- Email: [email protected]
Evaluating gillnetting protocols to characterize lacustrine fish communities
Fisheries Research 161 (2015) 320–329
Contents lists available at ScienceDirect
Fisheries Research journal homepage: www.elsevier.com/locate/fishres
Evaluating gillnetting protocols to characterize lacustrine fish communities Timothy J. Alexander a,b,∗ , Pascal Vonlanthen a,b , Guy Periat a , Franc¸ois Degiorgi c , Jean-Claude Raymond d , Ole Seehausen a,b a Department of Fish Ecology and Evolution, Centre of Ecology, Evolution and Biogeochemistry, EAWAG Swiss Federal Institute of Aquatic Science and Technology, Seestrasse 79, CH-6047 Kastanienbaum, Switzerland b Division of Aquatic Ecology and Evolution, Institute of Ecology and Evolution, University of Bern, Baltzerstrasse 6, CH-3012 Bern, Switzerland c CNRS-University of Franche-Comte/UMR 6249 Chrono-environment, La Bouloie, F-25030 Besanc¸on Cedex, France d ONEMA French National Agency for Water and Aquatic Environment, Unité Spécialisée Milieux Lacustres, Pisciculture de Rives 13, Quai de Rives 74200, Thonon-les-Bains, France
a r t i c l e
i n f o
Article history: Received 15 April 2014 Received in revised form 12 July 2014 Accepted 15 August 2014 Handling Editor George A. Rose Keywords: Multimesh gillnets CEN standard Vertical net protocol Perch Coregonus
a b s t r a c t Ecological research and monitoring of lacustrine ecosystems often requires a whole-lake assessment of fish communities. Gillnet sampling offers an efficient means of estimating abundance, biomass and fish community composition. However the choice of gillnet sampling protocol may influence lake characterization via physical properties of the nets and allocation of sampling effort between littoral, benthic and pelagic habitats. This paper compares two commonly used, whole-lake sampling protocols applied across 17 prealpine, subalpine and alpine European lakes ranging widely in size, depth and altitude to determine their relative strength for research and management applications. Effort-corrected estimates of abundance, biomass and species richness were correlated between the protocols and both distinguished the trout-dominated alpine communities from subalpine and prealpine lakes dominated by whitefish and perch. A considerable amount of variance remained unexplained between the two protocols however, which seemed to correspond with differences in the proportion of effort among benthic and pelagic habitats. We suggest that both the European standard (CEN) and vertical (VERT) netting protocols are suitable for assessing ecological status and monitoring changes in lake fish communities through time. However the details of each protocol should be kept in mind when comparing fish communities between lakes. Mesh sizes used in CEN nets produce a more even size frequency distribution, suggesting that this protocol is most appropriate for assessing size structure of fish assemblages. The high proportion of netting effort in benthic habitats shallower than 70 m depth under the CEN protocol means that, particularly in larger lakes, outcomes will be disproportionately influenced by the ecological condition of this habitat. The VERT protocol presumably provides a more accurate estimate of whole-lake CPUE and community composition because effort, in terms of net area, is more evenly distributed across the entire volume of the lake. This is particularly important in large and deep lakes where pelagic habitats occupy a high proportion of the lake volume. © 2014 Elsevier B.V. All rights reserved.
1. Introduction The European Union Water Framework Directive (WFD) requires that all member countries characterize, assess and if
∗ Corresponding author at: Department of Fish Ecology and Evolution, Centre of Ecology, Evolution and Biogeochemistry, EAWAG Swiss Federal Institute of Aquatic Science and Technology, Seestrasse 79, CH-6047 Kastanienbaum, Switzerland; Division of Aquatic Ecology and Evolution, Institute of Ecology and Evolution, University of Bern, Baltzerstrasse 6, CH-3012 Bern, Switzerland. Tel.: +41 58 765 2202. E-mail address: [email protected] (T.J. Alexander). http://dx.doi.org/10.1016/j.fishres.2014.08.009 0165-7836/© 2014 Elsevier B.V. All rights reserved.
necessary, improve the ecological status of their freshwater ecosystems by 2015. Fish, benthic invertebrates, phytoplankton and other aquatic flora form the basis of the biological component of the assessment. In particular, the longevity of many fish species makes them robust and temporally integrated indicators of ecosystem status (Vander Zanden and Vadeboncoeur, 2002). The WFD requires assessment of three aspects of the fish community: whole-lake estimate of catch per unit effort (CPUE), species composition and the age structure of fish assemblages. A multimesh gillnetting protocol has been adopted across Europe to conduct assessments under the WFD and facilitate intercalibration of quality thresholds between countries (hereafter refered to as CEN protocol;
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Appelberg, 2000; Comité Européen de Normalisation, 2005). The CEN protocol divides each lake into benthic and pelagic zones and samples these zones using gillnets deployed horizontally in the water column or on the lakebed. Netting effort is allocated in the benthic zone by randomly sampling within defined depth strata with replication determined according to the maximum depth and area of a lake. Pelagic netting effort is also conducted within depth strata but only within one column at the deepest point of the lake. A second gillnetting protocol has been used widely to survey whole-lake fish communities for management and research in eastern France. This protocol describes the use of vertically oriented gillnets that simultaneously survey from the water surface to the lake floor. The vertical nets were originally developed in the USA for studying depth distribution of fishes in lakes and reservoirs (Horak and Tanner, 1964; Lackey, 1968; Lynch et al., 1989). The vertical netting protocol (hereafter referred to as VERT protocol), introduced by Degiorgi et al. (1993) and amended in Degiorgi et al. (2001), describes the application of vertical nets to sample wholelake fish communities. Under the VERT protocol, replicate gillnet series are deployed within each type of littoral and offshore habitat present in the lake. Littoral habitats (<5 m deep) are classified based on substrate composition and particle size, macrophyte morphology and density, and proximity to an inwardly or outwardly flowing watercourse. Two sublittoral and three deep pelagic habitats are also defined relative to the maximum depth of the lake. The widespread use of the CEN sampling protocol provides an unprecedented opportunity for community- and macro-ecological research into natural and anthropogenic conditions influencing lacustrine fish communities (e.g. Brucet et al., 2013; Mehner et al., 2005, 2007). However, meaningful interpretation of the results of such research requires an understanding of the strengths, weaknesses and idiosyncrasies of the data on which it is based. Multiple authors have commented that the CEN protocol under-represents pelagic species in assessments of whole-lake fish communities (Achleitner et al., 2012; Deceliere-Vergès et al., 2009; DeceliereVergès and Guillard, 2008; Diekmann et al., 2005). Other authors have noted that sampling only at the deepest point of the lake probably overlooks horizontal variation in fish communities living in pelagic habitats (Lauridsen et al., 2008; Specziar et al., 2009). These issues are particularly important in large and deep lakes where pelagic habitats constitute the vast majority of the lake volume. This paper therefore aims to assess and compare the utility of CEN and VERT protocols for management and research into whole-lake fish
321
communities. The investigation will focus on the level of correspondence and discrepancy between the protocols and their relative strengths in providing reliable estimates of whole-lake fish CPUE, size structure, species richness and community composition.
2. Methods 2.1. Fish sampling Comparison of CEN and VERT protocols was conducted by sampling fish communities using both protocols in 17 lakes across eastern France, Switzerland and northern Italy (Fig. 1). Physical characteristics of the sampled lakes are provided in Table 1. Surveys were conducted between August and October 2010–2013. All nets were set before dusk and retrieved approximately 14 h later. Due to the logistics of the large-scale field schedule, soak times occasionally varied from this target. To reduce the influence of variation in soak time, catches were standardized to 14 h by dividing by the soak time (in decimal hours) and multiplying by 14. Biomass and number of fish (abundance) were also standardized by the area of net. The resulting biomass per unit effort (BPUE) and number of fish per unit effort (NPUE) therefore reflected fish catches per square meter per 14 h. Catch per unit effort (CPUE) is used in this paper as the collective term for both BPUE and NPUE. Catch rate and CPUE are also used interchangeably. Use of the terms biomass and abundance always refers to effort corrected values. Sampling under the CEN protocol was conducted according to the European Committee for Standardization standard EN14757:2005 (Comité Européen de Normalisation, 2005). Briefly, benthic netting effort was located randomly within defined depth strata with replication determined by the maximum depth and area of a lake. Pelagic nets were set suspended in the water column within the same depth strata at the deepest point of the lake over consecutive nights. Benthic habitats were sampled with Nordic type gillnets consisting of a series of contiguous panels of twelve mesh sizes following a geometric series: 5, 6.25, 8, 10, 12.5, 15.5, 19.5, 24, 35, 43, and 55 mm (measured knot to knot). Each mesh panel was 1.5 m high and 2.5 m wide. The combined multimesh net was 1.5 m high and 30 m long. Pelagic nets consisted of the same mesh sizes, minus the 5 mm mesh. Panels in pelagic nets were 6 m high making the combined net 6 m high and 27.5 m long.
Fig. 1. Map of sampled lakes across Switzerland, eastern France and northern Italy.
