EDP Sciences logo
Search
Menu
Home All issues Volume 24 (2024) Hydroécol. Appl., 24 (2024) 1 Full HTML
Open Access
Issue
Hydroécol. Appl.
Volume 24, 2024
Article Number 1
Number of page(s) 24
DOI https://doi.org/10.1051/hydro/2024001
Published online 19 March 2024

© EDF, hosted by EDP Sciences, 2024

Licence Creative CommonsThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1 Introduction

Reservoirs provide numerous services worldwide such as hydropower, water supply (e.g., irrigation, drinking water) or flood protection. With more than 45,000 large dams in the world at the beginning of the 20th century (Hjorth & Bengtsson, 2012), reservoirs need to be considered when managing global biodiversity (Liermann et al., 2012; McAllister et al., 2001; Wu et al., 2019). In reservoirs, and more generally in lentic ecosystems, littoral areas are characterized by a greater diversity of both abiotic and biotic structures than pelagic areas (Prchalová et al., 2009; Prchalová et al., 2008; Zohary & Gasith, 2014; Zohary & Ostrovsky, 2011). These structures constitute complex habitats and even hot spots of biodiversity in lakes (Gasith & Gafny, 1998; Schmieder, 2004), providing a high availability of trophic resources (Pierce et al., 1994). Consequently, in lakes and reservoirs, the achievement of many species life cycle depends on the littoral area.

The anthropogenic uses of reservoirs lead to artificial water level fluctuations (WLF) which are very different from those of natural lakes (Wetzel, 1990). Artificial WLF affect littoral habitat structure by bank erosion (Carmignani & Roy, 2017; Hirsch et al., 2017) and by modification of substrate and macrophytes compositions (Evtimova & Donohue, 2014). They can also temporarily prevent fish from accessing breeding (Hudon et al., 2005) or wintering habitats (Cott et al., 2008) and resting or refuge areas (Kaczka & Miranda, 2014; Logez et al., 2016). Thus, littoral area concentrates a lot of restoration projects to sustain fish fauna by improving habitat, e.g., through revegetation of banks, creation of shallow zones with habitat enhancement or deployment of floating islands (de Moraes et al., 2023; Halleraker et al., 2016; Radinger et al., 2023). Their success often remains uncertain (Vanderbosch & Galatowitsch, 2010), in part because it is dependent on multiple underlying management decisions, environmental conditions and difficulties to conduct such an evaluation. Predictive tools of fish population dynamics are valuable for improving success of restoration projects by testing the effects of different scenario, but they are still sparse (Strayer & Findlay, 2010).

Beyond specific reservoir ecosystems, many studies revealed the key role of habitat in population dynamics. Habitat quality drives the demographic parameters of fish populations, such as fecundity, growing or reproductive success (Farò et al., 2021; Hayes et al., 2009). To seek for an optimized physical habitat quality and availability, habitat is also involved in individual movements (Morales & Ellner, 2002; Nathan et al., 2008; Patterson et al., 2008). Finally, these movements contribute to the spatial distribution of species densities (Blanchard et al., 2021). Then, the modelling of individual fish movements frequently uses habitat quality as a main driver (Hovel & Regan, 2008; Railsback et al., 1999; Van Winkle et al., 1998).

In this study, we applied an individual-based approach to model fish movements in relation to habitat use. The individual approach is appropriate to simulate movements at the same resolution than habitat use (Grimm et al., 2006; Railsback et al., 1999). The main objective was to represent the spatial distribution of fish in a reservoir subject to water level fluctuations which frequently modify available habitats. The second objective was to keep a relatively simple calibration to facilitate future applications to other reservoirs concerned with restoration projects. The developed model, the 3PLake model, was supported by a high resolution dataset collected during previous studies on northern pike (Esox Lucius), European perch (Perca fluviatilis) and pikeperch (Sander lucioperca), three predatory fish species coexisting in a French reservoir (Roy, 2014; Westrelin et al., 2018). These species, common in European reservoirs (Overton et al., 2015; Skov & Nilsson, 2018), show contrasted habitat preferences. Northern pike are mainly littoral (Chapman & Mackay, 1984; Craig, 2008), European perch are more generalist (Imbrock et al., 2005; Pekcan-Hekim et al., 2005) and pikeperch are rather pelagic (Vehanen & Lahti, 2003). We hypothesized that these species showed different spatial distribution in relation with their habitat use and selectivity.

2 Material and methods

2.1 Study site

The Bariousses hydropower reservoir (45.33°N, 1.49°E) is located on the Vézère River in the western central part of France. The water level (WL) varied between altitudes 507.0 m and 513.5 m during the study from March 2012 to March 2014 (Fig. A.1 Appendix A). At the mean WL (511.5 m), the reservoir area was 0.866 km2, mean depth 7.1 m and maximum depth 19.4 m.

Based on the Secchi transparency depth, the littoral area of the reservoir was defined as the area connected to the bank with a depth lower than 2.5 m. The littoral area varied from 19.5% at low WL (507.0 m) to 13.3% at high WL (513.5 m) of the total reservoir area. Its habitats tended to become more homogeneous with a lowering structural complexity when the WL dropped: while gravel/pebble, immersed grass and emerging trees dominated at high levels, silt covered more than half the littoral area at low levels and emerging trees shrank (see Roy et al., 2021; and Westrelin et al., 2018 for a detailed description).

