Influence of the aquatic vegetation landscape on larval fish abundance

Journal of Great Lakes Research 41 (2015) 873–880

Contents lists available at ScienceDirect

Journal of Great Lakes Research journal homepage: www.elsevier.com/locate/jglr

Influence of the aquatic vegetation landscape on larval fish abundance Philippe Massicotte a,⁎, Andrea Bertolo a, Philippe Brodeur b, Christiane Hudon c, Marc Mingelbier d, Pierre Magnan a a Centre de recherche sur les interactions bassins versants-écosystèmes aquatiques (RIVE). Département des sciences de l'environnement, Université du Quèbec à Trois-Rivières, C. P. 500, Trois-Rivières, Québec, G9A 5H7, Canada b Ministère des Forêts, de la Faune et des Parcs, Direction régionale de la Mauricie et du Centre-du-Québec, Secteur de la faune et des parcs., 100 rue Laviolette, Bureau 207, Trois-Rivières, Québec, G9A 5S9, Canada c Environnement Canada, Division de la recherche sur l'hydrologie et l'écologie des bassins versants, Centre Saint-Laurent, 105 McGill, 7e étage, Montreal, Québec, H2Y 2E7, Canada d Ministère des Forêts, de la Faune et des Parcs, Direction de la faune aquatique, 24 Chemin Sainte-Foy, 2e étage, Québec, Québec G1S 4X4, Canada

a r t i c l e

i n f o

Article history: Received 27 October 2014 Accepted 2 May 2015 Available online 7 July 2015 Communicated by Edward Rutherford Keywords: Aquatic landscape Fish larvae Macrophytes Remote sensing Species-habitat associations Submerged and emergent aquatic vegetation

a b s t r a c t Aquatic vegetation (AV) is an important component of aquatic system, affecting habitat carrying capacity and fish productivity. By modulating habitat complexity, AV has the potential to influence the early ontogenetic stages of many fish species. However, understanding how variations in AV abundance influence fish distribution and abundance at the landscape scale of freshwater ecosystems is a considerable challenge due to the difficulty of collecting the high-resolution data needed for such analyses. We used yellow perch (Perca flavescens) as a biological model to test the hypothesis that larval fish abundance is positively related to a AV-rich landscape. Combining seven years of extensive larval sampling and satellite imagery, our results suggest that fish larvae tend to be associated with AV habitats after hatching. Based on zero-inflated binomial models, the probability of observing perch larvae increased from approximately 20% in low AV habitats to nearly 100% in dense AV habitats (pseudoR2 = 0.54). Our results also show that the probability of observing yellow perch larvae was lower during years of low water level and low AV biomass, likely limiting the availability of proper spawning or nursery habitats for fish. Furthermore, we demonstrate that larvae use both new AV as well as vegetative debris from the previous year that persisted through the winter. In agreement with a growing literature, our study gives additional support to the idea that AV plays a key role in freshwater fish recruitment. © 2015 International Association for Great Lakes Research. Published by Elsevier B.V. All rights reserved.

Introduction Aquatic vegetation (AV) increases habitat complexity in both marine (Boström et al., 2006) and freshwater (Jeppesen and Søndergaard, 1998) ecosystems. In addition to the many effects it has on ecosystem functioning by modulating oxygen availability (Caraco et al., 2006) and nutrient cycling (Kalff and Rooney, 2003; Tall et al., 2011), AV is a key component of physical habitat structure. Characteristics such as individual plant architecture as well as AV patch size and fragmentation structure the aquatic habitats at different spatial scales (Boström et al., 2006). By increasing habitat complexity, AV influences predator–prey interactions (Dionne and Folt, 1991; Diehl, 1992) as well as spatial distribution (Eklöv, 1997; Nurminen and Horppila, 2002), abundance (Boström et al., 2006) and community structure of aquatic organisms (Meerhoff et al., 2007). Even though these effects are relatively well studied at small and meso-scales, the effects of AV landscapes (i.e., at the ecosystem scale) ⁎ Corresponding author at: Department of Bioscience, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmark. E-mail address: [email protected] (P. Massicotte).

on animal ecology have received little attention and have been mainly studied in marine, coastal and estuarine ecosystems (reviewed in Boström et al., 2006). From these studies, it is generally accepted that AV increases the abundance of some animal populations, supports higher species richness and increases the survival of various animal taxa, including fishes (Smokorowski and Pratt, 2007). For example, many freshwater fish species occupy AV habitats at some stage of their life cycle (Petr, 2000). Near-shore habitats dominated by AV provide (i) refuge from predators, (ii) support for invertebrates, and (iii) foraging habitats for small fish (Pratt and Smokorowski, 2003; Yamaki and Yamamuro, 2013). It was also reported that larval and juvenile fish are more abundant in near-shore lacustrine habitats in mid and late summer, not only because of high temperature, but also because of the shelter and food associated to AV (Wang and Eckmann, 1994; Randall et al., 1996). Understanding how variations in AV at the landscape scale influence fish abundance in freshwater ecosystems is a considerable challenge due to the difficulty of collecting the high-resolution data needed for such analyses (Deaton et al., 2010). In these ecosystems, AV is often coarsely characterized using qualitative information (i.e., presence/ absence, density) estimated locally by visual inspection (Randall et al.,

http://dx.doi.org/10.1016/j.jglr.2015.05.010 0380-1330/© 2015 International Association for Great Lakes Research. Published by Elsevier B.V. All rights reserved.

874

P. Massicotte et al. / Journal of Great Lakes Research 41 (2015) 873–880

1996; Bertolo et al., 2012), which does not provide the fine-scale variations in AV biomass that is necessary for characterizing landscapes. Consequently, the influence of AV structure on predator–prey interactions has been mostly studied in laboratory (Christensen and Persson, 1993) or pond (Persson and Eklov, 1995) experiments, whereas studies on populations and communities were typically limited to comparing systems with and without AV (Jeppesen and Søndergaard, 1998). The global loss of AV beds is occurring rapidly in many freshwater ecosystems (Batiuk et al., 2000; Deaton et al., 2010; Vermaire et al., 2012), resulting in a decrease of their carrying capacity (Egertson and Downing, 2004). Negative impacts on fish communities are expected because in their ontogenetic stages, many species use AV for foraging or as shelter to avoid predation or cannibalism (Petr, 2000). A compelling example of the strong linkage between AV habitats and fish carrying capacity is provided in Lake Saint-Pierre (LSP, Hudon et al., 2012), where historically important commercial and sport yellow perch fisheries were supported until the mid-1990s (Guénette et al., 1994). Yellow perch experienced a dramatic collapse beginning in the mid-1990s (Guénette et al., 1994). Despite a major reduction in fishing effort, mainly between 2005 and 2010, yellow perch abundance continued to decrease, likely as a consequence of low recruitment (Magnan et al., 2008). The yellow perch stock has been considered as collapsed since 2011, when a five-year moratorium was imposed on both commercial and sport fishing in 2012. The decrease of aquatic habitat quality was suggested as one of the possible causes of the observed collapse. The progressive transformation of 50 km2 of spring-flooded marshland into perennial crops (1950–1964) and intensive annual cropland between 1964 and 1997 (Richard et al., 2011; de la Chenelière et al., 2014) combined with the shift from AV beds to benthic cyanobacteria documented in 2005–2007 (Vis et al., 2008; Hudon et al., 2012), have profoundly impacted this aquatic ecosystem. Given that many freshwater fishes have economic and recreational values, studies at the landscape level are critically needed to understand