322
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Table 1 Morphological characteristics of the study lakes. Lake
Altitude (m)
Area (km2 )
Max depth (m)
Mean depth (m)
Poschiavo Sils Brenet Remoray Chalain Saint-Point Joux Hallwil Morat Zug Neuchatel Geneva Walen Brienz Thun Lugano Maggiore
962 1797 1002 850 486 850 1004 449 429 414 429 372 419 564 558 271 194
2.0 4.1 0.8 1.0 2.3 5.2 9.5 10.3 22.8 38.3 218.3 582.0 24.2 29.8 48.3 48.7 212.5
85 71 18 27 32 43 32 48 45 198 152 310 151 261 217 288 372
61 35 9 14 17 16 16 28 23 84 64 152 105 173 136 134 177
Vertical net sampling was conducted according to the protocol proposed by Degiorgi et al. (1993) and amended in Degiorgi et al. (2001). The VERT protocol used gillnets series (separate nets for different mesh sizes) to simultaneously sample from the water surface to the lake floor. Each net in the series was 2 m wide and each series consisted of mesh sizes 10, 15, 20, 30, 40, 50, and 60 mm. Sampling effort was allocated according to the littoral and deep habitats present in the lake. Littoral habitats were mapped in each lake prior to the fish sampling event. Habitats were defined according to the dominant substrate composition, vegetation and proximity to a river or stream (Table 2). Deeper habitats were defined relative to the maximum depth of the lake. Gillnets were deployed at least three times in each littoral and deep habitat types present in the lake. Setting locations were randomly selected from within the available area of each habitat. The VERT protocol used different types of nets to sample littoral and deep habitats. Littoral habitats (to 5 m deep) were sampled with 2 m wide mesh panels for each mesh size deployed on the same lead and float line. A 2 m gap separated adjacent panels. The
Table 2 Habitat categories used to allocate sampling effort in the vertical netting protocol. Dimensions of the rock and sediment refer to the most common maximum dimension (length, width, height) of lithic particles at the site. Zmax is the maximum depth of the lake in meters. Some pelagic habitats may not be present in the lake if Zmax is less than 40 m. Sites with emergent macrophytes were further divided into dense and scattered based on a threshold of 10 cm between stems. Littoral habitats (<5 m) Emergent macrophytes Submerged macrophytes Floating macrophytes Woody debris or trees Leaf litter Rock slab or ledge Rocks or boulders Cobbles Gravel Gravel and cobbles Sand Fine sediment Out-flowing stream or river In-flowing stream or river
Details
Roots or branches in or touching water Solid rock/bedrock, no interstitial space >200 mm 20–200 mm 2–20 mm Mixture of gravel and cobbles 0.5–2 mm <0.5 mm
Deep habitats (>5 m)
Details
Sublittoral Deep sublittoral Min pelagic Med pelagic Max pelagic
5–10 m 10 m–0.3Zmax 0.3Zmax –0.6Zmax 0.6Zmax –0.9Zmax 0.9Zmax –Zmax
Volume (GL) 110.8 129.7 5.2 13.7 46.3 9.4 128.3 289.5 437.8 3,202.8 13,935.5 79,016.8 2,466.0 5,161.8 6,434.6 5,705.3 12,957.3
nets ranged in height from 1 m to 5 m and were selected to match the depth of the sampling site, ensuring that the net remains vertical in the water column. Gillnet series to sample deeper habitats consisted of separate nets for each mesh size, deployed as close to each other as weather conditions allowed. The nets were rolled onto lengths of PVC pipe (diameter 150 mm) filled with plastic bottles and waterproof expanding foam (Bartoo et al., 1973). The pipes provided floatation and allowed the nets to be deployed and retrieved using a customized winch. Two-meter wide aluminum spacers (10 mm hollow tubes), placed every 10–15 m along the net, ensured that it retained the appropriate width. Iron weights and a 25 mm diameter steel bar across the width at the base of the net provided the anchors to maintain the position of the lower end on the lake floor. Fish caught in both gillnet protocols were identified to species level; however due taxonomic uncertainty and difficulty to reliably distinguish between some species in the field, several taxa were aggregated to genera for analyses of abundance and biomass. These included: Coregonus spp. (C. albellus, C. alpinus, C. bondella, C. lavaretus, C. palea; referred to in this paper as whitefish), Rutilus spp. (R. rutilus, R. aula, R. pigus; referred to in this paper as roach), Alburnus spp. (A. alburnus, A. alborella), Scardinius spp. (S. hesperidicus, S. erythrophthalmus), Phoxinus spp. (P. phoxinus, P. lumaireul), Salmo spp. (S. trutta, S. marmorata, S. cenerinus, S. labrax), and Salvelinus spp. (S. umbla species complex).
2.2. Statistical analyses Protocols were compared by focusing on the three aspects of the fish community required for assessments under the WFD: wholelake estimate of catch per unit effort, species composition and the size structure of fish assemblages. Calculation of whole-lake CPUE differed slightly between the protocols. Mean CPUE, according to the CEN protocol (Comité Européen de Normalisation, 2005), involved calculating the catch rate for each individual net (abundance or biomass divided by net area, after correcting for soak time) and then calculating the mean CPUE across all nets set in the lake. Global CPUE, under the VERT protocol (Degiorgi et al., 2001), involved summing soak-time corrected fish abundance or biomass across all nets set in a lake and dividing by the summed surface area of all nets set in the lake. Mean CPUE reflects the average catch rate across all the nets set in the lake. Global CPUE describes the average catch rate across every square meter of net set in the lake. Note that mean CPUE and global CPUE give the same whole-lake estimate if net area is the same across all nets used in the calculation.
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fish size frequency distributions were used as a proxy for age structure in this study. Length and biomass frequencies distributions of fish collected under each protocol were compared without correcting for net selectivity. Fish community composition was summarized by species richness and Shannon diversity based on NPUE. We also compared between the protocols the relative abundance and biomass of common species: perch (Perca fluviatilis), roach (Rutilus spp.) and whitefish (Coregonus spp.). The correspondence between the protocols for all univariate metrics was determined using correlation analysis (Wilcoxon signed rank test). Multivariate differences between protocols in community composition were visualized by overlaying ellipses to distinguish 20, 40, and 60% clusters on non-metric multidimensional scaling (MDS) ordinations. MDS plots were based on Bray–Curtis resemblance on relativized species catches calculated by dividing the catch for each species by the summed CPUE for all species in the lake. One-way permutation analysis of variance (PERMANOVA; Type III (partial) sums of squares with unrestricted permutation of raw data) was used to test for differences between protocols across all lakes. Similarity percentage (SIMPER) identified which species contributed to differences between the protocols. All multivariate analyses were conducted in PRIMER 6 (PRIMER-E, 2008). To determine if the degree of similarity between the protocols was related to the physical characteristics of the lake (i.e.
323
morphology and altitude), the residuals between the CEN and VERT estimates of CPUE in each lake were regressed against the lake characteristics. The same analysis was also conducted for community composition using the Bray–Curtis similarity (common species; relativized) between the community composition of each lake as characterized by the VERT and CEN protocols. Values for physical characteristics were natural log transformed prior to analysis to stabilize their variance. A Bonferroni correction of ˛/n (where n is the number of tests), applied to accommodate the multiple comparisons, reduced ˛ to 0.01 for these tests. All univariate analyses were conducted in R (R Core Team, 2014) using ‘base’ package for univariate statistical analyses and ‘vegan’ for diversity calculations. Data were manipulated using ‘reshape’ (Wickham, 2007) and figures generated using the ‘base’ and ‘ggplot2’ (Wickham, 2009) packages. 3. Results 3.1. Catch per unit effort Fish biomass (BPUE) and abundance (NPUE) of each lake were significantly correlated between protocols (Fig. 2a and b; Wilcoxon signed rank test BPUE: V = 171, p < 0.0001, R2 = 0.66; NPUE: V = 171, p = 0.0001, R2 = 0.42). Despite the correlation, there were notable differences in the rank order of lakes between the protocols,
Fig. 2. Comparing estimates of whole-lake (a) BPUE, (b) NPUE, (c) taxon richness and (d) species diversity between CEN and VERT protocols. Dashed line indicates 1:1 agreement between protocols.
324
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 3. Percentage composition of fish communities by biomass sampled with (a) CEN and (b) VERT protocols, and by abundance sampled with (c) CEN and (d) VERT protocols.
particular in terms of abundance. For example, sampling with the VERT protocol suggested that Lake Brenet had the highest abundance and biomass of all lakes. Sampling with the CEN protocol placed the same lake at 6th highest for biomass and 9th highest for abundance. Abundance was also consistently higher under the CEN protocol with estimates ranging from 2 to 15 times greater than those of the VERT protocol in the same lake. Biomass was also higher in all lakes sampled by the CEN protocol but the difference was less severe (1.5–4.5 times). 3.2. Species composition Estimates of taxon richness for each lake were very similar between the protocols (Fig. 2c; Spearman’s rank correlation: S = 82.5, p < 0.0001, R2 = 0.93), with identical numbers of taxa recorded in 7 out of the 17 lakes. Species diversity was also weakly correlated between the protocols (Fig. 2d; Shannon diversity based on NPUE: Wilcoxon signed rank test: V = 149, p < 0.01, R2 = 0.09). The VERT protocol tended to estimate a more even community composition by abundance (Pielou’s evenness; data not shown) which resulted in higher estimates of species diversity than from CEN in 12 lakes. Fish community composition by biomass and abundance in fifteen of the sampled lakes was dominated by perch (P. fluviatilis; Percidae), whitefish (Coregonus spp; Salmonidae) and the Cyprinids roach (Rutilus spp; Cyprinidae) and chub (Squalius spp; Cyprinidae) in ratios that varied for each lake between the protocols (Fig. 3). The composition of the alpine lakes Sils and Poschiavo was strikingly different, where abundance and biomass were dominated by trout (Salmo spp; Salmonidae) and char (Salvelinus spp; Salmonidae), including introduced lake trout (Salvelinus namaycush; Salmonidae). The proportion of abundance and biomass of the three most frequently caught fish taxa was significantly correlated between protocols (excluding abundance of roach): perch (Fig. 4a; Wilcoxon signed rank test: biomass V = 0, p < 0.001, R2 = 0.68; abundance V = 8, p < 0.0001, R2 = 0.28), roach (Fig. 4b; biomass V = 20, p < 0.05, R2 = 0.28; abundance V = 85, p = 0.404, R2 = 0.28), and whitefish
(Fig. 4c; biomass V = 117, p < 0.01, R2 = 0.51; abundance V = 134, p < 0.0001, R2 = 0.56). The biomass of perch relative to other species was particularly strongly correlated between protocols but was consistently higher under the CEN protocol in all lakes. Roach was also recorded in higher proportions by biomass under the CEN protocol for the majority of lakes (13 of 15), while differences between protocols in proportion by abundance of this genus were less clear cut (CEN > VERT for 7 of 15 lakes). The proportion of whitefish was also very strongly correlated between protocols for abundance and biomass. The VERT protocol estimated a higher proportion of whitefish in almost every lake (abundance 14 of 15 lakes; biomass 10/15 lakes). Aside from perch and roach, burbot (Lota lota) also, tended to be caught more in CEN nets with higher relative biomass in all nine lakes where it was recorded (relative abundance CEN > VERT for 7 of 9 lakes). In two lakes, burbot were exclusively captured in CEN nets. Several species were also exclusively caught in one or the other protocol. Nets deployed under the CEN protocol captured the only representatives 7 species: stone loach (Barbatula barbatula; 1 lake), spiny loach (Cobitis sp.; 1 lake), southern pike (Esox cisalpinus; 1 lake), three-spined stickleback (Gasterosteus aculeatus; 1 lake), Padanian goby (Padogobius bonelli; 1 lake), freshwater blenny (Salaria fluviatilis; 2 lakes) and vairone (Telestes muticellus; 1 lake). In contrast, only one species, Italian barbel Barbus blebejus (1 lake) was caught exclusively in the VERT nets. Overall however, the number of occasions when one protocol detected a species in a lake that was overlooked by the other was approximately similar between protocols (24 for CEN and 28 for VERT). Overall, the CEN and VERT protocols characterized significantly different fish community composition by abundance (PERMANOVA: pseudo-F = 5.45, p > 0.001 with 999 unique permutations) and biomass (pseudo-F = 5.56, p > 0.003 with 998 unique permutations). Perch alone contributed over 30% of the difference in abundance between protocols, with lesser contributions from roach (Rutilus spp.), whitefish (Coregonus spp.), trout (Salmo spp.) and rudd (Scardinius spp.; SIMPER analysis; Table 3). Differences between the protocols by biomass were caused by a wider spectrum of species, but also with strong contributions from perch,
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
325
Fig. 4. Comparing the abundance and biomass of common species (a) perch, (b) whitefish, (c) roach and between protocols. Dashed line shows 1:1 agreement.