The ichtyofauna was dominated by two cyprinids, roach (Rutilus rutilus) and common bream (Abramis brama), and three percids, ruffe (Gymnocephalus cernua), European perch and pikeperch. Northern pike was another common piscivorous species living in the reservoir (Westrelin et al., 2021).

2.2 Species of interest

Northern pike (hereafter named pike) are an ambush predator preferring highly structured areas with vegetation (Chapman & Mackay, 1984; Cook & Bergersen, 1988; Keith et al., 2011): in lakes, they prefer the littoral zone and shallow waters (Craig, 1996; Westrelin et al., 2021), with some versatility observed among sites and sometimes a territorial behaviour (Eklöv, 1992; Skov & Koed, 2004). Pike are mostly active during day or twilight (Baktoft et al., 2012; Beaumont et al., 2005; Cook & Bergersen, 1988; Kobler et al., 2008). Some authors mention seasonal differences with a daytime activity in winter and a twilight activity in summer (Baktoft et al., 2012; Kobler et al., 2008; Říha et al., 2022). In addition, the structural complexity of habitats influences foraging behaviour and activity of pike (Říha et al., 2022).

European perch (hereafter named perch) frequent both littoral zones and neighboring deeper areas in summer depending on the phase of the day (Imbrock et al., 2005; Pekcan-Hekim et al., 2005; Westrelin et al., 2021). They prefer deeper waters in winter (Thorpe, 1977). Young European perch have a group foraging behaviour but large European perch are rather solitary (Craig, 1987; Eklöv, 1992).

Pikeperch prefer deep pelagic areas (Vehanen & Lahti 2003; Westrelin et al., 2021), to seek darkness (Craig, 1987) or prey (Huuskonen et al., 2019). They migrate to deep waters in winter (Říha et al., 2022; Vehanen & Lahti, 2003). Pikeperch are mostly active at dusk (Craig, 2000; Horký et al., 2008).

This knowledge highlights specific feeding strategies, habitat preferences or behaviour of the study species. Within each species, habitat preferences depend on the season (Říha et al., 2022; Westrelin et al., 2021) and the activity shows a diel rhythm (Baktoft et al., 2012; Horký et al., 2008; Jacobsen et al., 2015).

2.3 Tracking data

From March 2012 to March 2014, the three species were tracked by an array of 40 underwater VR2W 69kHz acoustic receivers in the Bariousses reservoir (France). Vemco V9P-2L (47 mm long, 6.3 g in the air) and V8-4L (20.5 mm long, 2 g in the air) acoustic transmitters were used. Specifically, data from 10 Northern pike, 22 European perch and 26 pikeperch (Appendix B) were retrieved to define habitat use and preferences based on their observed positions (Westrelin et al., 2021) and information on the littoral habitat as described in Westrelin et al. (2018). Detailed information on the tracking system and the tagging procedure is provided in Roy (2014), Roy et al., (2014) and Westrelin et al. (2018).

These tracking data were used to define, for each species, hourly and daily traveled distances, maps of habitat suitability index (HSI) that quantified the attractiveness of each reservoir location according to season and water level, and a density map.

Mean habitat preferences for the three species were quantified by selection ratios of various habitat characteristics: water depth over the whole lake, main substrate type, overhanging vegetation, tree stumps and emerging trees in the littoral zone. They were computed for each season, the season having a strong effect, and for each water level from altitude 507.0 to 513.5 m by 0.1 m. For details on this method, see Westrelin et al. (2018). For each species, these selection ratios were aggregated for each combination of season and water level in a HSI index that could be mapped over the lake on a 10 m x 10 m grid (264 maps per species, see Appendix C).

2.4 Model

The 3PLake model has been developed on the basis of a telemetry dataset with the willingness to use the fewest possible parameterization in order to avoid too much tuning and to facilitate the reproducibility. Only one parameter, the p selectivity coefficient, needs to be calibrated. The description of the 3Plake model below draws heavily from the ODD protocol (Overview, Design concepts, Details), but in a short version fitting to the relative simplicity of our model (Grimm et al., 2010; Grimm et al., 2020).

The 3Plake model has been developed in Java using SimAquaLife framework (Dumoulin, 2007).

2.4.1 Purpose

The purpose of the 3Plake model is to simulate individual movements of three fish species (pike, perch, pikeperch) depending on the season and the phase of the day that influence the activity of species depending on their diel cycle and the available habitat (synthesized in an HSI index based on water depth and littoral characteristics of substrate type, overhanging vegetation, tree stumps and emerging trees) (Fig. 1). Its objective is to predict the spatial distribution of fish individuals depending on the habitat availability which is closely linked to the water level. In this case study, the model was calibrated in the Bariousses reservoir.

thumbnail Fig. 1

Flow diagram of processes included in the 3Plake model. Input data are represented in green ellipses (dotted arrows indicate where they occurred). Tmax is the number of hourly time steps; Dmax is the hourly distance range; Nind is the number of individuals in the simulation; N Rep is the number of replicates of the simulation.

2.4.2 Entities, state variables, and scales

The reservoir area was discretized into 10 m x 10 m grid cells.

The entities are the three fish species of interest. Their movements depend on their habitat use and the habitat availability.

The time step is 1 h. The shift between seasons depends on the date: spring (from March 20 to 20 June), summer (from 21 June to 21 September), autumn (from 22 September to 20 December) and winter (from 21 December to 19 March). The phase of the day (dawn, day, dusk, night) changes with the hour according to sunrise/sunset hours specific to the date. Dawn, that includes the hour preceding the sunrise hour, the sunrise hour itself and the following hour, and dusk, that includes the hour preceding the sunset hour, the sunset hour itself and the following hour, both lasted 3 h.

