationally， electric fishing is widely used to sample fish in lakes and rivers， though this method is not common in China.Underwater observation is mainly employed to assess distribution of fishes in marine environments or deep，clear lakes because of its instantaneity and reliability.However， underwater observation method also has not been widely adopted in shallow lakes due to its greater requirements for equipment and operation[30].Poisoning is mainly used in small streams， lakes， and ponds， and is completely unselective for sizes or species; thus， total numbers and biomasses of species can be estimated directly with relatively high accuracy.Xie[31]used 20 mg/L calcium hypochlorite and 5 mg/L sodium pentachlorophenol to sample fishes in shallow Liangzi Lake and a total of 19 species has been recorded in this catch. However， the chemicals used in these applications cause water pollution， thus， poisoning cannot be widely used. Compared to methods above， netting gears， including pop nets， gillnets，purse seines， weirs， and trawls are the most commonly used fish sampling methods for use in shallow lakes. Besides， 10—14 species of small size fish has been collected by purse seine in a typical shallow Niushan Lake according to Ye[23]. 21 species of fish were identified in weir and multi-mesh gillnets catches in our study separately， while species caught by pop-net and purse seine was obviously less than these.Moreover， the use of pop-net and purse seine has higher requirements on habitat and involves complicated operation. In general， they can only achieve higher efficiency under shallow lakes with flat lakebed and macrophytic habitat. In contrast， multi-mesh gillnet used in our study can complete sampling under various complex habitats with high efficiency and simple operations. It has been used in fish assemblage investigations in several complex habitats， such as deep canyon[32]， shallow reed habitat[11]， deep lakes[8，13]andcoastal zone of shallow lakes[12]， and the sampling efficiency of this multi-mesh gillnets turned out to be very satisfied.