roach, whitefish, trout and rudd. Both protocols were sensitive to major geographical differences in fish assemblage composition among lakes (i.e. between trout-dominated Sils and Poschiavo and other lakes), but were rarely in complete agreement as to
which of the lakes with more similar community composition clustered together (MDS; Fig. 5). In terms of biomass, both protocols identified that the fish community composition of whitefishdominated lakes, Thun and Walen, were distinct; however there
326
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 5. MDS plots show clustering among lakes according to Bray–Curtis similarity of fish assemblage composition for biomass as sampled by (a) CEN and (b) VERT protocols and for abundance by (c) CEN and (d) VERT protocols. The extreme dissimilarity between Sils and Poschiavo and the other lakes was consistent between protocols; consequently plots exclude these two lakes to reveal clusters among the more similar lakes. Ellipses describe 20, 40 and 60% similarity in assemblage composition (shown in green, blue and red respectively).
Table 3 Similarity percentages analysis identified that perch (Perca fluviatilis), whitefish (Coregonus spp.), roach (Rutilus spp.), rudd (Scardinius spp.) and trout (Salmo spp.) contributed most strongly to differences between the protocols in fish community composition by (a) biomass and (b) abundance. (a) Species
Perca fluviatilis Coregonus spp. Scardinius spp. Rutilus spp. Salmo spp. Squalius spp. Tinca tinca Salvelinus namaycush Leuciscus leuciscus Salvelinus spp. Sander lucioperca Esox lucius Abramis brama
Mean % biomass
Contribution %
CEN
VERT
27.3 9.8 8.9 19.3 7.9 3.6 4.8 2.9 3.1 1.8 1.6 2.4 1.6
9.5 17.4 15.8 12.3 8.6 12.0 7.3 2.8 1.1 1.7 1.3 2.2 1.1
16.2 12.6 12.5 11.2 10.7 8.1 5.9 3.9 2.6 2.4 1.9 1.8 1.8
(b) Species
Perca fluviatilis Rutilus spp. Coregonus spp. Salmo spp. Scardinius spp. Phoxinus spp. Gymnocephalus cernuus Salvelinus spp. Alburnus spp.
Mean % abundance CEN
VERT
55.6 18.0 5.3 6.0 1.3 2.8 2.2 2.1 1.2
32.2 22.3 15.7 7.5 8.2 1.2 1.9 1.6 2.4
Contribution %
31.4 15.5 14.5 10.6 7.0 3.3 3.2 3.0 2.5
was little similarity between protocols in clustering among the remaining lakes. Whitefish also seemed to separate the lake clusters in terms of abundance. Fish assemblage composition in Brienz and Walen was distinct from the other lakes under the CEN protocol, while the higher proportion of whitefish caught in the VERT nets meant that this group also included Thun and Chalain under the VERT protocol.
3.3. Size structure Fish length frequency distributions were significantly different between the two protocols (two-sample Kolmogorov–Smirnov test: D = 0.405, p < 0.001). The CEN nets returned a relatively smooth frequency distribution of fish lengths across all lakes (Fig. 6; mode = 75 mm, mean = 103 mm, range = 32–875 mm). The frequency distribution of the VERT protocol was multimodal with the most commonly caught fish length slightly larger at 85 mm (mean = 138 mm, range = 32–1008 mm). Secondary peaks in frequency occurred at 125 mm and 155 mm. The frequency distribution of fish biomass displayed similar patterns between the protocols (not shown).
3.4. Lake characteristics influencing differences between protocols Lake volume was the only aspect of lake morphology or altitude significantly related to differences in CPUE between the protocols. Estimates of biomass tended to be more similar between the protocols in lakes that were larger by volume (F1,15 = 10.8, p < 0.01, R2 = 0.42; however residuals). No such trend was seen in abundance (F1,15 = 3.49, p = 0.08, R2 = 0.19). Differences in the multivariate
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 6. The geometric series of mesh sizes in CEN nets (a) produces a relatively even frequency distribution of fish lengths and weights across all lakes. Sampling according to the VERT protocol returned a multimodal distribution. Plots cropped at 400 mm.
community composition between the protocols were not significantly related to any aspects of lake morphology.
4. Discussion The CEN and VERT protocols were both developed to characterize whole-lake fish communities (Appelberg, 2000; Comité Européen de Normalisation, 2005; Degiorgi et al., 1993, 2001) and our results showed that fish abundance, biomass and species richness were indeed correlated between the protocols across the surveyed lakes. Despite this correspondence however, we also revealed considerable variability in the rank-ordering and magnitude of difference between lakes for the various fish metrics. Differences between the protocols in the proportion of net area sampling littoral, benthic and pelagic habitats appeared to influence the way each protocol represented the CPUE and species composition of the lake. The higher proportion of effort in benthic habitats under the CEN protocol is reflected in the high proportion of benthic and demersal fish species. In particular, the proportion of perch was consistently higher in CEN than VERT nets. The data presented here suggest that analyses of fish abundance data collected using the CEN protocol will be heavily influenced by perch as this species constituted 50–75% of the community in most lakes (Fig. 3c). The higher level of effort in benthic habitats is also reflected by the CEN protocol catching the only individuals of several strictly benthic species including stone loach, spiny loach, Padanian goby and the freshwater blenny. The exclusive capture of these species in CEN nets perhaps reflects that some benthic species are found in very low numbers and require a large net area in contact with the lake floor to detect them. An alternative explanation is that many of the CEN-exclusive species possessed long thin body morphologies such that they perhaps require a large amount of effort before an individual of these species is caught and retained in the net.
327
Fig. 7. Under the VERT protocol (black circles), the proportion of the net area (i.e. sampling effort) allocated to pelagic waters (defined as >3 m from the lake floor) increases in correspondence with the proportion of the lake occupied by this habitat. The CEN protocol (gray circles) requires one column of pelagic nets at the deepest point of the lake; consequently the proportion of netting effort occurring in pelagic waters remains approximately stable with increasing proportional volume of pelagic water.