The model is designed to be applied on an annual cycle. The number of individuals is kept constant in all simulations with no birth and no death.

2.4.3 Process overview and scheduling

One typical realistic year has been built for which at each hour, the season was set from the breakdown exposed in 2.4.2, the phase of the day from sunrise/sunset hours in 2012 and the water level from the time series observed in 2013. For each season combined to each water level and to each species corresponded a HSI map. For each season combined to each phase of the day and each species corresponded an hourly distance range that depends on the level of activity of the species and that defines available cells in which to move in the next hour. HSI maps and hourly distance range, associated with a selectivity coefficient (p) (more details in paragraph 2.4.6), drive individual fish movement.

The three species were modeled in independent simulations integrating all individuals from one species.

2.4.4 Design concepts

At each time step, individuals move to a cell which is in the hourly distance range from the previous occupied cell. It depends on the attractiveness of the cell for the chosen species, quantified by the HSI map, and on the p selectivity coefficient. The fish movement is assumed to be stochastic among the best available HSI cells. At each time step, the cell in which to move is randomly chosen among a fraction p (the selectivity coefficient) of the best HSI cells individual could access, depending on the hourly distance range. Individuals can stay in their initial cell, without moving: it has the same status as other cells, it is only described by its HSI value.

There is no interaction between individuals for perch and pikeperch. To consider the territorial behaviour of pike, two individuals of this species cannot share a same cell.

The position (a cell identifier) of each individual fish is recorded at each time step (a date and an hour).

2.4.5 Initialization

A total of 50 individuals, with similar characteristics to those given in Appendix B, is included in the simulation. This is based on an estimate made from a full draining of the reservoir in 1997. Initialization is a stochastic process. The starting cell is chosen randomly among all cells.

2.4.6 Input parameters and input data

Five input parameters are used in the 3Plake model (Tab. I).

The p coefficient is the only parameter used to calibrate the model. The calibration procedure is detailed in the Model evaluation paragraph (2.4.8.).

In addition, the 3Plake model needs four input data:

  • The ephemeris of the year to access to the sunrise and the sunset hours to define the phases of the day;

  • The hourly water level time series (see Appendix A for this case study);

  • The HSI map for each species and for each combination season*water level (264 maps per species) (see Appendix C for this case study);

  • The hourly distance range (Dmax) that defines the maximum distance between two consecutive positions of an individual in the model. It was calculated using the telemetry dataset. For each species, there is one value for each combination season*phase of the day (16 values per species) (Appendix D for this case study). This hourly distance range is set as the mean hourly distance estimated over data multiplied by 1.7. This value was chosen (i) to simulate mean hourly distances that approached the empirical values and (ii) to get 80% of the observed hourly distances smaller than Dmax for all combinations of season*phase of the day.

Table I

Input parameters of the 3Plake model and values used in the simulations for the case study.

2.4.7 Submodels

Environment and species characteristics are updated at each time step:

  • The date and the hour define the current season and phase of the day;

  • The water level changes following the hourly water level time series;

  • The species HSI map changes depending on the season and the water level;

  • The species Dmax is updated according to the season and phase of the day.

At each time step, individuals move to a cell within Dmax range from the previous position and which is in the lake at the water level corresponding to the time step. The candidate cells depend on their attractiveness, quantified by their HSI value, and on the p selectivity coefficient of the simulation. A lower value of p corresponds to a more selective habitat choice (towards higher quality). A higher value of p corresponds to a less selective habitat choice; a value of 100% for p means that the habitat choice is random among the accessible cells.

We ran 10 replicates for each value of p, giving a total of 210 simulations per species (Tab. I). All simulations were run with 50 individuals and the time period covered lasted one year.

We calculated the following metrics with simulation outputs:

  • The hourly traveled distance, which is the Euclidean distance between two consecutively occupied cells.

  • The daily traveled distance, which is the Euclidean distance between the cell of an individual at the first hour of a calendar day and the cell of the same individual 24 h later. The daily distance measures the net effect of the prospecting behaviour at a daily scale.

  • The species density map, which is the total number of positions (from all individuals of the considered species) recorded in each cell over all time steps of a simulation.

2.4.8 Model evaluation

The traveled distance metrics were used to evaluate the range of suitable p values for each species. The objective of the 3Plake model was to simulate the spatial distribution of each species. Therefore, the comparison between simulated and observed density maps was used for the final calibration of p for each species.

Hourly distance

Yearly mean hourly distance

For hourly distances, the evaluation of the simulations was first based on a comparison of the simulated means by species (mean and range across the 10 replicates for each value of p) with the observed means.

Because of non-continuous observed data, the observed hourly distance was calculated as the Euclidean distance between the first position of an hour and the first position of the next hour when the time lag separating these positions was between half an hour and one hour and a half. We assigned a zero value to all distances smaller or equal to 3 m, which was the telemetry system mean position error (Roy et al., 2014).

For a fair comparison of the simulations with the observations, the yearly mean observed hourly distance was defined as the mean of the four seasonal means to equally weigh the season. Then we calculated a 95% confidence interval for this observed mean using bootstrap (1,000 resamplings with replacement; the 95% confidence interval is thus given by the 25th and 975th values sorted in increasing order).