the role of AV on fish abundance and productivity. Here we used yellow perch (Perca flavescens) as a biological model to test the hypothesis that larval fish abundance is positively related to AV abundance at the landscape level. To test this hypothesis, we modeled the probability of observing perch larvae in LSP using data from a seven-year survey of larval abundance (2005–2011) that included more than 200 stations per year in conjunction with an annual index of aquatic vegetation derived from satellite imagery. Materials and methods Study site LSP is the largest fluvial lake (≈ 400 km2 ) of the St. Lawrence River (Fig. 1). This lake, formed by a widening in the river, has a relatively slow current (b0.5 m × s− 1) and is shallow (mean depth = 3 m), except for a man-made navigation channel (depth N 11 m) that concentrates much of the river flow into its central portion (57–88% of total discharge; Hudon and Carignan, 2008). LSP receives the inflow of five major tributaries (cumulative watershed area = 21,127 km2) distinguished on the basis of their physical and biogeochemical characteristics (Frenette et al., 2012). The LSP floodplain itself covers approximately 140 km2, typically for 5–9 weeks during the spring freshet and represents an important spawning ground for yellow perch (Bertolo et al., 2012). LSP is characterized by large sheltered bays and shallow sloping shorelines that support large emergent marshes and extensive beds of submerged aquatic vegetation to a maximum depth of about 3 m outside the navigation channel (Hudon, 1997). During summer, aquatic plants cover 260 km2 (85%) of the lake surface area (Vis et al., 2007) and represent an important habitat for juvenile fish. The submerged vegetation is dominated by American eel-grass (Vallisneria americana Michx) and various pondweeds (Potamogeton richardsonii A. Bennett,

Fig. 1. Study system and sampling sites in Lake Saint-Pierre (LSP). Blue and orange symbols show the locations of the sampling sites located at the 60–80 cm and 100–120 cm isobaths, respectively. Site positions were adjusted each year depending on the water level to make sure they were located at the selected isobaths. For more details on lake bathymetry, see Fig. 1 in Bertolo et al. (2012).


Stuckenia pectinata L.), whereas bulrushes (Schoenoplectus fluviatilis (Torr.) M.T. Strong, Schoenoplectus lacustris (L.) Palla, Schoenoplectus pungens (Vahl) Palla), broad-fruited bur-reed (Sparganium eurycarpum Engelm.) and broad-leaved arrow-leaf (Sagittaria latifolia Willd.) are commonly found in shallow emergent marshes. The amplitude of annual water level variation in LSP ranged between 1.2 and 2.7 m (2000– 2010), reflecting the cumulative influence of the regulated St. Lawrence River, the Ottawa River, and smaller tributaries, with a small (b 0.3 m) semi-diurnal tidal effect. During the study, water levels during the growth season (May–September) varied between 3.53 and 7.11 m (Electronic Supplementary Material (ESM) Table S1, ESM Fig. S1). Daily water level data at Lake Saint-Pierre Curve no. 2 station (gauging station no. 15975, m above the International Great Lakes Datum of 1985) were obtained for the 2004–2012 period. Larval sampling Yellow perch larvae were collected approximately 2–3 weeks after hatching (≈ end of May, see ESM Fig. S1) from 2005 to 2011 following the protocol presented in Bertolo et al. (2012). Each year, sampling sites (n = 216) were located all along the LSP perimeter on two isobaths (60–80 cm and 100–120 cm; Fig. 1). Larvae were sampled using pushnets (2 m long plankton-type net, 0.4 × 0.4 m square mount, 500 μm mesh size) at a velocity of 1 m × s−1. A 50 m long transect parallel to the shore was sampled at each site. Larvae were sampled in the top 40 cm from the surface for the 60–80 cm isobath (two pushnets juxtaposed) and the top 80 cm for the 100–120 cm isobath (two pushnets superposed). Hence, the volumes sampled at the two isobaths were identical. After capture, larvae were narcotized with tricaine methanesulfonate (MS-222™ — Sandoz) and immediately preserved in 75% or 95% ethanol for further laboratory analyses. Larvae were individually identified and measured to the nearest 0.1 mm under a binocular microscope. Abundance data were expressed in number per unit surface area (ind. × m−2) rather than per volume (ind. m−3) to standardize data between isobaths. A total of 101,093 perch larvae were sampled from 2005 to 2011, with body sizes ranging between 5.99 and 27.97 mm (13.72 ± 3.42 mm, mean ± SD).

875

potential radiometric noise using the extract() function from the raster package in R. The detection of AV by satellite in early spring is nearly impossible due to the relatively low AV biomass and high water turbidity of the area. Thus, we calculated EVI using satellite images taken at biomass peak (July/August) as a proxy of available AV at the time of hatching (ESM Table S1). AV data from both the current year and the previous year were used to predict larval fish abundance. We consider that these predictors are complementary because AV is represented by both (i) overwintering senescent stems, related to the distribution of AV during the previous summer, and (ii) new shoots, related to the distribution of AV during the summer of sampling. We evaluated the ability of EVI to predict in situ AV biomass using field measurements collected at 25 sites between 7 and 9 August 2012 on the south shore of LSP (ESM Fig. S2, panel I). EVI values were calculated using a Landsat 7 image taken on 8 July 2012 and regressed against in situ AV biomass. The image used to examine the relationship between EVI and AV was taken one month before AV sampling because it was the best available image (i.e. clear atmosphere). In LSP, AV growth is typically initiated in late May and biomass is maximal between mid-July and mid-September; such variability in the timing of peak biomass likely results from differences in species composition among sites and environmental conditions among years (Hudon, personal observations). We expected that the large buffer size (90 × 90 m) of remote sensing images would attenuate the effects of small-scale spatial and seasonal variability on our assessment of AV biomass. At each site, submerged vascular macrophytes and metaphyton (filamentous cyanobacteria and chlorophytes) were collected using a double-headed rake (35 cm wide) dragged on the bottom over a length of approximately 1 m (Yin et al., 2000). At each site, replicate rake samples were collected around the boat, which were used to estimate mean biomass per site. The sampled vegetation was thoroughly rinsed, sorted among major plant types (major taxa of vascular plants, filamentous cyanobacteria, and chlorophytes) and weighed after drying (50 °C). Vegetation dry mass at each site (mean g DM m− 2 ± S.D.) was estimated for each plant type by averaging values of 3–10 individual replicate rake samples. Modeling spatial patterns of perch larvae

Remote sensing of aquatic vegetation From 2005 to 2011, AV biomass was estimated using the enhanced vegetation index (EVI) derived from Landsat 5 and 7 imageries (row 14, path 28, 7911 × 7211 pixels at a resolution of 30 × 30 m; http:// earthexplorer.usgs.gov). EVI was calculated as follows (Nagler et al., 2005): EVI ¼ 2:5