The VERT protocol returned a higher proportion of whitefish across all lakes reflecting the higher proportion of sampling effort (i.e. net area) in pelagic habitats, especially in large and deep lakes. Pelagic habitats of the large deep lakes surveyed in this study support commercial fisheries (mainly whitefish; Gerdeaux et al., 2006) and the high proportional volume of pelagic habitat in these lakes (Fig. 7) reflects its importance to whole-lake ecosystem functioning. Depending on the trophic condition of the lake, deep pelagic habitats supported considerable fish biomass (e.g. whitefish in Lakes Walen and Brienz) or were almost fishless (e.g. hypoxic conditions of deep pelagic waters in Lake Lugano). Gillnet sampling to characterize the fish community therefore needs sufficient effort in pelagic habitats in order to adequately represent pelagic communities (Jurvelius et al., 2011; Specziar et al., 2009) and their contribution to the size and composition of fish community for the whole lake (Diekmann et al., 2005). Of the two protocols tested here, the VERT protocol is most likely to produce an estimate of whole-lake fish community which is closer to reality because effort, in terms of net area, is more evenly distributed across the entire volume of the lake. The proportion of netting effort in pelagic waters under the VERT protocol increased in correspondence with the proportion by volume of this habitat within the lake (Fig. 7). On the other hand, the CEN protocol required only one column of pelagic nets at the deepest point of the lake. Consequently the proportion of netting effort occurring in pelagic waters remained approximately stable with increasing lake volume under the CEN protocol. To accommodate this artifact in analyses based on the CEN protocol, Lauridsen et al. (2008) proposed weighting whole-lake estimates of CPUE by the relative volume of each habitat compartment and the proportion of effort in the habitat. This was intended to shift CPUE closer to reflecting the true catch rate and species composition across a lake. However since CEN nets are set only at one location in the lake and do not accommodate horizontal spatial variation, estimates of pelagic CPUE may not be accurate across the lake. Any discrepancies from the true catch rate will be magnified when calculating a
328
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
weighted whole-lake CPUE owing to the high proportional volume of pelagic waters. Other studies also generally report agreement when the CEN protocol is compared against alternative methods to characterize lacustrine fish communities. Specziar et al. (2009) conducted a comparison between nets that simultaneously sampling the whole water column and CEN gillnets in a large, shallow lake and concluded that both netting methods provided similar species richness, abundance and biomass for benthic, but not pelagic species. CEN benthic nets alone were sufficient to identify species presence in eleven French lakes, although inclusion of pelagic nets was necessarily to accurately represent relative fish abundance throughout the lake (Deceliere-Vergès and Guillard, 2008). Achleitner et al. (2012) agreed that, in 14 Austrian lakes, CEN benthic nets detected far more species than pelagic nets, but netting effort in open water was required to detect several exclusively pelagic species and to better represent whole-lake fish community. Other authors have reported correspondence between CEN gillnets and hydroacoustics within (Yule et al., 2013) and between (Emmrich et al., 2012) lakes. However, Emmrich et al. (2010) added the caveat that the CEN protocol does not provide a reliable estimate of fish density in deep lakes. Taking into account the results of our study and the published research, we suggest that intensive gillnet sampling in littoral and benthic habitats, ideally in combination with electrofishing ˝ et al., 2009), (Achleitner et al., 2012; Diekmann et al., 2005; Eros is an effective method to generate a species list for a lake. Netting effort in open-water habitats is essential to detect predominantly pelagic species. Where the objective of sampling is to represent the abundance or biomass of the fish community throughout the lake, pelagic habitats should be sampled with sufficient net surface area to reflect their volumetric contribution to the lake in combination with spatial replication to accommodate vertical and horizontal gradients. An alternative approach is to calculate a volume-weighted CPUE (e.g. Lauridsen et al., 2008). In this case, sampling should aim to provide an accurate representation of the fish assemblage within volume-based habitat compartments, which can then be adjusted by the volumetric contribution of each compartment to reflect fish communities throughout the lake. Ultimately, for the purposes of determining current ecological status and monitoring changes in the fish community of a lake through time, any thoughtfully designed, statistically robust and consistently applied sampling protocol is sufficient. A manager must be aware however that the scope of an assessment is constrained to the sampled habitats. Also, if one habitat is sampled more intensively than others, then an assessment of ecological status will be disproportionately weighted toward the ecological conditions of this habitat. The results of our study suggest that the CEN provides a sound basis for ecological assessments of the fish community, particularly where size-structure is important, however the heavy benthic emphasis means that outcomes may be disproportionately influenced by ecological conditions in benthic habitats.
5. Conclusion Use of the CEN protocol has the benefit of a massive database of lakes against which results based on CEN sampling can be compared. This is invaluable for comparative ecological assessment and management applications which require comparison to lakes with near-natural conditions. However the heavy benthic emphasis of the CEN protocol means that, in its current form, it may not be the best sampling protocol for many research questions which require reliable estimates of fish density/biomass or species-specific CPUE throughout the entire waterbody. In this regard, the vertical
netting protocol, which simultaneously samples the entire water column within multiple depth zones, may be more appropriate for calculating whole-lake estimates of CPUE and to determine shifting fish-habitat preferences along physico-chemical gradients.
Acknowledgements Funding was provided for field surveys by Swiss Cantons, Regione Lombardia (Italy), the French recreational fishing federation, Swiss Federal office of the Environment, University of Besanc¸on and Eawag. We acknowledge the Swiss Fisheries wardens and the French “Office National des eaux et des Milieux aquatiques” (ONEMA) for field assistance and for supporting the project. We also thank the many students, civil servants, helpers, and Eawag technical staff who assisted in the field. This research benefited from the thoughtful and constructive comments of two anonymous reviewers.
References Achleitner, D., Gassner, H., Luger, M., 2012. Comparison of three standardised fish sampling methods in 14 alpine lakes in Austria. Fish. Manage. Ecol. 19, 352–361. Appelberg, M., 2000. Swedish Standard Methods for Sampling Freshwater Fish with Multi-mesh Gillnets, Fiskeriverket Information, 1. Bartoo, N.W., Hansen, R.G., Wydoski, R.S., 1973. A portable vertical gill-net system. Prog. Fish Cult. 35, 231–233. Brucet, S., Pédron, S., Mehner, T., Lauridsen, T.L., Argillier, C., Winfield, I.J., Volta, P., Emmrich, M., Hesthagen, T., Holmgren, K., Benejam, L., Kelly, F., Krause, T., Palm, A., Rask, M., Jeppesen, E., 2013. Fish diversity in European lakes: geographical factors dominate over anthropogenic pressures. Freshw. Biol. 58, 1779–1793. Comité Européen de Normalisation, 2005. Water Quality – Sampling of Fish with Multi-mesh Gillnets (CEN 14757). European Committee for Standardization, Brussels. Deceliere-Vergès, C., Argillier, C., Lanoiselée, C., De Bortoli, J., Guillard, J., 2009. Stability and precision of the fish metrics obtained using CEN multi-mesh gillnets in natural and artificial lakes in France. Fish. Res. 99, 17–25. Deceliere-Vergès, C., Guillard, J., 2008. Assessment of the pelagic fish populations using CEN multi-mesh gillnets: consequences for the characterization of the fish communities. Know. Manage. Aquat. Ecosyst. 389, 1–16. Degiorgi, F., Grandmottet, J.P., Chanteloube, P., Pardon, C., Rousselet, A., Suat, J.F., Vandelle, J.P., 1993. Relations entre la topographie aquatique et l’organisation spatiale de l’ichtyofaune lacustre: définition des modalités spatiales d’une stratégie de prélèvement reproductible. Bull. Fr. de Pisc. 329, 199–220. Degiorgi, F., Grandmottet, P.J., Raymond, J.C., Rivier, J., 2001. Échantillonnage de l’ichtyofaune lacustre: engins passifs et protocole de prospection. In: Gerdeaux, D. (Ed.), Gestion piscicole des grands plans d’eau. Quae, Paris, pp. 151–182. Diekmann, M., Brämick, U., Lemcke, R., Mehner, T., 2005. Habitat-specific fishing revealed distinct indicator species in German lowland lake fish communities. J. Appl. Ecol. 42, 901–909. Emmrich, M., Helland, I.P., Busch, S., Schiller, S., Mehner, T., 2010. Hydroacoustic estimates of fish densities in comparison with stratified pelagic trawl sampling in two deep, coregonid-dominated lakes. Fish. Res. 105, 178–186. Emmrich, M., Winfield, I.J., Guillard, J., Rustadbakken, A., VergÈS, C., Volta, P., Jeppesen, E., Lauridsen, T.L., Brucet, S., Holmgren, K., Argillier, C., Mehner, T., 2012. Strong correspondence between gillnet catch per unit effort and hydroacoustically derived fish biomass in stratified lakes. Freshw. Biol. 57, 2436–2448. ˝ T., Specziár, A., Bíró, P., 2009. Assessing fish assemblages in reed habitats of a Eros, large shallow lake – a comparison between gillnetting and electric fishing. Fish. Res. 96, 70–76. Gerdeaux, D., Anneville, O., Hefti, D., 2006. Fishery changes during reoligotrophication in 11 peri-alpine Swiss and French lakes over the past 30 years. Acta Oecol.: Int. J. Ecol. 30, 161–167. Horak, D.L., Tanner, H.A., 1964. The use of vertical gill nets in studying fish depth distribution, Horsetooth Reservoir, Colorado. Trans. Am. Fish. Soc. 93, 137–145. Jurvelius, J., Kolari, I., Leskelä, A., 2011. Quality and status of fish stocks in lakes: gillnetting, seining, trawling and hydroacoustics as sampling methods. Hydrobiologia 660, 29–36. Lackey, R.T., 1968. Vertical gill nets for studying depth distribution of small fish. Trans. Am. Fish. Soc. 97, 296–299. Lauridsen, T.L., Landkildehus, F., Jeppesen, E., Jørgensen, T.B., Søndergaard, M., 2008. A comparison of methods for calculating Catch Per Unit Effort (CPUE) of gill net catches in lakes. Fish. Res. 93, 204–211. Lynch, W.E., Gerber, J.M., Johnson, D.L., Reutter, J.R., 1989. Management brief: a quickly deployed vertical gill-net system. N. Am. J. Fish. Manage. 9, 119–121.
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329 Mehner, T., Diekmann, M., Brämick, U., Lemcke, R., 2005. Composition of fish communities in German lakes as related to lake morphology, trophic state, shore structure and human-use intensity. Freshw. Biol. 50, 70–85. Mehner, T., Holmgren, K., Lauridsen, T.L., Jeppesen, E., Diekmann, M., 2007. Lake depth and geographical position modify lake fish assemblages of the European ‘Central Plains’ ecoregion. Freshw. Biol. 52, 2285–2297. PRIMER-E, 2008. Primer 6 Version 6.1.11 and PERMANOVA+ Version 1.0.1. Plymouth Marine Laboratory, Roborough, Plymouth, UK. R Core Team, 2014. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ˝ T., György, Á.I., Tatrai, I., Bíró, P., 2009. A comparison between Specziar, A., Eros, the benthic Nordic gillnet and whole water column gillnet for characterizing
329
fish assemblages in the shallow Lake Balaton. Ann. Limnol.: Int. J. Limnol. 45, 171–180. Vander Zanden, M.J., Vadeboncoeur, Y., 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83, 2152–2161. Wickham, H., 2007. Reshaping data with the {reshape} package. J. Stat. Softw. 21, 1–20. Wickham, H., 2009. ggplot2: Elegant Graphics for Data Analysis. Springer, New York. Yule, D.L., Evrard, L.M., Cachera, S., Colon, M., Guillard, J., 2013. Comparing two fish sampling standards over time: largely congruent results but with caveats. Freshw. Biol. 58, 2074–2088.