For each species we identified an optimal value of p simulating the least biased mean hourly distance, i.e., the value for which the mean of simulated hourly distances was the closest to the observed one (Westrelin et al., 2021). We also calculated 1,000 bootstrap values of the difference between the observed and simulated yearly mean hourly distances (resampling the individuals for the observed means, and sampling randomly one of the 10 replicates for the simulated means). The percentiles 2.5 and 97.5 of these 1,000 values defined a 95% confidence interval of this difference. The difference between the observed and simulated yearly mean hourly distances was significant when this 95% confidence interval did not include zero.

Hourly distance distribution

Then, for each species, we used boxplots to compare the distribution of simulated hourly distances for each value of p with that of observed ones.

For each species and each value of p, we built a boxplot with 40,000 simulated hourly distances that were randomly chosen out of all simulated replicates and time steps.

Additionally, we built a boxplot of observed hourly distances for each species. To equally weigh each season similarly to the simulations, we sampled 10,000 values with replacement for each of the four seasons, and we used these 40,000 values to build the boxplot.

Daily distance

We calculated the observed daily distance with a duration separating both positions comprised between half a day and a day and a half, because of non-continuous observed data. We applied the same principles as for the hourly distances to compare simulated and observed daily distances.

Density map

For the observed density map, we first calculated a density map for each season, for which the density in a cell was the total number of positions recorded in this cell (corrected by the probability of positioning in this cell by the telemetry system, see Roy et al. (2014) for this probability map). Then, we calculated the yearly map as the cell-by-cell weighed mean of the four seasonal maps. The weights were the inverse of the total number of recorded positions in the season to give each season the same weight. See Appendix E.

Next, we centered and standardized the simulated and observed density maps considering only cells with a non-null probability of positioning by the telemetry system (Roy et al., 2014). Then, for each cell, we computed the square differences of simulated and observed centered and standardized densities and summed them over all cells of the map. The Euclidean distance was the square root of this sum. Thus, for each species and each p, we obtained a mean and a range of this distance based on the 10 replicates. We also calculated a randomized model of this distance by randomizing the cell positions in the simulated density map before calculating its distance to the observed one (the used replicate of the simulation was randomly chosen each time). We repeated this randomization 1,000 times to calculate a mean and a 95% confidence interval. This bootstrap on the null model allowed to identify when the simulated distribution was closer to the observed distribution than would be a randomized distribution, in other words when the 3Plake model was more informative than the null model.

For each species, we identified the value of p and its 95% confidence interval for this density map distance using bootstrap. For each value of p, we randomly sampled one value of the distance between density maps among the 10 replicates. Then we identified the value of p which led to the minimum distance for this sample. We repeated this 1,000 times to get 1,000 values for p. For each species, the best value of p was the mean across these 1,000 bootstrap values, and the 95% confidence interval was defined by the percentiles 2.5 and 97.5.

Finally, in addition to the density map distance, we mapped the differences between simulated and observed centered and standardized densities for the optimal p value of each species to spatially identify them.

3 Results

3.1 Hourly distance

Yearly mean of simulated hourly distances increased with the selectivity parameter p. The three species presented the same general pattern: a steep rise for small values of p (up to 5–10%) and then a slight linear increase (Fig. 2 and Appendix F Fig. F1). The simulated distances were very close between the ten replicates (amplitude less than 1 m). But there was a larger uncertainty in the yearly mean of observed hourly distances (i.e., a large confidence interval; Fig. 2). Hence, the mean of simulated distances was not significantly different from the mean of observed distances for a large range of values of p. The numerical values obtained by bootstrap for the p values giving unbiased estimate of this yearly mean were p = 70% (range 10–100%) for pike, p = 60% (range 20–100%) for perch and p = 30% (range 10–100%) for pikeperch (Tab. II).

The distribution of simulated hourly distances was less dispersed than that of the observed ones whatever the value of p in this range for all species (Fig. 2 and Appendix F Fig. F1). Specifically, simulations captured fewer short distances (null distances in the observations) and could not get the largest ones.

thumbnail Fig. 2

Yearly mean of simulated hourly distances for European perch as a function of the selectivity parameter, p, varying from 1% to 100%. The solid red line represents the simulated mean (across all time steps, all individuals and all replicates) and the dashed red lines its range over the ten replicates (very close to the mean). The solid blue line is the yearly mean of observed hourly distances over all data and the dashed blue lines are its 95% confidence interval. Each grey boxplot represents the simulated distribution of the hourly distances for the 365 × 50 × 10 × 24 h. The blue boxplot stands for observed values. The extreme values of the observed boxplot are truncated and the maximum value, 1140, is displayed. Boxplots represent extreme values and 25%, 50% and 75% quantiles.
Table II

Parameter p (optimal value and range obtained by bootstrap) for which simulation and observation minimized the chosen metrics for each of the following criteria: mean hourly distance, mean daily distance and distance between density maps.

3.2 Daily distance

The yearly mean of simulated daily distances increased with p for all species. For a given p value, replicates gave very close distances (Fig. 3 and Appendix F Fig. 2). But there was a larger uncertainty in the observed yearly mean of daily distances. The mean of simulated distances was not significantly different from the mean of observed distances for a large range of p values. By bootstrapping, the optimal p values were p = 70% (range 10–100%) for pike, p = 20% (range 10–65%) for perch and p = 95% (range 50–100%) for pikeperch (Tab. II). In general, there was a large overlap of p value ranges between hourly and daily distances (Tab. II).