NIR−RED NIR þ 6RED−7:5BLUE þ 1

ð1Þ

where NIR, RED, and BLUE represent the Landsat near infrared (band 4), red (band 3), and blue (band 1) spectral bands, respectively. For each year, the best image candidate (i.e., when the atmosphere was clear) was selected between July and August (ESM Table S1, ESM Fig. S1). Prior to analysis, radiometric calibrations were applied to convert 8bit satellite-quantized digital numbers to at-satellite top-of-atmosphere reflectance 32-bit values (Canty et al., 2004; Schroeder et al., 2006; Chander et al., 2009). All images (Landsat 5 and Landsat 7) were corrected for sensor degradation using the most current radiometric calibration coefficients (Chander et al., 2009). We also applied a pseudo-invariant procedure to attenuate atmospheric effects in order to standardize any atmosphere-related changes in top-of-atmosphere reflectance for the image set. The detailed procedure and references can be found in Massicotte et al. (2013). Clouds and cloud shadows were removed from each image in Matlab using the Fmask tool (Zhu and Woodcock, 2012). Mean EVI values were extracted at each sampling location using a buffer of 90 × 90 m (3 × 3 pixels) to remove

In LSP, yellow perch nurseries are mainly located in a few bays that are regularly distributed along the shore at intervals of ≈ 10 km (Bertolo et al., 2012). Larvae show large-scale aggregation patterns likely related to the location of the main spawning grounds (Bertolo et al., 2012, see Fig. 1), so we used a spatial autocorrelation term to take into account such spatial dependency (Table 1). The explanatory spatial variables (eigenfunctions) used to model the distribution of larval perch abundances at broad scales (Table 1) were generated by asymmetric eigenvector map (AEM) analysis (Blanchet et al., 2008). In brief, the AEM analysis allows modeling of larval spatial distributions generated by asymmetric and directional physical processes such as water flow. The analysis generates explanatory spatial variables (eigenfunctions) based on a directional graph between sampling sites (ESM Fig. S3). The AEM directional graphs between sampling sites were generated so that each site is connected with the six nearest neighbors to ensure that there were no unlinked groups of stations (ESM Fig. S3). One graph per year and per shore was generated. We treated the shores separately because the navigation channel (see Fig. 1) is a high-velocity (N0.5 m × s−1) zone that prevents lateral connectivity between north and south shore for organisms with limited swimming capacity, such as yellow perch larvae (Fortin and Magnin, 1972; Houde, 1969). Furthermore, genetic differences between north and south shore yellow perch larvae have been reported (Leung et al., 2011). Moran's I coefficients, which measure the degree of correlation among neighboring observations in the network, were calculated for each eigenfunction produced by the AEM procedure. For each combination of year/shore, the first four spatially structured eigenvectors (AEM) with a positive

876


Table 1 Description of the model candidates used in this study (number of observations = 1505). Variables from the experimental design (isobaths, year) were also included as explanatory variables in the models. In all models, the part to the left of the | represents the count model (GLM using negative binomial distribution) whereas the right side represents the binomial model. The enhanced vegetation index (EVI) was derived from Landsat 5 and 7 imageries to estimate macrophyte biomass. Asymmetric eigenvector map (AEM) variables were used to model the spatial distribution of larval fish abundances over a broad scale. AIC is the Akaike information criterion. The best model is presented in bold. Model candidate

Model specification

Description

AIC

ΔAIC

M1 M2 M3 M4

Abundance ~ AEM|1 Abundance ~ AEM + EVIyear − 1|EVIyear − 1 Abundance ~ AEM + EVIyear|EVIyear Abundance ~ AEM + EVIyear + EVIyear − 1|EVIyear

Includes only spatial variables generated by the AEM analysis. Same as M1 but with EVI calculated for the previous year (EVIyear − 1). Same as M1 but with EVI calculated for the current year (EVIyear). Same as M1 but with EVI calculated for the previous and current year.

112,384.81 95,095.50 89,841.38 89,321.20

23,063.61 5774.30 520.18 0.00

Moran's I index between 0.70 and 1 were kept for subsequent analyses. This choice was motivated because these AEM represent spatial patterns occurring at roughly the same scale to which nurseries are distributed around the LSP (Bertolo et al., 2012). Modeling framework Larval abundances were modeled using zero-inflated models to account for the excessive number of zero abundances observed in the data (Zuur et al., 2009). This type of model has two parts: (1) a negative binomial or Poisson generalized linear model (GLM) for the count process and (2) a binomial model for predicting the excess zeros (zeroinflation model). In this study, we used a negative binomial distribution for the count process because we found that the variances of larval abundance were larger than the mean. Four candidate models were developed to take into account different a priori hypotheses about the effects of AV on larval distribution. All models integrated all the data available (seven years, n = 1505). The base model (M1) includes isobath and year, since these variables were part of the experimental design, as well as spatial variables representing aggregation patterns at broad scales to account for large-scale effects likely related to the location of spawning grounds (see below for details). Because larvae in the spring tend to be associated with AV that overwinter or emerge early from the substrate, presumably in search for food and/or cover (Čech et al., 2009; Paradis et al., 2014), we developed a second model (M2) that is similar to M1 but that also includes AV biomass calculated for the previous year (EVIyear − 1). The third model (M3) is identical to M1 but includes EVI calculated the year of sampling (EVIyear). Finally, the fourth model (M4) includes EVI calculated both for the previous and the current year. Including these two proxies of AV abundances (one estimated the year before representing early spring cover opportunities and a second at the end of the current summer representing cover available in summer and early fall) in the same model should give a better representation of relative AV abundance at the time of larvae sampling (i.e., late spring at the end of May). Candidate models M1, M2, M3, and M4 were fitted using the zeroinfl() function from the pscl package (Zeileis et al., 2008). As suggested by Zuur et al. (2009), the best zero-inflated candidate model was selected using the Akaike information criterion (AIC), which is based on the principle of parsimony and identifies the model that accounts for the most variation with the fewest variables (Burnham and Anderson, 2002). We thus used the AIC criterion to select the best models and indirectly infer about the relative importance of variables. The pseudo-R2 values of each model were extracted using the pR2() function from the pscl package (Zeileis et al., 2008). All statistical analyses were performed in R 3.0.1 (R Development Core Team 2013).

function from the DAAG R package showed a good predictive accuracy of the model and consequently its reliability to predict EVI from 2004 and 2011 (unpublished data). EVI calculated from 2004 to 2011 ranged between − 0.025 and 0.101; highest values were observed in the nearshore areas and bays of LSP, where emergent vegetation is abundant (ESM Fig. S2). The lowest EVI values were found toward the navigation channel in the center of LSP, where water depth and velocity are generally too high to allow AV growth.

Model selection AIC revealed that all model candidates that included AV biomass (M2–M4) performed better than the basic model (M1) for predicting larval perch abundance (Table 1). The inclusion of either the AV biomass of the previous or current year clearly improved the plausibility of models M2 and M3, respectively, compared to model M1. Furthermore, the model that included both the previous and current year estimates of AV biomass (M4) had the lowest AIC, indicating it was the best of the four models (Table 1). The regression model obtained by the binomial GLM of the retained zero-inflated model (M4) showed an increasing probability of observing perch larvae as EVI increased (pseudo-R2 = 0.54, p b 0.0001; Fig. 3). For the lower values of EVI (≈−0.02), the probability of observing perch larvae was approximately 20%; it increased to nearly 100% for the highest values of EVI (≈0.1). The probabilities of observing fish larvae for individual years (colored lines Fig. 3) were consistent with those obtained by pooling all years in the same analysis (black thick line Fig. 3). Our results also indicated that the probability of observing yellow perch larvae was lower during years when water level was low during sampling (2007 and 2010; ESM Table S1, ESM Fig. S1). For comparable water level conditions, the probability of observing yellow perch larvae was lower during years of low AV biomass (2008 vs 2009; ESM Figs. S2E vs S21F and Fig. 3).