Contents lists available at ScienceDirect
Fisheries Research journal homepage: www.elsevier.com/locate/fishres
Evaluating gillnetting protocols to characterize lacustrine fish communities Timothy J. Alexander a,b,∗ , Pascal Vonlanthen a,b , Guy Periat a , Franc¸ois Degiorgi c , Jean-Claude Raymond d , Ole Seehausen a,b a Department of Fish Ecology and Evolution, Centre of Ecology, Evolution and Biogeochemistry, EAWAG Swiss Federal Institute of Aquatic Science and Technology, Seestrasse 79, CH-6047 Kastanienbaum, Switzerland b Division of Aquatic Ecology and Evolution, Institute of Ecology and Evolution, University of Bern, Baltzerstrasse 6, CH-3012 Bern, Switzerland c CNRS-University of Franche-Comte/UMR 6249 Chrono-environment, La Bouloie, F-25030 Besanc¸on Cedex, France d ONEMA French National Agency for Water and Aquatic Environment, Unité Spécialisée Milieux Lacustres, Pisciculture de Rives 13, Quai de Rives 74200, Thonon-les-Bains, France
a r t i c l e
i n f o
Article history: Received 15 April 2014 Received in revised form 12 July 2014 Accepted 15 August 2014 Handling Editor George A. Rose Keywords: Multimesh gillnets CEN standard Vertical net protocol Perch Coregonus
a b s t r a c t Ecological research and monitoring of lacustrine ecosystems often requires a whole-lake assessment of fish communities. Gillnet sampling offers an efficient means of estimating abundance, biomass and fish community composition. However the choice of gillnet sampling protocol may influence lake characterization via physical properties of the nets and allocation of sampling effort between littoral, benthic and pelagic habitats. This paper compares two commonly used, whole-lake sampling protocols applied across 17 prealpine, subalpine and alpine European lakes ranging widely in size, depth and altitude to determine their relative strength for research and management applications. Effort-corrected estimates of abundance, biomass and species richness were correlated between the protocols and both distinguished the trout-dominated alpine communities from subalpine and prealpine lakes dominated by whitefish and perch. A considerable amount of variance remained unexplained between the two protocols however, which seemed to correspond with differences in the proportion of effort among benthic and pelagic habitats. We suggest that both the European standard (CEN) and vertical (VERT) netting protocols are suitable for assessing ecological status and monitoring changes in lake fish communities through time. However the details of each protocol should be kept in mind when comparing fish communities between lakes. Mesh sizes used in CEN nets produce a more even size frequency distribution, suggesting that this protocol is most appropriate for assessing size structure of fish assemblages. The high proportion of netting effort in benthic habitats shallower than 70 m depth under the CEN protocol means that, particularly in larger lakes, outcomes will be disproportionately influenced by the ecological condition of this habitat. The VERT protocol presumably provides a more accurate estimate of whole-lake CPUE and community composition because effort, in terms of net area, is more evenly distributed across the entire volume of the lake. This is particularly important in large and deep lakes where pelagic habitats occupy a high proportion of the lake volume. © 2014 Elsevier B.V. All rights reserved.
1. Introduction The European Union Water Framework Directive (WFD) requires that all member countries characterize, assess and if
∗ Corresponding author at: Department of Fish Ecology and Evolution, Centre of Ecology, Evolution and Biogeochemistry, EAWAG Swiss Federal Institute of Aquatic Science and Technology, Seestrasse 79, CH-6047 Kastanienbaum, Switzerland; Division of Aquatic Ecology and Evolution, Institute of Ecology and Evolution, University of Bern, Baltzerstrasse 6, CH-3012 Bern, Switzerland. Tel.: +41 58 765 2202. E-mail address: [email protected] (T.J. Alexander). http://dx.doi.org/10.1016/j.fishres.2014.08.009 0165-7836/© 2014 Elsevier B.V. All rights reserved.
necessary, improve the ecological status of their freshwater ecosystems by 2015. Fish, benthic invertebrates, phytoplankton and other aquatic flora form the basis of the biological component of the assessment. In particular, the longevity of many fish species makes them robust and temporally integrated indicators of ecosystem status (Vander Zanden and Vadeboncoeur, 2002). The WFD requires assessment of three aspects of the fish community: whole-lake estimate of catch per unit effort (CPUE), species composition and the age structure of fish assemblages. A multimesh gillnetting protocol has been adopted across Europe to conduct assessments under the WFD and facilitate intercalibration of quality thresholds between countries (hereafter refered to as CEN protocol;
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Appelberg, 2000; Comité Européen de Normalisation, 2005). The CEN protocol divides each lake into benthic and pelagic zones and samples these zones using gillnets deployed horizontally in the water column or on the lakebed. Netting effort is allocated in the benthic zone by randomly sampling within defined depth strata with replication determined according to the maximum depth and area of a lake. Pelagic netting effort is also conducted within depth strata but only within one column at the deepest point of the lake. A second gillnetting protocol has been used widely to survey whole-lake fish communities for management and research in eastern France. This protocol describes the use of vertically oriented gillnets that simultaneously survey from the water surface to the lake floor. The vertical nets were originally developed in the USA for studying depth distribution of fishes in lakes and reservoirs (Horak and Tanner, 1964; Lackey, 1968; Lynch et al., 1989). The vertical netting protocol (hereafter referred to as VERT protocol), introduced by Degiorgi et al. (1993) and amended in Degiorgi et al. (2001), describes the application of vertical nets to sample wholelake fish communities. Under the VERT protocol, replicate gillnet series are deployed within each type of littoral and offshore habitat present in the lake. Littoral habitats (<5 m deep) are classified based on substrate composition and particle size, macrophyte morphology and density, and proximity to an inwardly or outwardly flowing watercourse. Two sublittoral and three deep pelagic habitats are also defined relative to the maximum depth of the lake. The widespread use of the CEN sampling protocol provides an unprecedented opportunity for community- and macro-ecological research into natural and anthropogenic conditions influencing lacustrine fish communities (e.g. Brucet et al., 2013; Mehner et al., 2005, 2007). However, meaningful interpretation of the results of such research requires an understanding of the strengths, weaknesses and idiosyncrasies of the data on which it is based. Multiple authors have commented that the CEN protocol under-represents pelagic species in assessments of whole-lake fish communities (Achleitner et al., 2012; Deceliere-Vergès et al., 2009; DeceliereVergès and Guillard, 2008; Diekmann et al., 2005). Other authors have noted that sampling only at the deepest point of the lake probably overlooks horizontal variation in fish communities living in pelagic habitats (Lauridsen et al., 2008; Specziar et al., 2009). These issues are particularly important in large and deep lakes where pelagic habitats constitute the vast majority of the lake volume. This paper therefore aims to assess and compare the utility of CEN and VERT protocols for management and research into whole-lake fish
321
communities. The investigation will focus on the level of correspondence and discrepancy between the protocols and their relative strengths in providing reliable estimates of whole-lake fish CPUE, size structure, species richness and community composition.
2. Methods 2.1. Fish sampling Comparison of CEN and VERT protocols was conducted by sampling fish communities using both protocols in 17 lakes across eastern France, Switzerland and northern Italy (Fig. 1). Physical characteristics of the sampled lakes are provided in Table 1. Surveys were conducted between August and October 2010–2013. All nets were set before dusk and retrieved approximately 14 h later. Due to the logistics of the large-scale field schedule, soak times occasionally varied from this target. To reduce the influence of variation in soak time, catches were standardized to 14 h by dividing by the soak time (in decimal hours) and multiplying by 14. Biomass and number of fish (abundance) were also standardized by the area of net. The resulting biomass per unit effort (BPUE) and number of fish per unit effort (NPUE) therefore reflected fish catches per square meter per 14 h. Catch per unit effort (CPUE) is used in this paper as the collective term for both BPUE and NPUE. Catch rate and CPUE are also used interchangeably. Use of the terms biomass and abundance always refers to effort corrected values. Sampling under the CEN protocol was conducted according to the European Committee for Standardization standard EN14757:2005 (Comité Européen de Normalisation, 2005). Briefly, benthic netting effort was located randomly within defined depth strata with replication determined by the maximum depth and area of a lake. Pelagic nets were set suspended in the water column within the same depth strata at the deepest point of the lake over consecutive nights. Benthic habitats were sampled with Nordic type gillnets consisting of a series of contiguous panels of twelve mesh sizes following a geometric series: 5, 6.25, 8, 10, 12.5, 15.5, 19.5, 24, 35, 43, and 55 mm (measured knot to knot). Each mesh panel was 1.5 m high and 2.5 m wide. The combined multimesh net was 1.5 m high and 30 m long. Pelagic nets consisted of the same mesh sizes, minus the 5 mm mesh. Panels in pelagic nets were 6 m high making the combined net 6 m high and 27.5 m long.
Fig. 1. Map of sampled lakes across Switzerland, eastern France and northern Italy.