Similarly to the hourly distance, the distribution of simulated daily distances was less dispersed than that of the observed ones whatever the p value (Fig. 3 and Appendix F Fig. F2).

thumbnail Fig. 3

Yearly mean of simulated daily distances for perch as a function of the values of p varying from 1% to 100%. The solid red line is the simulated mean (across all time steps, all individuals and all replicates) and the dashed red lines its range over the ten replicates (very close to the mean). The solid blue line is the yearly mean of observed daily distances over all data for Perch, and the dashed blue lines its 95% confidence interval. Each grey boxplot represents the simulated distribution of the daily distances of the 365 × 50 × 10 days. The blue boxplot stands for observed values. The extreme values of the boxplots are truncated and the maximum value is displayed on each boxplot. Boxplots represent extreme values and 25%, 50% and 75% quantiles.

3.3 Distance between simulated and observed density maps

The distance to observed density maps was generally smaller for the 3Plake model simulations than for the null model (Fig. 4 and Appendix F Fig. F3), except when p>60% for pike. Based on the bootstrap, the optimal p values (minimizing the distance between observed and simulated maps) depended on the species: a smaller value for the pike (10%, range 10–10%), an intermediate value for the perch (20%, range 5–50%) and a higher value for the pikeperch (61%, range 25–90%) (Tab. II). The graphic visualization based on the mean over the 10 replicates (Fig. 4) can differ from the numerical results given by the bootstrap (Tab. II). So, pike seemed to have a spatial distribution linked to quite a high selectivity (low p) whereas pikeperch seemed to have a weaker selectivity with respect to HSI. Importantly, the best value identified for p for the density map criteria was in the range of the ones identified for hourly and daily distances (Tab. II). In other words, for the best p value identified for the density criteria, the yearly means of both hourly and daily distances did not differ significantly from the observed ones.

thumbnail Fig. 4

Distance between simulated and observed densities for perch as a function of p. The solid blue line is the mean over the 10 replicates and the dashed blue lines the range over the 10 replicates. The null model of this distance is also represented (mean over 1,000 repetitions and 95% confidence interval corresponding to the 2.5 and 97.5 quantiles of these 1,000 values, respectively in green solid line and dashed green lines).

3.4 Density maps

Differences between simulated and observed density maps were displayed for each species, for the optimal value of the p selectivity coefficient (Fig. 5). For the three species, some specific areas of the reservoir revealed higher simulated densities compared to observations. These areas were located mostly in the upstream half of the lake for perch (Fig. 5b) and pikeperch (Fig. 5c), whereas lower densities (compared to observations) were simulated in the downstream part of the lake, especially for perch (Fig. 5b). However, for pike (Fig. 5a) those areas of higher simulated densities (compared to observed ones) were situated in narrow littoral portions both upstream and downstream. Observed density maps are displayed in Appendix E.

thumbnail Fig. 5

Difference between simulated and observed density maps (unitless) for the optimal p value for each species, (a) pike (p = 10%), (b) perch (p = 20%) and (c) pikeperch (p = 61%). Positive values indicate larger density values for simulation than for observation. Negative values indicate smaller density values for simulation than for observation. Values smaller than −4 and larger than 4 are truncated to −4 and 4 on the map. Grey color indicates blind areas where no fish could be tracked by the telemetry system. Upstream part is to the North-East and downstream part to the South-West.

4 Discussion

4.1 Input data

The 3Plake model needs few and easy to get input data, apart from the HSI maps. In addition, HSI maps are key drivers of this model that assumes that fish movements are mainly driven by the habitat use. Telemetry data are very suitable to compute HSI maps, but they could also be calculated from physical data coupled with available habitat preference models (Laura et al., 2020; Rosenfeld, 2017).

4.2 Calibration of the 3Plake model

The calibration of the 3Plake model is only based on the p selectivity coefficient (fraction p of the cells with the highest HSI within the hourly distance range). The traveled distance metrics were used to evaluate the range of p values whereas, in the end, the density maps were used to choose the p value for each species. In the Bariousses reservoir and for each of the three species, the best p value evaluated by density maps matched with the range of p values according to hourly and daily distances. They were quite different between species with p = 10% for pike, p = 20% for perch and p = 61% for pikeperch. This illustrates that the three fish species showed different degrees of habitat selectivity (higher habitat selectivity for lower p values). For pike and perch, the values of p that led to the best match between simulated and observed density maps were lower than for pikeperch. This was in agreement with the higher habitat selectivity of pike and perch than pikeperch. It was particularly the case for pike (p = 10%, a value given by all 1,000 bootstrap iterations), a species that is known for being strongly attracted to highly structured habitats with vegetation and shallow waters (Inskip, 1982; Kobler et al., 2008). Perch revealed a slightly lower habitat selectivity (p = 20%, with a larger range of 5–50). Indeed, even if the species was attracted by littoral and near-littoral areas (Westrelin et al., 2018), it could use the pelagic zone (Jacobsen et al., 2015; Westrelin et al., 2021) and showed a more plastic behaviour (Craig, 2000). Furthermore, it was shown to adapt its habitat use to minimize interactions with other competitors as pike or pikeperch (Schulze et al., 2006). Pikeperch was the species with the lowest habitat selectivity (p = 61%) and with the largest confidence interval for p (25–90). Pikeperch frequents deeper waters (Craig, 1987) and the pelagic zone (Vehanen & Lahti, 2003) but can also be found in shallower waters depending on the season (Vehanen & Lahti, 2003; Westrelin et al., 2021). The density maps observed in the Bariousses reservoir confirmed the larger spectrum of habitats used by pikeperch (Appendix E, Fig. E1). Indeed, this species is known to live in a large range of aquatic ecosystems, including those widely anthropized (Kangur et al., 2007; Poulet et al., 2005).