Results Prediction of AV biomass In 2012, EVI was found to be a relatively good predictor of in situ AV biomass in LSP (Fig. 2, R2 = 0.44, p = 0.00033, n = 25). A crossvalidation procedure (number of folds = 5) using the CVlm()

Fig. 2. Relationship between the enhanced vegetation index (EVI) and macrophyte biomass. Aquatic vegetation for biomass measurements was collected between 7 and 9 August 2012 on the south shore of Lake Saint-Pierre (see Fig. 1). EVI values were calculated from a Landsat 7 image taken on 8 July 2012.


1.0

877

Characterizing the AV landscape in freshwater ecosystems

Influence of AV landscape on larval fish abundance

Studies aiming to evaluate the relationship between AV and fish abundance in freshwaters have commonly compared systems with and without AV. Such an approach is limited by the fact that field estimates of AV cover and biomass are time-consuming and reflect important small-scale variability (Vis et al., 2007). Traditional in situ sampling techniques such as rakes or SCUBA are limited in their ability to fully describe heterogeneity at the landscape level because they are time consuming and adapted to small-scale studies. Because detailed characterization of the AV landscape represents a considerable

Many factors can influence fish abundance and community composition in aquatic ecosystems (Jackson et al., 2001). Among these, it was recently suggested that AV was consistently a better predictor of fish community composition than water quality variables (Cvetkovic et al., 2010). For instance, empirical studies have found that fish larvae in freshwater AV beds could be as much as 160 times more abundant than in adjacent open waters, and that larvae concentrate inside aquatic plant beds rather than in the transition zones between open and vegetated habitats (Paller, 1987; Deaton et al., 2010). Our results are

Probability of observing fish larvae

Combining seven years of extensive larval sampling (n = 1505) and satellite imageries, and after controlling for spatial auto-correlation among nursery areas at a broad scale, our results clearly suggest that yellow perch larvae tend to be associated with AV habitats within a few weeks of hatching. Whereas the position of spawning grounds is known only at a rough/large scale (Bertolo et al., 2012, see Fig. 1), it is clear that in LSP yellow perch spawns at the beginning of spring in temporary flooded areas were most often exposed to air at the time of our sampling. Giving that spawning occur in a slow flowing 30–100 cm water column in densely vegetated areas that rapidly warms in the spring (Mingelbier and Morin, 2005), our study thus lends additional support to the idea that the AV cover is associated positively with larval fish abundance at the landscape scale. Fish–habitat relationships are challenging to quantify because little is known about the functional links between fish and habitats, suggesting that life-history approaches are needed (Petr, 2000; Randall et al., 2012). To our knowledge, the results we present here provide the first evidence of a positive relationship between larval fish abundance and AV at the landscape scale; perhaps an inferred preference or evolutionary benefit. Our results thus expand the findings by Bertolo et al. (2012), that found similar relationships but measured only a coarse index of AV abundance over a shorter-time period (3 years). The comparison with the results by Bertolo et al. (2012) demonstrates the robustness of the remote sensing approach and suggests that habitat modeling based on this technique is a promising tool to predict larval habitat quality when field sampling is not possible.

challenge (Deaton et al., 2010), studies aiming to elucidate the effects of AV on fish abundance usually consider either the spatial or temporal scale, but rarely both (Boström et al., 2006). Here we show how remote sensing can be used to overcome these shortcomings by allowing large areas to be examined with spatial and temporal resolutions fine enough to detect local interannual changes in vegetation cover at the landscape level (Nelson et al., 2006; Albright and Ode, 2010; Visser et al., 2013). Water level conditions in LSP at the time of Landsat image acquisition ranged between 3.78 m (2007) and 5.01 m (2008), resulting in important differences in water depth, turbidity, light intensity reaching the bottom and most likely, AV biomass among years. For example, bottom areas lying at a depth of 1 m on the 2007 Landsat image would be under 2.28 m of water in 2008 whereas the bottom area under 1 m of water on the image from 2008 would have been completely dry the previous year (ESM Fig. S2). Because Landsat images used for EVI calibration (2012) and AV habitat characterization (2004–2011) represented conditions from a single day for each year, we investigated other conditions within the lake that could potentially affect EVI scores independent of AV coverage. In this respect, the effects of water color and CDOM on optical water properties have been well documented in the past (Frenette et al., 2012; Massicotte et al., 2013). In LSP, variations in the concentration of CDOM likely result from spatial proximity and seasonal variations in the inflow of tributary waters, which indeed fluctuate daily, seasonally and among years. The Saint-François River, whose waters flow along the nearshore southern LSP, exhibited a mean monthly discharge ranging from 39 m3 s− 1 in August 2012 (the year of AV field sampling and Landsat calibration of EVI estimation) to 426 m3 × s−1 in August 2008. Water color and CDOM concentrations in the SaintFrançois River also likely varied in time as a positive function of discharge, as previously documented for DOC in several Quebec rivers (Hudon et al., 1996). Nevertheless, in spite of the large inter-annual variations in water level and in optical water properties in LSP, the relationship between EVI and larval probability of occurrence was highly significant, which strongly supports our interpretation of the importance of AV as a critical habitat for fish larvae. It should be noted that EVI scores derived from the 2012 Landsat image and field data used in the calibration exercise cover a narrower range (− 0.15 to 0.05, Fig. 2) than the EVI scores derived from the Landsat images from 2004–2011 (−0.02 to 0.10, ESM Fig. S2). This difference is partly attributable to the low water level conditions prevailing in the summer of 2012 (3.6–3.8 m) in comparison with the previous years (3.8–5.2 m) and also to interannual differences in the optical quality of water masses within LSP. In LSP, upwelling radiometric contributions from the bottom is unlikely to influence reflectance and EVI values since the sampling stations were located in highly turbid areas (thus not influencing the longer wavelengths that are used to calculate EVI). Remote sensing of AV in shallow water aquatic ecosystems is still under rapid development, especially in clear water environments where reflectance values can be affected by the bed of the water body. A new generation of multispectral sensors, such as Landsat-8 satellite (OLI), are now being deployed which will provide spectral and radiometric resolutions much higher than their predecessors. These improvements should allow the development of algorithms for better assessments of AV biomass in shallow water ecosystems such as the LSP.

0.8

0.6

0.4 2005 2006 2007 2008 2009 2010 2011

0.2

0.0 −0.02

0.00 0.02 0.04 0.06 0.08 Enhanced vegetation index (EVI)

0.10

Fig. 3. Probability of observing perch larvae as a function of the enhanced vegetation index (EVI). The thick black line represents the fitted GLM binomial on pooled data (2005–2011) whereas the colored lines are obtained using year as a covariable (pseudo-R2 = 0.56, pEVI b 0.0001, p2006 = 0.0277, p2007 b 0.0001, p2008 = 0.0739, p2009 = 0.6632, p2010 b 0.0001, p2011 = 0.0055). Dots are the observed values (presence/absence).