322
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Table 1 Morphological characteristics of the study lakes. Lake
Altitude (m)
Area (km2 )
Max depth (m)
Mean depth (m)
Poschiavo Sils Brenet Remoray Chalain Saint-Point Joux Hallwil Morat Zug Neuchatel Geneva Walen Brienz Thun Lugano Maggiore
962 1797 1002 850 486 850 1004 449 429 414 429 372 419 564 558 271 194
2.0 4.1 0.8 1.0 2.3 5.2 9.5 10.3 22.8 38.3 218.3 582.0 24.2 29.8 48.3 48.7 212.5
85 71 18 27 32 43 32 48 45 198 152 310 151 261 217 288 372
61 35 9 14 17 16 16 28 23 84 64 152 105 173 136 134 177
Vertical net sampling was conducted according to the protocol proposed by Degiorgi et al. (1993) and amended in Degiorgi et al. (2001). The VERT protocol used gillnets series (separate nets for different mesh sizes) to simultaneously sample from the water surface to the lake floor. Each net in the series was 2 m wide and each series consisted of mesh sizes 10, 15, 20, 30, 40, 50, and 60 mm. Sampling effort was allocated according to the littoral and deep habitats present in the lake. Littoral habitats were mapped in each lake prior to the fish sampling event. Habitats were defined according to the dominant substrate composition, vegetation and proximity to a river or stream (Table 2). Deeper habitats were defined relative to the maximum depth of the lake. Gillnets were deployed at least three times in each littoral and deep habitat types present in the lake. Setting locations were randomly selected from within the available area of each habitat. The VERT protocol used different types of nets to sample littoral and deep habitats. Littoral habitats (to 5 m deep) were sampled with 2 m wide mesh panels for each mesh size deployed on the same lead and float line. A 2 m gap separated adjacent panels. The
Table 2 Habitat categories used to allocate sampling effort in the vertical netting protocol. Dimensions of the rock and sediment refer to the most common maximum dimension (length, width, height) of lithic particles at the site. Zmax is the maximum depth of the lake in meters. Some pelagic habitats may not be present in the lake if Zmax is less than 40 m. Sites with emergent macrophytes were further divided into dense and scattered based on a threshold of 10 cm between stems. Littoral habitats (<5 m) Emergent macrophytes Submerged macrophytes Floating macrophytes Woody debris or trees Leaf litter Rock slab or ledge Rocks or boulders Cobbles Gravel Gravel and cobbles Sand Fine sediment Out-flowing stream or river In-flowing stream or river
Details
Roots or branches in or touching water Solid rock/bedrock, no interstitial space >200 mm 20–200 mm 2–20 mm Mixture of gravel and cobbles 0.5–2 mm <0.5 mm
Deep habitats (>5 m)
Details
Sublittoral Deep sublittoral Min pelagic Med pelagic Max pelagic
5–10 m 10 m–0.3Zmax 0.3Zmax –0.6Zmax 0.6Zmax –0.9Zmax 0.9Zmax –Zmax
Volume (GL) 110.8 129.7 5.2 13.7 46.3 9.4 128.3 289.5 437.8 3,202.8 13,935.5 79,016.8 2,466.0 5,161.8 6,434.6 5,705.3 12,957.3
nets ranged in height from 1 m to 5 m and were selected to match the depth of the sampling site, ensuring that the net remains vertical in the water column. Gillnet series to sample deeper habitats consisted of separate nets for each mesh size, deployed as close to each other as weather conditions allowed. The nets were rolled onto lengths of PVC pipe (diameter 150 mm) filled with plastic bottles and waterproof expanding foam (Bartoo et al., 1973). The pipes provided floatation and allowed the nets to be deployed and retrieved using a customized winch. Two-meter wide aluminum spacers (10 mm hollow tubes), placed every 10–15 m along the net, ensured that it retained the appropriate width. Iron weights and a 25 mm diameter steel bar across the width at the base of the net provided the anchors to maintain the position of the lower end on the lake floor. Fish caught in both gillnet protocols were identified to species level; however due taxonomic uncertainty and difficulty to reliably distinguish between some species in the field, several taxa were aggregated to genera for analyses of abundance and biomass. These included: Coregonus spp. (C. albellus, C. alpinus, C. bondella, C. lavaretus, C. palea; referred to in this paper as whitefish), Rutilus spp. (R. rutilus, R. aula, R. pigus; referred to in this paper as roach), Alburnus spp. (A. alburnus, A. alborella), Scardinius spp. (S. hesperidicus, S. erythrophthalmus), Phoxinus spp. (P. phoxinus, P. lumaireul), Salmo spp. (S. trutta, S. marmorata, S. cenerinus, S. labrax), and Salvelinus spp. (S. umbla species complex).
2.2. Statistical analyses Protocols were compared by focusing on the three aspects of the fish community required for assessments under the WFD: wholelake estimate of catch per unit effort, species composition and the size structure of fish assemblages. Calculation of whole-lake CPUE differed slightly between the protocols. Mean CPUE, according to the CEN protocol (Comité Européen de Normalisation, 2005), involved calculating the catch rate for each individual net (abundance or biomass divided by net area, after correcting for soak time) and then calculating the mean CPUE across all nets set in the lake. Global CPUE, under the VERT protocol (Degiorgi et al., 2001), involved summing soak-time corrected fish abundance or biomass across all nets set in a lake and dividing by the summed surface area of all nets set in the lake. Mean CPUE reflects the average catch rate across all the nets set in the lake. Global CPUE describes the average catch rate across every square meter of net set in the lake. Note that mean CPUE and global CPUE give the same whole-lake estimate if net area is the same across all nets used in the calculation.
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fish size frequency distributions were used as a proxy for age structure in this study. Length and biomass frequencies distributions of fish collected under each protocol were compared without correcting for net selectivity. Fish community composition was summarized by species richness and Shannon diversity based on NPUE. We also compared between the protocols the relative abundance and biomass of common species: perch (Perca fluviatilis), roach (Rutilus spp.) and whitefish (Coregonus spp.). The correspondence between the protocols for all univariate metrics was determined using correlation analysis (Wilcoxon signed rank test). Multivariate differences between protocols in community composition were visualized by overlaying ellipses to distinguish 20, 40, and 60% clusters on non-metric multidimensional scaling (MDS) ordinations. MDS plots were based on Bray–Curtis resemblance on relativized species catches calculated by dividing the catch for each species by the summed CPUE for all species in the lake. One-way permutation analysis of variance (PERMANOVA; Type III (partial) sums of squares with unrestricted permutation of raw data) was used to test for differences between protocols across all lakes. Similarity percentage (SIMPER) identified which species contributed to differences between the protocols. All multivariate analyses were conducted in PRIMER 6 (PRIMER-E, 2008). To determine if the degree of similarity between the protocols was related to the physical characteristics of the lake (i.e.
323
morphology and altitude), the residuals between the CEN and VERT estimates of CPUE in each lake were regressed against the lake characteristics. The same analysis was also conducted for community composition using the Bray–Curtis similarity (common species; relativized) between the community composition of each lake as characterized by the VERT and CEN protocols. Values for physical characteristics were natural log transformed prior to analysis to stabilize their variance. A Bonferroni correction of ˛/n (where n is the number of tests), applied to accommodate the multiple comparisons, reduced ˛ to 0.01 for these tests. All univariate analyses were conducted in R (R Core Team, 2014) using ‘base’ package for univariate statistical analyses and ‘vegan’ for diversity calculations. Data were manipulated using ‘reshape’ (Wickham, 2007) and figures generated using the ‘base’ and ‘ggplot2’ (Wickham, 2009) packages. 3. Results 3.1. Catch per unit effort Fish biomass (BPUE) and abundance (NPUE) of each lake were significantly correlated between protocols (Fig. 2a and b; Wilcoxon signed rank test BPUE: V = 171, p < 0.0001, R2 = 0.66; NPUE: V = 171, p = 0.0001, R2 = 0.42). Despite the correlation, there were notable differences in the rank order of lakes between the protocols,
Fig. 2. Comparing estimates of whole-lake (a) BPUE, (b) NPUE, (c) taxon richness and (d) species diversity between CEN and VERT protocols. Dashed line indicates 1:1 agreement between protocols.
324
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 3. Percentage composition of fish communities by biomass sampled with (a) CEN and (b) VERT protocols, and by abundance sampled with (c) CEN and (d) VERT protocols.
particular in terms of abundance. For example, sampling with the VERT protocol suggested that Lake Brenet had the highest abundance and biomass of all lakes. Sampling with the CEN protocol placed the same lake at 6th highest for biomass and 9th highest for abundance. Abundance was also consistently higher under the CEN protocol with estimates ranging from 2 to 15 times greater than those of the VERT protocol in the same lake. Biomass was also higher in all lakes sampled by the CEN protocol but the difference was less severe (1.5–4.5 times). 3.2. Species composition Estimates of taxon richness for each lake were very similar between the protocols (Fig. 2c; Spearman’s rank correlation: S = 82.5, p < 0.0001, R2 = 0.93), with identical numbers of taxa recorded in 7 out of the 17 lakes. Species diversity was also weakly correlated between the protocols (Fig. 2d; Shannon diversity based on NPUE: Wilcoxon signed rank test: V = 149, p < 0.01, R2 = 0.09). The VERT protocol tended to estimate a more even community composition by abundance (Pielou’s evenness; data not shown) which resulted in higher estimates of species diversity than from CEN in 12 lakes. Fish community composition by biomass and abundance in fifteen of the sampled lakes was dominated by perch (P. fluviatilis; Percidae), whitefish (Coregonus spp; Salmonidae) and the Cyprinids roach (Rutilus spp; Cyprinidae) and chub (Squalius spp; Cyprinidae) in ratios that varied for each lake between the protocols (Fig. 3). The composition of the alpine lakes Sils and Poschiavo was strikingly different, where abundance and biomass were dominated by trout (Salmo spp; Salmonidae) and char (Salvelinus spp; Salmonidae), including introduced lake trout (Salvelinus namaycush; Salmonidae). The proportion of abundance and biomass of the three most frequently caught fish taxa was significantly correlated between protocols (excluding abundance of roach): perch (Fig. 4a; Wilcoxon signed rank test: biomass V = 0, p < 0.001, R2 = 0.68; abundance V = 8, p < 0.0001, R2 = 0.28), roach (Fig. 4b; biomass V = 20, p < 0.05, R2 = 0.28; abundance V = 85, p = 0.404, R2 = 0.28), and whitefish
(Fig. 4c; biomass V = 117, p < 0.01, R2 = 0.51; abundance V = 134, p < 0.0001, R2 = 0.56). The biomass of perch relative to other species was particularly strongly correlated between protocols but was consistently higher under the CEN protocol in all lakes. Roach was also recorded in higher proportions by biomass under the CEN protocol for the majority of lakes (13 of 15), while differences between protocols in proportion by abundance of this genus were less clear cut (CEN > VERT for 7 of 15 lakes). The proportion of whitefish was also very strongly correlated between protocols for abundance and biomass. The VERT protocol estimated a higher proportion of whitefish in almost every lake (abundance 14 of 15 lakes; biomass 10/15 lakes). Aside from perch and roach, burbot (Lota lota) also, tended to be caught more in CEN nets with higher relative biomass in all nine lakes where it was recorded (relative abundance CEN > VERT for 7 of 9 lakes). In two lakes, burbot were exclusively captured in CEN nets. Several species were also exclusively caught in one or the other protocol. Nets deployed under the CEN protocol captured the only representatives 7 species: stone loach (Barbatula barbatula; 1 lake), spiny loach (Cobitis sp.; 1 lake), southern pike (Esox cisalpinus; 1 lake), three-spined stickleback (Gasterosteus aculeatus; 1 lake), Padanian goby (Padogobius bonelli; 1 lake), freshwater blenny (Salaria fluviatilis; 2 lakes) and vairone (Telestes muticellus; 1 lake). In contrast, only one species, Italian barbel Barbus blebejus (1 lake) was caught exclusively in the VERT nets. Overall however, the number of occasions when one protocol detected a species in a lake that was overlooked by the other was approximately similar between protocols (24 for CEN and 28 for VERT). Overall, the CEN and VERT protocols characterized significantly different fish community composition by abundance (PERMANOVA: pseudo-F = 5.45, p > 0.001 with 999 unique permutations) and biomass (pseudo-F = 5.56, p > 0.003 with 998 unique permutations). Perch alone contributed over 30% of the difference in abundance between protocols, with lesser contributions from roach (Rutilus spp.), whitefish (Coregonus spp.), trout (Salmo spp.) and rudd (Scardinius spp.; SIMPER analysis; Table 3). Differences between the protocols by biomass were caused by a wider spectrum of species, but also with strong contributions from perch,
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
325
Fig. 4. Comparing the abundance and biomass of common species (a) perch, (b) whitefish, (c) roach and between protocols. Dashed line shows 1:1 agreement.