The various degrees of habitat selectivity of species highlight their adaptability to their environment. Pike seemed to be the most specialist species, with the lowest p value and p value range. As a result, this species is potentially more vulnerable to pressure than perch or pikeperch. Indeed, throughout its native distribution area, in some cases, stocking supports pike populations to compensate environmental degradation (Guillerault et al., 2021; Pierce, 2012). As the reproductive success of pike is very dependent on littoral habitats, the species is particularly vulnerable in reservoirs subject to water level fluctuations (Casselman & Lewis 1996; Crane et al., 2015; Hudon et al., 2010); it can also be impacted by invasive species (Rehage et al., 2016). In contrast, pikeperch revealed a low habitat selectivity (higher p value), coherent with its recent increasing abundance in lakes and reservoirs (Huuskonen et al., 2019; Jeppesen et al., 2012) also linked to beneficial effects of climate change on this species (Pekcan-Hekim et al., 2005).

4.3 Gaps between observations and simulations

We observed some gaps between observations and simulations on evaluation metrics: the 3Plake model (i) did not represent the longest observed hourly distances, (ii) underestimated the motionless behaviour and (iii) slightly diverged from the observed density map in certain areas depending on species.

Hourly simulated distances were limited by maximum values (Dmax) that hindered the model from simulating the longest observed hourly distances. However, these maximum values were chosen to control the mean distance, and so, to relax this constraint by increasing these maxima would have led to an overestimation of the mean simulated distance compared to the observation. Daily distances were less impacted by this constraint and observed and simulated daily patterns were more in agreement than hourly ones. Even if daily distance is not defined as the numerical addition of 24-hourly segments, it takes into account these 24 simulated hourly movements which could help to simulate daily movements at longer distances, closer to the observed longer daily distances.

Motionless behaviours were also less often simulated than observed, especially in the hourly movements. Independent of HSI, since the probability to move to a cell was the same for all cells within the maximum distance range, the probability to simulate immobility (fish that stays at the same position) was low compared to the observations. Actually, in the input data, neither HSI nor maximum distance range are informative on this motionless behaviour. To capture this feature could be a further improvement of the model.

The 3Plake model tended to slightly overestimate the densities in the upstream part and some narrow littoral areas of the reservoir at the expense of the downstream part. Except for perch, the highest densities simulated towards the main tributary were in agreement with numerous studies based on observations (e.g., Brosse, 1999; Prchalová et al., 2009; Prchalová et al., 2008; Świerzowski et al., 2000; Vašek et al., 2016), in relation with the riverine origin of the fish fauna and the presence of an upstream-downstream productivity gradient (Vašek et al., 2006). The comparison between observed and simulated density maps (visually and with the Euclidean distance) could have been limited by the fact that, for each species, observed maps were built on a subsample of the species population. The subsamples could have been too small, at least in some seasons, to be representative. This could be enhanced in species showing territorial behaviour as pike (Eklöv, 1992; Skov & Koed, 2004), which moreover had the lowest sample size of ten individuals throughout the whole study, and artificially led to null density zones in the observed map. On the other hand, this could illustrate the complementarity of simulations and observations: by extracting habitat preferences from a set of individuals and then applying them to a representative sample of the population, the model is able to catch, or at least to give insight into, features that could not be handled with few observations.

4.4 Perspectives

The 3Plake model represents quite well the spatial distribution of three fish species in the Bariousses reservoir. However, the model has been designed to be applied on an annual scale and, as such, is not appropriate to focus on the movements during the spawning period, which constitutes a zoom on a particular period. Now, a process of spatial search and selection of spawning habitat could be added to the model and activated in the species adequate periods (Baetens et al., 2013; Clark et al., 1998; Clark et al., 2008). Such a model could be used to predict the effects of habitat restoration or water level scenarios on the spawning success of the three fish species. Similarly to spawning habitat, habitat restoration would lead to new HSI maps that could be built with fish habitat preferences already available in the reservoir. Comparing model outputs from simulations based on these restored maps with simulations run with non-restored ones would give insights into theoretical impacts of the restoration on spawning success. The same approach could be applied to compare different WLF scenarios.

5 Conclusions

The 3Plake model, calibrated using high resolution data, is only based on HSI maps and maximum hourly distances. Though quite simple, it simulated some interesting patterns of movement and habitat selection as discussed above. It is generic and could be applied to many other fish species as long as HSI maps and hourly distance range are available at each time step. It implicitly integrates some biotic interactions among individuals or among fish species, because model inputs are based on high resolution observations when all three species coexisted in the reservoir. Other biotic factors, such as prey distribution in the reservoir, are also implicitly partly integrated into HSI maps, because in relation with physical habitat.

The 3Plake spatially explicit IBM simulated yearly mean hourly and daily distances for pike, perch and pikeperch. Specific values of the selectivity parameter made it possible to simulate fish spatial densities close enough to the observed ones to be informative and consistent with the available habitat. A next step will be the addition of a spawning module to predict the impact of habitat restoration or water level management on the population dynamics. This final model could be a useful tool for stakeholders, to get insights into the effects of environmental measures on fish densities and fish spawning success.