Discussion

878


consistent with these findings and show that the global probability of observing yellow perch larvae increases from approximately 20% with low AV abundance to nearly 100% in dense AV habitats (Fig. 3). We also showed that the probability of observing yellow perch larvae was lower during years of relatively high AV biomass but of low spring water levels. This result suggests that low water levels could limit fish access to essential spawning and nursery habitats composed of dense vegetation (Mingelbier and Morin, 2005) and/or could enhance egg or larva mortality just after reproduction. Low water levels could render egg strands more vulnerable to mechanical shock induced by high winds (Dumont, 1996), reduce access of larvae to important foraging AV habitats thus limiting growth and survival (Glemet and Rodriguez, 2007), or enhance larval predation and cannibalism (Sass et al., 2006). Episodes of increasingly low water levels in close succession (2007, 2010 and 2012, see ESM Fig. S1) resulted in a sequence of alternately drying out and flooding of shallow riparian areas over successive years, resulting in a fringe of bare bottom where submerged organisms previously thrived. Such change in shallow riparian habitat may have contributed to limit yellow perch recruitment in LSP and further points out the importance of predictable seasonal water level fluctuations to maintain both shallow-water AV habitats and floodplain accessibility for fish productivity. Our results thus suggest that episodes of low water levels (2007, 2010 and 2012) in conjunction with the hypothesized loss of large AV beds observed in recent years could have contributed to limit yellow perch recruitment in LSP. It also highlights the importance of AV habitats and their accessibility (with high water levels) for fish productivity. The spatial distribution and biomass of AV beds varied greatly among years in LSP, even under similar water level conditions (ESM Table S1, ESM Figs. S2E vs S2F). These relatively large interannual variations have also been observed in other large river ecosystems, like the Amazon (Silva et al., 2013). In LSP, the distribution, composition and productivity of AV beds are largely influenced by the inflow of nutrients from watershed land use (Hudon and Carignan, 2008), primarily agriculture, variation in transparency (Hudon et al., 2000), water retention time, geomorphology (i.e. depth, bottom type, exposure, slope, Vis et al., 2007) and interannual water level variations (Hudon, 1997; Vis et al., 2007; Hudon and Carignan, 2008; Hudon et al., 2012). Littoral habitats and AV have also been affected by cumulative human interventions over the past 150 years, including progressive channel excavation (1854–1998), control of ice-jams using icebrakers since 1976 to maintain the navigation channel free of ice, water discharge and level regulation (since 1960), rising water temperatures (since the early 1970s) and the development of intensive agriculture in the floodplain (since the early 1990s, Hudon, 2004; Richard et al., 2011). Over the past 10 years, a reduction in the abundance of submerged aquatic vegetation (de la Chenelière et al., 2014) and the proliferation of benthic cyanobacteria (Vis et al., 2007) have been observed in LSP. These natural and anthropogenic changes influencing AV cover indirectly affect larval fish dynamics by modulating habitat quality and quantity as well as the position of their niche in these littoral habitats (Randall et al., 2012). For comparable water level conditions, the lower probability of observing yellow perch larvae during years of low AV biomass (2008 vs 2009; ESM Fig. S2) shows that the loss of vegetation beds at the landscape scale can have a great impact on yellow perch recruitment. Further evidence of the importance of AV on yellow perch was provided by Hudon et al. (2012) who documented how a shift from submerged AV to benthic cyanobacteria coincided with a reduction in benthic food resources, low juvenile perch growth and adult fish biomass. Our study confirms the findings by Bertolo et al. (2012) and Paradis et al. (2014), which showed that yellow perch larvae, usually considered pelagic until they reach the juvenile stage (Urho, 1996), can show a strong association with AV. The extensive shallow areas and the lack of a truly pelagic zone in LSP probably explain this exception. Spawning yellow perch aggregate in the main bays of LSP which are densely vegetated and offer spawning grounds in shallow flooded

areas (Mingelbier and Morin, 2005). The abundance of vegetation in the spawning grounds of flooded bays could thus contribute to the strong association observed between perch larvae and AV shortly after hatching. Hence, the functional relationship between AV and fishes are likely to depend on the species and life stage (Randall et al., 2012). Taking these traits into account might improve our predictions of fish recruitment. AV landscapes are highly dynamic. Their distribution is partially determined by factors such as depth and light penetration and also by the distribution of AV beds from previous years. It has been argued that during the spring, fish may associate with both new vegetation and overwintered aquatic plants for food and cover (Killgore et al., 1991). In agreement with these results, using AV biomass estimated for the previous (EVIyear − 1, M2) or the current (EVIyear, M3) year provided better support compared to the basic model (M1), which did not include AV biomass, while the best model (M4) included both EVIyear − 1 and EVIyear. Thus, our results might also suggest that larvae use both new vegetation and AV debris from the previous year that persisted through the winter (Table 1). Including both current and past AV summer cover in our analysis not only gave us the possibility to have a proxy of AV cover in spring, but also provided interesting insights into the ecology of the AV-larval fish association. Although either passive or active habitat selection is possible, it was suggested that yellow perch larvae actively select AV habitats after hatching, at least in systems lacking a pelagic zone (Bertolo et al., 2012; Paradis et al., 2014). This effect might be greater in aquatic ecosystems with extensive littoral zones (Weaver et al., 1997) such as LSP, where water currents are moderate (except for the navigation channel; Fig. 1). In such areas, the zooplankton consumption rate of juvenile perch is generally higher in habitats with dense AV (Winfield, 1986), again emphasizing the use of AV habitats for foraging. Even though we cannot pinpoint a particular mechanism for active habitat selection, by explicitly modeling spatial connectivity at a broad scale (as a proxy for dispersion potential; ESM Fig. S3), our results provide evidence that AV biomass plays a key role in determining the distribution and abundance of yellow perch larvae in the system. In addition to total biomass, AV community composition can also affect fish distribution (Chick and Mclvor, 1994; Weaver et al., 1997; Cvetkovic et al., 2010; Silva et al., 2013). Given that remote sensing and aerial imagery are becoming more available, the use of image texture metrics as a proxy of AV community composition may also provide additional and valuable information to evaluate the effect of AV landscape on fish communities. Our results indicate that remote sensing of AV biomass can be a useful tool to monitor the evolution of AV landscapes in shallow systems over longer periods of time. Based on the technique developed in our study, further work should explicitly look at the historical distribution and abundance of AV beds in LSP, in order to test the hypothesis of an impact of recent AV loss on yellow perch recruitment. The negative influence of water level and of low AV biomass on larval yellow perch abundance that we show in our study suggests potential impacts of future climate change on fish recruitment. Under climate change scenarios, warmer water temperatures and lower water levels in the St. Lawrence River are anticipated, with important consequences on riparian habitats Hudon (2004). Acknowledgments This work was supported by the Centre de recherche sur les interactions bassins versants-écosystèmes aquatiques (RIVE) (1480000) and the Canada Research Chair in Freshwater Ecology (both from the Université du Québec à Trois-Rivières, Québec, Canada) (950-205922) and the Ministère des Forêts de la Faune et des Parcs du Québec. We thank Mr. Rémi Bacon and Nicolas Auclair for larval identification and the staff and research assistants from the Ministère des Forêts de la Faune et des Parcs du Québec and the Université du Québec à TroisRivières for field sampling and laboratory work. Additional support


was given by the Groupe de recherche interuniversitaire en limnologie et environnement aquatique (GRIL). P. Massicotte was supported by a postdoctoral fellowship from RIVE. Contribution of RIVE and the Groupe de recherche interuniversitaire en limnologie et environnement aquatique (GRIL).

Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jglr.2015.05.010.