roach, whitefish, trout and rudd. Both protocols were sensitive to major geographical differences in fish assemblage composition among lakes (i.e. between trout-dominated Sils and Poschiavo and other lakes), but were rarely in complete agreement as to
which of the lakes with more similar community composition clustered together (MDS; Fig. 5). In terms of biomass, both protocols identified that the fish community composition of whitefishdominated lakes, Thun and Walen, were distinct; however there
326
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 5. MDS plots show clustering among lakes according to Bray–Curtis similarity of fish assemblage composition for biomass as sampled by (a) CEN and (b) VERT protocols and for abundance by (c) CEN and (d) VERT protocols. The extreme dissimilarity between Sils and Poschiavo and the other lakes was consistent between protocols; consequently plots exclude these two lakes to reveal clusters among the more similar lakes. Ellipses describe 20, 40 and 60% similarity in assemblage composition (shown in green, blue and red respectively).
Table 3 Similarity percentages analysis identified that perch (Perca fluviatilis), whitefish (Coregonus spp.), roach (Rutilus spp.), rudd (Scardinius spp.) and trout (Salmo spp.) contributed most strongly to differences between the protocols in fish community composition by (a) biomass and (b) abundance. (a) Species
Perca fluviatilis Coregonus spp. Scardinius spp. Rutilus spp. Salmo spp. Squalius spp. Tinca tinca Salvelinus namaycush Leuciscus leuciscus Salvelinus spp. Sander lucioperca Esox lucius Abramis brama
Mean % biomass
Contribution %
CEN
VERT
27.3 9.8 8.9 19.3 7.9 3.6 4.8 2.9 3.1 1.8 1.6 2.4 1.6
9.5 17.4 15.8 12.3 8.6 12.0 7.3 2.8 1.1 1.7 1.3 2.2 1.1
16.2 12.6 12.5 11.2 10.7 8.1 5.9 3.9 2.6 2.4 1.9 1.8 1.8
(b) Species
Perca fluviatilis Rutilus spp. Coregonus spp. Salmo spp. Scardinius spp. Phoxinus spp. Gymnocephalus cernuus Salvelinus spp. Alburnus spp.
Mean % abundance CEN
VERT
55.6 18.0 5.3 6.0 1.3 2.8 2.2 2.1 1.2
32.2 22.3 15.7 7.5 8.2 1.2 1.9 1.6 2.4
Contribution %
31.4 15.5 14.5 10.6 7.0 3.3 3.2 3.0 2.5
was little similarity between protocols in clustering among the remaining lakes. Whitefish also seemed to separate the lake clusters in terms of abundance. Fish assemblage composition in Brienz and Walen was distinct from the other lakes under the CEN protocol, while the higher proportion of whitefish caught in the VERT nets meant that this group also included Thun and Chalain under the VERT protocol.
3.3. Size structure Fish length frequency distributions were significantly different between the two protocols (two-sample Kolmogorov–Smirnov test: D = 0.405, p < 0.001). The CEN nets returned a relatively smooth frequency distribution of fish lengths across all lakes (Fig. 6; mode = 75 mm, mean = 103 mm, range = 32–875 mm). The frequency distribution of the VERT protocol was multimodal with the most commonly caught fish length slightly larger at 85 mm (mean = 138 mm, range = 32–1008 mm). Secondary peaks in frequency occurred at 125 mm and 155 mm. The frequency distribution of fish biomass displayed similar patterns between the protocols (not shown).
3.4. Lake characteristics influencing differences between protocols Lake volume was the only aspect of lake morphology or altitude significantly related to differences in CPUE between the protocols. Estimates of biomass tended to be more similar between the protocols in lakes that were larger by volume (F1,15 = 10.8, p < 0.01, R2 = 0.42; however residuals). No such trend was seen in abundance (F1,15 = 3.49, p = 0.08, R2 = 0.19). Differences in the multivariate
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
Fig. 6. The geometric series of mesh sizes in CEN nets (a) produces a relatively even frequency distribution of fish lengths and weights across all lakes. Sampling according to the VERT protocol returned a multimodal distribution. Plots cropped at 400 mm.
community composition between the protocols were not significantly related to any aspects of lake morphology.
4. Discussion The CEN and VERT protocols were both developed to characterize whole-lake fish communities (Appelberg, 2000; Comité Européen de Normalisation, 2005; Degiorgi et al., 1993, 2001) and our results showed that fish abundance, biomass and species richness were indeed correlated between the protocols across the surveyed lakes. Despite this correspondence however, we also revealed considerable variability in the rank-ordering and magnitude of difference between lakes for the various fish metrics. Differences between the protocols in the proportion of net area sampling littoral, benthic and pelagic habitats appeared to influence the way each protocol represented the CPUE and species composition of the lake. The higher proportion of effort in benthic habitats under the CEN protocol is reflected in the high proportion of benthic and demersal fish species. In particular, the proportion of perch was consistently higher in CEN than VERT nets. The data presented here suggest that analyses of fish abundance data collected using the CEN protocol will be heavily influenced by perch as this species constituted 50–75% of the community in most lakes (Fig. 3c). The higher level of effort in benthic habitats is also reflected by the CEN protocol catching the only individuals of several strictly benthic species including stone loach, spiny loach, Padanian goby and the freshwater blenny. The exclusive capture of these species in CEN nets perhaps reflects that some benthic species are found in very low numbers and require a large net area in contact with the lake floor to detect them. An alternative explanation is that many of the CEN-exclusive species possessed long thin body morphologies such that they perhaps require a large amount of effort before an individual of these species is caught and retained in the net.
327
Fig. 7. Under the VERT protocol (black circles), the proportion of the net area (i.e. sampling effort) allocated to pelagic waters (defined as >3 m from the lake floor) increases in correspondence with the proportion of the lake occupied by this habitat. The CEN protocol (gray circles) requires one column of pelagic nets at the deepest point of the lake; consequently the proportion of netting effort occurring in pelagic waters remains approximately stable with increasing proportional volume of pelagic water.