Supplementary Material

An individual-based model to predict the spatial distribution of northern pike, European perch and pikeperch from habitat use in a reservoir subject to water level fluctuations Un modèle individu-centré pour prédire la distribution spatiale du brochet, de la perche commune et du sandre, à partir de l’utilisation de l’habitat dans un réservoir soumis à des fluctuations de niveau d’eau Access here

Acknowledgements

We would like to thank Guillaume Garbay who developed a first version of the model during his Master internship Nicolas Dumoulin who developed SimAquaLife and co-supervised this Master Gérard Henry who helped in transferring the model to Linux and Coralie Garron for a careful reading and English corrections. And we sincerely thank the two reviewers for their valuable comments that helped to improve the manuscript.

Fundings

This work was supported by HYNES (INRAE − EDF R&D) program.

References

  • Baetens J.M., Van Nieuland S., Pauwels I.S., De Baets B., Mouton A.M. & Goethals P.L.M., 2013. An individual-based model for the migration of pike (Esox lucius) in the river Yser, Belgium. Ecol Modell 258: 40–52. [CrossRef] [Google Scholar]
  • Baktoft H., Aarestrup K., Berg S., Boel M., Jacobsen L., Jepsen N., Koed A., Svendsen J.C. & Skov C., 2012. Seasonal and diel effects on the activity of northern pike studied by high-resolution positional telemetry. Ecol Freshw Fish 21 (3): 386–394. [CrossRef] [Google Scholar]
  • Beaumont W., Hodder K., Masters J., Scott L. & Welton J., 2005. Activity patterns in pike (Esox lucius), as determined by motion-sensing telemetry. Aquat Telemetry: Adv Appl. Rome: FAO/COISPA: 231–243. [Google Scholar]
  • Blanchard J.R., Santos R.O. & Rehage J.S., 2021. Sociability interacts with temporal environmental variation to spatially structure metapopulations: a fish dispersal simulation in an ephemeral landscape. Ecol Modell 443: 109458. [CrossRef] [Google Scholar]
  • Brosse S., 1999. Habitat, dynamique spatiale et structure des communautés pisciaires en milieu lacustre − Etude de la retenue de Pareloup (Aveyron, France). 203 pp. [Google Scholar]
  • Carmignani J.R. & Roy A.H., 2017. Ecological impacts of winter water level drawdowns on lake littoral zones: a review. Aquat Sci 79 (4): 803–824. [CrossRef] [Google Scholar]
  • Casselman J. & Lewis C.A., 1996. Habitat requirements of northern pike (Esox lucius). Can J Fish Aquat Sci 53: 161–174. [CrossRef] [Google Scholar]
  • Chapman C. & Mackay W., 1984. Direct observation of habitat utilization by Northern Pike. Copeia 1984: 255. [CrossRef] [Google Scholar]
  • Chapman C. & Mackay W.C., 1984. Versatility in habitat use by a top aquatic predator, Esox lucius L. J Fish Biol 25: 109–115. [CrossRef] [Google Scholar]
  • Clark M., Rose K., Chandler J., Richter T., Orth D. & Winkle W., 1998. Simulating smallmouth bass reproductive success in reservoirs. Environ Biol Fishes 51: 161–174. [CrossRef] [Google Scholar]
  • Clark M.E., Rose K.A., Chandler J.A., Richter T.J., Orth D.J. & Van Winkle W., 2008. Water-level fluctuation effects on centrarchid reproductive success in reservoirs: a modeling analysis. North Am J Fish Manag 28 (4): 1138–1156. [CrossRef] [Google Scholar]
  • Cook M.F. & Bergersen E.P., 1988. Movements, habitat selection, and activity periods of Northern Pike in Eleven Mile Reservoir, Colorado. Trans Am Fish Soc 117 (5): 495–502. [CrossRef] [Google Scholar]
  • Cott P.A., Sibley P.K., Somers W.M., Lilly M.R. & Gordon A.M., 2008. A review of water level fluctuations on aquatic biota with an emphasis on fishes in ice-covered lakes. J Am Water Resour Assoc 44 (2): 343–359. [CrossRef] [Google Scholar]
  • Craig J., 1996. Pike: Biology and Exploitation. Springer Netherlands. [Google Scholar]
  • Craig J., 2000. Percid Fishes: Systematics, Ecology and Exploitation. Wiley, 370 pp. [Google Scholar]
  • Craig J.F., 1987. The Biology of Perch and Related Fish. London: Croom Helm. [Google Scholar]
  • Craig J.F., 2008. A Short Review of Pike Ecology. Hydrobiologia 601 (1): 5–16. [CrossRef] [Google Scholar]
  • Crane D.P., Miller L.M., Diana J.S., Casselman J.M., Farrell J.M., Kapuscinski K.L. & Nohner J.K., 2015. Muskellunge and northern pike ecology and management: important issues and research needs. Fisheries 40 (6): 258–267. [CrossRef] [Google Scholar]
  • de Moraes K.R., Souza A.T., Muška M., Hladík M., Čtvrtlíková M., Draštík V., Kolařík T., Kučerová A., Krolová M., Sajdlová Z., Šmejkal M. & Kubečka J., 2023. Artificial floating islands: a promising tool to support juvenile fish in lacustrine systems. Hydrobiologia 850 (9): 1969–1984. [CrossRef] [Google Scholar]