References Albright, T.P., Ode, D.J., 2010. Monitoring the dynamics of an invasive emergent macrophyte community using operational remote sensing data. Hydrobiologia 661 (1), 469–474. http://dx.doi.org/10.1007/s10750-010-0537-8 (Nov.). Batiuk, R., Bergstrom, P., Kemp, W.M., Koch, E., Murray, L., Stevenson, J.C., Bartleson, R., Carter, V., Rybicki, N., Landwehr, J., Gallegos, C., Karrh, L., Naylor, M., Wilcox, D., Moore, K.A., Ailstock, S., Teichberg, M., 2000. Chesapeake Bay submerged aquatic vegetation water quality and habitat-based requirement and restoration targets: a second technical synthesis. Tech. Rep. August, EPA, Chesapeake Bay Program, Annapolis. Bertolo, A., Blanchet, G., Magnan, P., Brodeur, P., Mingelbier, M., Legendre, P., 2012. Inferring processes from spatial patterns: the role of directional and non-directional forces in shaping fish larvae distribution in a freshwater lake system. PLoS One 7 (11), e50239 (Jan., URL http://www.pubmedcentral.nih.gov/articlerender.fcgi?art id = 3502328&tool = pmcentrez&rendertype = abstract). Blanchet, F.G., Legendre, P., Borcard, D., 2008. Modelling directional spatial processes in ecological data. Ecol. Model. 215 (4), 325–336 (URL b Go to ISIN://000257979200007). Boström, C., Jackson, E.L., Simenstad, C.a., 2006. Seagrass landscapes and their effects on associated fauna: a review. Estuar. Coast. Shelf Sci. 68 (3-4), 383–403 (Jul., URL http://linkinghub.elsevier.com/retrieve/pii/S0272771406 001004). Burnham, K.P., Anderson, D.R., 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd edition. Springer, New York. Canty, M., Nielsen, A., Schmidt, M., 2004. Automatic radiometric normalization of multitemporal satellite imagery. Remote Sens. Environ. 91 (3-4), 441–451 (Jun., URL http://linkinghub.elsevier.com/retrieve/pii/S0034425704 001208 http://www. sciencedirect.com/science/article/pii/S0034425704001208). Caraco, N., Cole, J., Findlay, S., Wigand, C., 2006. Vascular plants as engineers of oxygen in aquatic systems. Bioscience 56 (3), 219–225 (URL http://dx.doi.org/10.1641/00063568(2006)056[0219:VPAEO O]2.0.CO;2 http://bioscience.oxfordjournals.org/ content/56/3/219.short). Čech, M., Peterka, J., Říha, M., Jůza, T., Kubečka, J., 2009. Distribution of egg strands of perch (Perca fluviatilis L.) with respect to depth and spawning substrate. Hydrobiologia 630 (1), 105–114 (May, URL http://link.springer.com/10.1007/ s10750-009-9783-z). Chander, G., Markham, B.L., Helder, D.L., 2009. Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM +, and EO-1 ALI sensors. Remote Sens. Environ. 113, 893–903 (URL b Go to ISIN://000265716000002). Chick, J., Mclvor, C., 1994. Patterns in the abundance and composition of fishes among beds of different macrophytes: viewing a littoral zone as a landscape. Can. J. Fish. Aquat. Sci. 51, 2873–2882 (URL http://www.nrcresearchpress.com/doi/pdf/10.1139/ f94-286). Christensen, B., Persson, L., 1993. Species-specific antipredatory behaviors — effects on prey choice in different habitats. Behav. Ecol. Sociobiol. 32, 1–9. Cvetkovic, M., Wei, A., Chow-Fraser, P., 2010. Relative importance of macrophyte community versus water quality variables for predicting fish assemblages in coastal wetlands of the Laurentian Great Lakes. J. Great Lakes Res. 36 (1), 64–73 (Mar., URL http:// www.sciencedirect.com/science/article/pii/S03801 33009002135). de la Chenelière, V., Brodeur, P., Mingelbier, M., 2014. Restauration des habitats du lac Saint-Pierre : un prérequis au rétablissement de la perchaude. Le Nat. Can. 138 (2), 50 (URL http://id.erudit.org/iderudit/1025070ar). Deaton, A.S., Chappell, W.S., Hart, K., O'Neal, J., Boutin, B., 2010. North Carolina Coastal Habitat Protection Plan. Tech. Rep. , p. 659. Diehl, S., 1992. Fish predation and benthic community structure: the role of omnivory and habitat complexity. Ecology 73 (5), 1646–1661 (URL http://www.jstor.org/stable/10. 2307/1940017). Dionne, M., Folt, C.L., 1991. An experimental analysis of macrophyte growth forms as fish foraging habitat. Can. J. Fish. Aquat. Sci. 48 (1), 123–131 (Jan., URL http://www. nrcresearchpress.com/doi/abs/10.1139/f91-017 #.U62wQP7VOF4). Dumont, P., 1996. Comparaison de la dynamique des populations de perchaudes (Perca flavescens) soumises à des niveaux différents de stress anthropique (Ph.D. thesis). Université du Québec à Montréal. Egertson, C., Downing, J., 2004. Relationship of fish catch and composition to water quality in a suite of agriculturally eutrophic lakes. Can. J. Fish. Aquat. Sci. 61, 1784–1796 (URL http://www.nrcresearchpress.com/doi/abs/10.1139/f04-109). Eklöv, P., 1997. Effects of habitat complexity and prey abundance on the spatial and temporal distributions of perch (Perca fluviatilis) and pike (Esox lucius). Can. J. Fish. Aquat. Sci. 54, 1520–1531 (Apr., URL http://www.nrcresearchpress.com/doi/abs/10.1139/ f97-059 #.U62xFv7VOF4).