The VERT protocol returned a higher proportion of whitefish across all lakes reflecting the higher proportion of sampling effort (i.e. net area) in pelagic habitats, especially in large and deep lakes. Pelagic habitats of the large deep lakes surveyed in this study support commercial fisheries (mainly whitefish; Gerdeaux et al., 2006) and the high proportional volume of pelagic habitat in these lakes (Fig. 7) reflects its importance to whole-lake ecosystem functioning. Depending on the trophic condition of the lake, deep pelagic habitats supported considerable fish biomass (e.g. whitefish in Lakes Walen and Brienz) or were almost fishless (e.g. hypoxic conditions of deep pelagic waters in Lake Lugano). Gillnet sampling to characterize the fish community therefore needs sufficient effort in pelagic habitats in order to adequately represent pelagic communities (Jurvelius et al., 2011; Specziar et al., 2009) and their contribution to the size and composition of fish community for the whole lake (Diekmann et al., 2005). Of the two protocols tested here, the VERT protocol is most likely to produce an estimate of whole-lake fish community which is closer to reality because effort, in terms of net area, is more evenly distributed across the entire volume of the lake. The proportion of netting effort in pelagic waters under the VERT protocol increased in correspondence with the proportion by volume of this habitat within the lake (Fig. 7). On the other hand, the CEN protocol required only one column of pelagic nets at the deepest point of the lake. Consequently the proportion of netting effort occurring in pelagic waters remained approximately stable with increasing lake volume under the CEN protocol. To accommodate this artifact in analyses based on the CEN protocol, Lauridsen et al. (2008) proposed weighting whole-lake estimates of CPUE by the relative volume of each habitat compartment and the proportion of effort in the habitat. This was intended to shift CPUE closer to reflecting the true catch rate and species composition across a lake. However since CEN nets are set only at one location in the lake and do not accommodate horizontal spatial variation, estimates of pelagic CPUE may not be accurate across the lake. Any discrepancies from the true catch rate will be magnified when calculating a
328
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329
weighted whole-lake CPUE owing to the high proportional volume of pelagic waters. Other studies also generally report agreement when the CEN protocol is compared against alternative methods to characterize lacustrine fish communities. Specziar et al. (2009) conducted a comparison between nets that simultaneously sampling the whole water column and CEN gillnets in a large, shallow lake and concluded that both netting methods provided similar species richness, abundance and biomass for benthic, but not pelagic species. CEN benthic nets alone were sufficient to identify species presence in eleven French lakes, although inclusion of pelagic nets was necessarily to accurately represent relative fish abundance throughout the lake (Deceliere-Vergès and Guillard, 2008). Achleitner et al. (2012) agreed that, in 14 Austrian lakes, CEN benthic nets detected far more species than pelagic nets, but netting effort in open water was required to detect several exclusively pelagic species and to better represent whole-lake fish community. Other authors have reported correspondence between CEN gillnets and hydroacoustics within (Yule et al., 2013) and between (Emmrich et al., 2012) lakes. However, Emmrich et al. (2010) added the caveat that the CEN protocol does not provide a reliable estimate of fish density in deep lakes. Taking into account the results of our study and the published research, we suggest that intensive gillnet sampling in littoral and benthic habitats, ideally in combination with electrofishing ˝ et al., 2009), (Achleitner et al., 2012; Diekmann et al., 2005; Eros is an effective method to generate a species list for a lake. Netting effort in open-water habitats is essential to detect predominantly pelagic species. Where the objective of sampling is to represent the abundance or biomass of the fish community throughout the lake, pelagic habitats should be sampled with sufficient net surface area to reflect their volumetric contribution to the lake in combination with spatial replication to accommodate vertical and horizontal gradients. An alternative approach is to calculate a volume-weighted CPUE (e.g. Lauridsen et al., 2008). In this case, sampling should aim to provide an accurate representation of the fish assemblage within volume-based habitat compartments, which can then be adjusted by the volumetric contribution of each compartment to reflect fish communities throughout the lake. Ultimately, for the purposes of determining current ecological status and monitoring changes in the fish community of a lake through time, any thoughtfully designed, statistically robust and consistently applied sampling protocol is sufficient. A manager must be aware however that the scope of an assessment is constrained to the sampled habitats. Also, if one habitat is sampled more intensively than others, then an assessment of ecological status will be disproportionately weighted toward the ecological conditions of this habitat. The results of our study suggest that the CEN provides a sound basis for ecological assessments of the fish community, particularly where size-structure is important, however the heavy benthic emphasis means that outcomes may be disproportionately influenced by ecological conditions in benthic habitats.
5. Conclusion Use of the CEN protocol has the benefit of a massive database of lakes against which results based on CEN sampling can be compared. This is invaluable for comparative ecological assessment and management applications which require comparison to lakes with near-natural conditions. However the heavy benthic emphasis of the CEN protocol means that, in its current form, it may not be the best sampling protocol for many research questions which require reliable estimates of fish density/biomass or species-specific CPUE throughout the entire waterbody. In this regard, the vertical
netting protocol, which simultaneously samples the entire water column within multiple depth zones, may be more appropriate for calculating whole-lake estimates of CPUE and to determine shifting fish-habitat preferences along physico-chemical gradients.
Acknowledgements Funding was provided for field surveys by Swiss Cantons, Regione Lombardia (Italy), the French recreational fishing federation, Swiss Federal office of the Environment, University of Besanc¸on and Eawag. We acknowledge the Swiss Fisheries wardens and the French “Office National des eaux et des Milieux aquatiques” (ONEMA) for field assistance and for supporting the project. We also thank the many students, civil servants, helpers, and Eawag technical staff who assisted in the field. This research benefited from the thoughtful and constructive comments of two anonymous reviewers.
References Achleitner, D., Gassner, H., Luger, M., 2012. Comparison of three standardised fish sampling methods in 14 alpine lakes in Austria. Fish. Manage. Ecol. 19, 352–361. Appelberg, M., 2000. Swedish Standard Methods for Sampling Freshwater Fish with Multi-mesh Gillnets, Fiskeriverket Information, 1. Bartoo, N.W., Hansen, R.G., Wydoski, R.S., 1973. A portable vertical gill-net system. Prog. Fish Cult. 35, 231–233. Brucet, S., Pédron, S., Mehner, T., Lauridsen, T.L., Argillier, C., Winfield, I.J., Volta, P., Emmrich, M., Hesthagen, T., Holmgren, K., Benejam, L., Kelly, F., Krause, T., Palm, A., Rask, M., Jeppesen, E., 2013. Fish diversity in European lakes: geographical factors dominate over anthropogenic pressures. Freshw. Biol. 58, 1779–1793. Comité Européen de Normalisation, 2005. Water Quality – Sampling of Fish with Multi-mesh Gillnets (CEN 14757). European Committee for Standardization, Brussels. Deceliere-Vergès, C., Argillier, C., Lanoiselée, C., De Bortoli, J., Guillard, J., 2009. Stability and precision of the fish metrics obtained using CEN multi-mesh gillnets in natural and artificial lakes in France. Fish. Res. 99, 17–25. Deceliere-Vergès, C., Guillard, J., 2008. Assessment of the pelagic fish populations using CEN multi-mesh gillnets: consequences for the characterization of the fish communities. Know. Manage. Aquat. Ecosyst. 389, 1–16. Degiorgi, F., Grandmottet, J.P., Chanteloube, P., Pardon, C., Rousselet, A., Suat, J.F., Vandelle, J.P., 1993. Relations entre la topographie aquatique et l’organisation spatiale de l’ichtyofaune lacustre: définition des modalités spatiales d’une stratégie de prélèvement reproductible. Bull. Fr. de Pisc. 329, 199–220. Degiorgi, F., Grandmottet, P.J., Raymond, J.C., Rivier, J., 2001. Échantillonnage de l’ichtyofaune lacustre: engins passifs et protocole de prospection. In: Gerdeaux, D. (Ed.), Gestion piscicole des grands plans d’eau. Quae, Paris, pp. 151–182. Diekmann, M., Brämick, U., Lemcke, R., Mehner, T., 2005. Habitat-specific fishing revealed distinct indicator species in German lowland lake fish communities. J. Appl. Ecol. 42, 901–909. Emmrich, M., Helland, I.P., Busch, S., Schiller, S., Mehner, T., 2010. Hydroacoustic estimates of fish densities in comparison with stratified pelagic trawl sampling in two deep, coregonid-dominated lakes. Fish. Res. 105, 178–186. Emmrich, M., Winfield, I.J., Guillard, J., Rustadbakken, A., VergÈS, C., Volta, P., Jeppesen, E., Lauridsen, T.L., Brucet, S., Holmgren, K., Argillier, C., Mehner, T., 2012. Strong correspondence between gillnet catch per unit effort and hydroacoustically derived fish biomass in stratified lakes. Freshw. Biol. 57, 2436–2448. ˝ T., Specziár, A., Bíró, P., 2009. Assessing fish assemblages in reed habitats of a Eros, large shallow lake – a comparison between gillnetting and electric fishing. Fish. Res. 96, 70–76. Gerdeaux, D., Anneville, O., Hefti, D., 2006. Fishery changes during reoligotrophication in 11 peri-alpine Swiss and French lakes over the past 30 years. Acta Oecol.: Int. J. Ecol. 30, 161–167. Horak, D.L., Tanner, H.A., 1964. The use of vertical gill nets in studying fish depth distribution, Horsetooth Reservoir, Colorado. Trans. Am. Fish. Soc. 93, 137–145. Jurvelius, J., Kolari, I., Leskelä, A., 2011. Quality and status of fish stocks in lakes: gillnetting, seining, trawling and hydroacoustics as sampling methods. Hydrobiologia 660, 29–36. Lackey, R.T., 1968. Vertical gill nets for studying depth distribution of small fish. Trans. Am. Fish. Soc. 97, 296–299. Lauridsen, T.L., Landkildehus, F., Jeppesen, E., Jørgensen, T.B., Søndergaard, M., 2008. A comparison of methods for calculating Catch Per Unit Effort (CPUE) of gill net catches in lakes. Fish. Res. 93, 204–211. Lynch, W.E., Gerber, J.M., Johnson, D.L., Reutter, J.R., 1989. Management brief: a quickly deployed vertical gill-net system. N. Am. J. Fish. Manage. 9, 119–121.
T.J. Alexander et al. / Fisheries Research 161 (2015) 320–329 Mehner, T., Diekmann, M., Brämick, U., Lemcke, R., 2005. Composition of fish communities in German lakes as related to lake morphology, trophic state, shore structure and human-use intensity. Freshw. Biol. 50, 70–85. Mehner, T., Holmgren, K., Lauridsen, T.L., Jeppesen, E., Diekmann, M., 2007. Lake depth and geographical position modify lake fish assemblages of the European ‘Central Plains’ ecoregion. Freshw. Biol. 52, 2285–2297. PRIMER-E, 2008. Primer 6 Version 6.1.11 and PERMANOVA+ Version 1.0.1. Plymouth Marine Laboratory, Roborough, Plymouth, UK. R Core Team, 2014. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ˝ T., György, Á.I., Tatrai, I., Bíró, P., 2009. A comparison between Specziar, A., Eros, the benthic Nordic gillnet and whole water column gillnet for characterizing
329
fish assemblages in the shallow Lake Balaton. Ann. Limnol.: Int. J. Limnol. 45, 171–180. Vander Zanden, M.J., Vadeboncoeur, Y., 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83, 2152–2161. Wickham, H., 2007. Reshaping data with the {reshape} package. J. Stat. Softw. 21, 1–20. Wickham, H., 2009. ggplot2: Elegant Graphics for Data Analysis. Springer, New York. Yule, D.L., Evrard, L.M., Cachera, S., Colon, M., Guillard, J., 2013. Comparing two fish sampling standards over time: largely congruent results but with caveats. Freshw. Biol. 58, 2074–2088.