879

Fortin, R., Magnin, E., 1972. Dynamique d'un groupement de perchaudes, Perca flavescens (Mitchill) dans la Grande Anse de l'Île Perrot, au lac Saint-Louis. Nat. Can. 99, 367–380. Frenette, J.-J., Massicotte, P., Lapierre, J.-f., 2012. Colorful niches of phytoplankton shaped by the spatial connectivity in a large river ecosystem: a riverscape perspective. PLoS One 7 (4), e35891 (Jan., URL http://www.ncbi.nlm.nih.gov/pubmed/22558259 http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid = 3340396&tool = pmce ntrez&rendertype = abstract). Glemet, H., Rodriguez, M.A., 2007. Short-term growth (RNA/DNA ratio) of yellow perch (Perca flavescens) in relation to environmental influences and spatio-temporal variation in a shallow fluvial lake. Can. J. Fish. Aquat. Sci. 64 (12), 1646–1655 (Dec., URL http://apps.webofknowledge.com/full_record.do?product = WOS&search_mode = GeneralSearch&qid = 1&SID = 2EszzoFrBECxqcCaEsG&page = 1&doc = 1). Guénette, S., Mailhot, Y., Mcquinn, I., Lamoureux, P., Fortin, T., 1994. Paramètres biologiques, exploitation commerciale et modélisation de la population de perchaude (Perca flavescens) du lac Saint-Pierre. Tech. rep. Ministère de l'environnement et de la faune, Direction régionale Mauricie-Bois-Francs et Université du Québec à Montréal. Houde, E.D., 1969. Sustained swimming ability of larvae of walleye (Stizostedion vitreum vitreum) and yellow perch (Perca flavescens). J. Fish. Board Can. 26 (6), 1647–1659 (URL b Go to ISIN://A1969D667500018). Hudon, C., 1997. Impact of water level fluctuations on St. Lawrence River aquatic vegetation. Can. J. Fish. Aquat. Sci. 54, 2853–2865 (URL http://www.nrcresearchpress.com/ doi/abs/10.1139/f97-201). Hudon, C., 2004. Managing St. Lawrence River discharge in times of climatic uncertainty: how water quantity affects wildlife, recreation and the economy. Transactions of the 69th North American Wildlife and Natural Resources Conference. Vol. 69. wildlife management inst, 1146 19th st, nw, ste 700, Washington, DC 20036 USA, Spokane, WA, pp. 165–181 (URL http://apps.webofknowledge.com/full_ record.do?product = WOS&search_mode = GeneralSearch&qid = 6&SID = 2EszzoFrBECxqcCaEsG&page = 1&doc = 1). Hudon, C., Carignan, R., 2008. Cumulative impacts of hydrology and human activities on water quality in the St. Lawrence River (Lake Saint-Pierre, Quebec, Canada). Can. J. Fish. Aquat. Sci. 65 (6), 1165–1180 (Jun., URL http://www.nrcresearchpress.com/ doi/abs/10.1139/F08-069). Hudon, C., Morin, R., Bunch, J., Harland, R., 1996. Carbon and nutrient output from the Great Whale River (Hudson Bay) and a comparison with other rivers around Quebec. Hudon, C., Lalonde, S., Gagnon, P., 2000. Ranking the effects of site exposure, plant growth form, water depth, and transparency on aquatic plant biomass. Can. J. Fish. Aquat. Sci. 57 (S1), 31–42 (URL http://www.nrc.ca/cgi-bin/cisti/journals/rp/rp2_abst_ e?cjfas_ f99-232_57_ns_nf_cjfas57-00). Hudon, C., Cattaneo, A., Tourville Poirier, A.-M., Brodeur, P., Dumont, P., Mailhot, Y., Amyot, J.-P., Despatie, S.-P., Lafontaine, Y., 2012. Oligotrophication from wetland epuration alters the riverine trophic network and carrying capacity for fish. Aquat. Sci. 74 (3), 495–511 (Dec., URL http://www.springerlink.com/index/10. 1007/s00027-011-02 43-2). Jackson, D.A., Peres-Neto, P.R., Olden, J.D., 2001. What controls who is where in freshwater fish communities: the roles of biotic, abiotic, and spatial factors. Can. J. Fish. Aquat. Sci. 58 (1), 157–170 (Jan., URL http://www.nrcresearchpress.com/doi/abs/10.1139/ f00-239 #.U6290v7VOF4). Jeppesen, E., Søndergaard, M., 1998. The structuring role of submerged macrophytes in lakes, Springer Edition. Ecological Studies Vol. 131. Springer New York, New York, NY (URL http://forskningsbasen.deff.dk/Share.external?sp = S9e465 3f0-7e9c11dd-a5a8-000ea68e967b&sp = Sau http://www.springerlink.com/index/10.1007/ 978-1-4612-0695-8). Kalff, J., Rooney, N., 2003. Submerged macrophyte-bed effects on water-column phosphorus, chlorophyll a, and bacterial production. Ecosystems 6 (8), 797–807 (Dec., URL http://link.springer.com/10.1007/s10021-003-0184-2). Killgore, K.J., Hoover, J.J., Morgan, Raymond, P.I., 1991. Aquatic plant control research program. Habitat value of aquatic plants for fishes. Tech. rep., US Army Corps of Engineers, Vicksburg (Jun., URL http://oai.dtic.mil/oai/oai?verb = getRecord&metadataPre fix = html&identifier = ADA239569). Leung, C., Magnan, P., Angers, B., 2011. Generic evidence for sympatric populations of yellow perch (Perca flavescens) in Lake Saint-Pierre (Canada): the crucial first step in developing a fishery management. J. Aquaculture Res. Dev. S6 (001), 1–7 (URL http://www.omicsonline.org/2155-9546/pdfdownload.php?do wnload = 21559546-S6-001.pdf&&aid = 3669). Magnan, P., Mailhot, Y., Dumont, P., 2008. État du stock de perchaude du lac SaintPierre en 2007 et efficacité du plan de gestion de 2005. Tech. rep. Comité aviseur scientifique sur la gestion de la perchaude du lac Saint-Pierre, Université du Québec à Trois-Rivières et ministère des Ressources naturelles et de la Faune du Québec. Massicotte, P., Gratton, D., Frenette, J.-J., Assani, A.A., 2013. Spatial and temporal evolution of the St. Lawrence River spectral profile: a 25-year case study using landsat 5 and 7 imagery. Remote Sens. Environ. 136, 433–441 (Sep., URL http://linkinghub.elsevier.com/retrieve/pii/S0034425713 00182X http://www. sciencedirect.com/science/article/pii/S003442571300182X). Meerhoff, M., Iglesias, C., De Mello, F.T., Clemente, J.M., Jensen, E., Lauridsen, T.L., Jeppesen, E., 2007. Effects of habitat complexity on community structure and predator avoidance behaviour of littoral zooplankton in temperate versus subtropical shallow lakes. Freshw. Biol. 52 (6), 1009–1021. Mingelbier, M., Morin, J., 2005. Modélisation numérique 2D de l'habitat des poissons du Saint-Laurent fluvial pour évaluer l'impact des changements climatiques et de la régularisation. Le Nat. Can. 129 (1), 96–102. Nagler, P., Scott, R., Westenburg, C., Cleverly, J., Glenn, E., Huete, a., 2005. Evapotranspiration on western U.S. rivers estimated using the Enhanced Vegetation Index from

880


MODIS and data from eddy covariance and Bowen ratio flux towers. Remote Sens. Environ. 97 (3), 337–351 (Aug., URL http://linkinghub.elsevier.com/retrieve/pii/ S0034425705 001616). Nelson, A.S., Cheruvelil, K.K.S.K., Soranno, A.P., Nelson, a.C., S. S., Soranno, P.a., 2006. Satellite remote sensing of freshwater macrophytes and the influence of water clarity. Aquat. Bot. 85 (4), 289–298 (Nov., URL http://linkinghub.elsevier.com/ retrieve/pii/S0304377006 000921 http://www.sciencedirect.com/science/ article/pii/S0304377006000921). Nurminen, L.K.L., Horppila, J.A., 2002. A diurnal study on the distribution of filter feeding zooplankton: effect of emergent macrophytes, pH and lake trophy. Aquat. Sci. 64 (2), 198–206. Paller, M.H., 1987. Distribution of larval fish between macrophyte beds and open channels in a southeastern floodplain swamp. J. Freshw. Ecol. 4 (2), 191–200 (Dec., URL http:// dx.doi.org/10.1080/02705060.1987.9664652 http://www.tandfonline.com/doi/abs/ 10.1080/02705060.1987.9664652). Paradis, Y., Bertolo, A., Mingelbier, M., Brodeur, P., Magnan, P., 2014. What controls distribution of larval and juvenile yellow perch? The role of habitat characteristics and spatial processes in a large, shallow lake. J. Great Lakes Res. 40 (1), 172–178 (Mar., URL http://www.sciencedirect.com/science/article/pii/S03801 33013001809 http:// linkinghub.elsevier.com/retrieve/pii/S0380133013001809). Persson, L., Eklov, P., 1995. Prey refuges affecting interactions between piscivorous perch and juvenile perch and roach. Ecology 76 (1), 70 (Jan., URL http://www.jstor.org/ stable/1940632?origin = crossref). Petr, T., 2000. Interactions between fish and aquatic macrophytes in inland waters: a review. Tech. rep. Food and Agriculture Organization of the United Nations, Roma (URL http://books.google.com/books?id = 7tvoQ8-_PB4C&pgis = 1). Pratt, T.C., Smokorowski, K.E., 2003. Fish habitat management implications of the summer habitat use by littoral fishes in a north temperate, mesotrophic lake. URL http:// www.nrcresearchpress.com/doi/pdf/10.1139/f03-022. Randall, R., Minns, C., Cairns, V., Moore, J., 1996. The relationship between an index of fish production and submerged macrophytes and other habitat features at three littoral areas in the Great Lakes. Can. J. Fish. Aquat. Sci. 53 (S1), 35–44 (URL http://www. nrc.ca/cgi-bin/cisti/journals/rp/rp2_abst_ e?cjfas_f95-271_53_ns_nf_cjfas53-96). Randall, R., Brousseau, C., Hoyle, J., 2012. Effect of aquatic macrophyte cover and fetch on spatial variability in the biomass and growth of littoral fishes in bays of Prince Edward County, Lake Ontario. Aquat. Ecosyst. Health Manage. 15 (4), 37–41 (URL http://www.tandfonline.com/doi/abs/10.1080/14634988.201 2.734691). Richard, G., Côté, D., Mingelbier, M., Jobin, B., Morin, J., Brodeur, P., 2011. Utilisation du sol dans la plaine inondable du lac Saint-Pierre (fleuve Saint-Laurent) durant les périodes 1950, 1964 et 1997: interprétation de photos aériennes, numérisation et préparation d'une base de données géoréférencées. Tech. rep., Gouvernement du Québec, Québec. Sass, G.G., Gille, C.M., Hinke, J.T., Kitchell, J.F., 2006. Whole-lake influences of littoral structural complexity and prey body morphology on fish predator-prey interactions. Ecol. Freshw. Fish 15 (3), 301–308 (Sep., URL http://apps.webofknowledge.com/full_ record.do?product = WOS&search_mode = GeneralSearch&qid = 4&SID = 2EszzoFrBECxqcCaEsG&page = 1&doc = 1). Schroeder, T., Cohen, W., Song, C., Canty, M., Yang, Z., 2006. Radiometric correction of multi-temporal Landsat data for characterization of early successional forest patterns in western Oregon. Remote Sens. Environ. 103 (1), 16–26 (Jul., URL http://linkinghub. elsevier.com/retrieve/pii/S0034425706 001179 http://www.sciencedirect.com/ science/article/pii/S0034425706001179). Silva, T.S., Melack, J.M., Novo, E.M.L.M., 2013. Responses of aquatic macrophyte cover and productivity to flooding variability on the Amazon floodplain. Glob. Chang. Biol. 19 (11), 3379–3389 Nov., URL http://www.ncbi.nlm.nih.gov/pubmed/23818397.

Smokorowski, K., Pratt, T., 2007. Effect of a change in physical structure and cover on fish and fish habitat in freshwater ecosystems — a review and meta-analysis. Environ. Rev. 15, 15–41 (Dec., URL http://www.nrcresearchpress.com/doi/abs/10.1139/a06007). Tall, L., Caraco, N., Maranger, R., 2011. Denitrification hot spots: dominant role of invasive macrophyte Trapa natans in removing nitrogen from a tidal river. Ecol. Appl. 21 (8), 3104–3114 (Dec., URL http://www.esajournals.org/doi/abs/10.1890/11-0061.1). Urho, L., 1996. Habitat shifts of perch larvae as survival strategy. Ann. Zool. Fenn. 62 (3), 419–460 (May, URL http://www.ingentaselect.com/rpsv/cgi-bin/cgi?ini = xref% &body = linker&reqdoi = 10.1357/0022240041446173 http://www.annzool.net/ PDF/anzf33/anzf33-329p.pdf). Vermaire, J.C., Prairie, Y., Gregory-Eaves, I., 2012. Diatom-inferred decline of macrophyte abundance in lakes of southern Quebec, Canada. Can. J. Fish. Aquat. Sci. 69 (3), 511–524 (Mar., URL http://www.nrcresearchpress.com/doi/abs/10.1139/f2011-1 69#.U62ysf7VOF5). Vis, C., Hudon, C., Carignan, R., Gagnon, P., 2007. Spatial analysis of production by macrophytes, phytoplankton and epiphyton in a large river system under different water-level conditions. Ecosystems 10 (2), 293–310 (URL b Go to ISI N:// 000248135200009). Vis, C., Cattaneo, A., Hudon, C., 2008. Shift from chlorophytes to cyanobacteria in benthic macroalgae along a gradient of nitrate depletion. J. Phycol. 44 (1), 38–44. Visser, F., Wallis, C., Sinnott, A.M., 2013. Optical remote sensing of submerged aquatic vegetation: opportunities for shallow clearwater streams. Limnol. Ecol. Manage. Inland Waters 43 (5), 388–398 (Sep., URL http://linkinghub.elsevier.com/retrieve/pii/ S0075951113 000467). Wang, N., Eckmann, R., 1994. Distribution of perch (Perca fluviatilis L.) during their first year of life in Lake Constance. Hydrobiologia 277 (3), 135–143 (Mar., URL http:// link.springer.com/10.1007/BF00007295). Weaver, M.J., Magnuson, J.J., Clayton, M.K., 1997. Distribution of littoral fishes in structurally complex macrophytes. Can. J. Fish. Aquat. Sci. 54 (10), 2277–2289 (Oct., URL http://www.nrcresearchpress.com/doi/abs/10.1139/f97-130). Winfield, I.J., 1986. The influence of simulated aquatic macrophytes on the zooplankton consumption rate of juvenile roach, Rutilus rutilus, rudd, Scardinius erythrophthalmus, and perch, Perca fluviatilis. J. Fish Biol. 29 (sa), 37–48 (Dec., URL http://onlinelibrary. wiley.com/doi/10.1111/j.1095-8649. 1986.tb04997.x/abstract http://doi.wiley.com/ 10.1111/j.1095-8649.1986.tb04997.x). Yamaki, A., Yamamuro, M., 2013. Floating-leaved and emergent vegetation as habitat for fishes in a eutrophic temperate lake without submerged vegetation. Limnology 14 (3), 257–268 (Mar., URL http://link.springer.com/10.1007/s10201-013-0403-2). Yin, Y., Winkelman, J., Langrehr, H., 2000. Long term resource monitoring program procedures: aquatic vegetation monitoring. Tech. rep., U.S. Geological Survey, Upper Midwest Environmental Sciences Center, La Crosse, WI. Zeileis, A., Kleiber, C., Jackman, S., 2008. Regression models for count data in R. J. Stat. Softw. 27 (8), 1–25 (Jul., URL http://apps.webofknowledge.com/full_record.do?product = WOS&search_mode = GeneralSearch&qid = 11&SID = 2EszzoFrBECxqcCaEsG&page = 1&doc = 1). Zhu, Z., Woodcock, C.E., 2012. Object-based cloud and cloud shadow detection in Landsat imagery. Remote Sens. Environ. 118, 83–94 (Mar., URL http://dx.doi.org/10.1016/j. rse.2011.10.028 http://linkinghub.elsevier.com/retrieve/pii/S0034425711003853). Zuur, A.F., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M., 2009. Mixed effects models and extensions in ecology with R. Springer New York, New York, NY (URL http:// link.springer.com/10.1007/978-0-387-87458-6).