  • Dumoulin N., 2007. SimAquaLife: a framework for aquatic organisms spatial dynamics modelling. Revue des Sciences et Technologies de l’Information − Série TSI : Technique et Science Informatiques, 26 (6): 701–721. [Google Scholar]
  • Eklöv P., 1992. Group foraging versus solitary foraging efficiency in piscivorous predators: the perch, Perca fluviatilis, and pike, Esox lucius, patterns. Anim Behav 44: 313–326. [CrossRef] [Google Scholar]
  • Evtimova V. & Donohue I., 2014. Quantifying ecological responses to amplified water level fluctuations in standing waters: an experimental approach. J Appl Ecol 51. [Google Scholar]
  • Farò D., Zolezzi G. & Wolter C., 2021. How much habitat does a river need? A spatially-explicit population dynamics model to assess ratios of ontogenetical habitat needs. J Environ Manag 286: 112100. [CrossRef] [Google Scholar]
  • Gasith A. & Gafny S., 1998. Importance of physical structures in lakes: the case of lake kinneret and general implications. In: The Structuring Role of Submerged Macrophytes in Lakes (E. Jeppesen, M. Søndergaard, M. Søndergaard, K. Christoffersen Eds.) pp. 331–338. [CrossRef] [Google Scholar]
  • Grimm V., Berger U., Bastiansen F., Eliassen S., Ginot V., Giske J., Goss-Custard J., Grand T., Heinz S.K., Huse G., Huth A., Jepsen J.U., Jørgensen C., Mooij W.M., Müller B., Pe’er G., Piou C., Railsback S.F., Robbins A.M., Robbins M.M., Rossmanith E., Rüger N., Strand E., Souissi S., Stillman R.A., Vabø R., Visser U. & DeAngelis D.L., 2006. A standard protocol for describing individual-based and agent-based models. Ecol Modell 198 (1): 115–126. [CrossRef] [Google Scholar]
  • Grimm V., Berger U., DeAngelis D.L., Polhill J.G., Giske J. & Railsback S.F., 2010. The ODD protocol: a review and first update. Ecol Modell 221 (23): 2760–2768. [CrossRef] [Google Scholar]
  • Grimm V., Railsback S.F., Vincenot C.E., Berger U., Gallagher C., DeAngelis D.L., Edmonds B., Ge J., Giske J., Groeneveld J., uuml, rgen, Johnston A.S.A., Milles A., Nabe-Nielsen J., Polhill J.G., Radchuk V., Rohw, auml, der M.-S., Stillman R.A., Thiele J. C., Ayll, oacute D. n., 2020. The ODD protocol for describing agent-based and other simulation models: a second update to improve clarity, replication, and structural realism. J Artif Soc Soc Simul 23 (2): 7. [CrossRef] [PubMed] [Google Scholar]
  • Guillerault N., Loot G., Blanchet S., Millet P., Musseau C., Cucherousset J. & Santoul F., 2021. Efficiency of Northern pike (Esox lucius) stocking in metropolitan France at large spatial and temporal scales. Fish Manag Ecol 28 (5): 486–495. [CrossRef] [Google Scholar]
  • Halleraker J.H., van de Bund F W., Bussettini M., Gosling R., Döbbelt-Grüne S., Hensman J., Kling J., Koller-Kreimel V. & Pollard P., 2016. Working group ECOSTAT report on common understanding of using mitigation measures for reaching good ecological potential for heavily modified water bodies. 104. [Google Scholar]
  • Hayes D., Jones M., Lester N., Chu C., Doka S., Netto J., Stockwell J., Thompson B., Minns C., Shuter B. & Collins N., 2009. Linking fish population dynamics to habitat conditions: insights from the application of a process-oriented approach to several Great Lakes species. Rev Fish Biol Fish 19 (3): 295–312. [CrossRef] [Google Scholar]
  • Hirsch P.E., Eloranta A.P., Amundsen P.-A., Brabrand Å., Charmasson J., Helland I.P., Power M., Sánchez-Hernández J., Sandlund O.T., Sauterleute J.F., Skoglund S., Ugedal O. & Yang H., 2017. Effects of water level regulation in alpine hydropower reservoirs: an ecosystem perspective with a special emphasis on fish. Hydrobiologia 794 (1): 287–301. [CrossRef] [Google Scholar]
  • Hjorth P. & Bengtsson L., 2012. Large dams, statistics and critical review. In: Encyclopedia of Lakes and Reservoirs (L. Bengtsson, R.W. Herschy,R.W. Fairbridge Eds.) pp. 475–479. [CrossRef] [Google Scholar]
  • Horký P., Slavík O. & Bartoš L., 2008. A telemetry study on the diurnal distribution and activity of adult pikeperch, Sander lucioperca (L.), in a riverine environment. Hydrobiologia 614 (1): 151. [CrossRef] [Google Scholar]
  • Hovel K.A. & Regan H.M., 2008. Using an individual-based model to examine the roles of habitat fragmentation and behavior on predator-prey relationships in seagrass landscapes. Landsc Ecol 23 (1): 75–89. [CrossRef] [Google Scholar]
  • Hudon C., Armellin A., Gagnon P. & Patoine A., 2010. Variations in water temperatures and levels in the St. Lawrence River (Québec, Canada) and potential implications for three common fish species. Hydrobiologia 647 (1): 145–161. [CrossRef] [Google Scholar]
  • Hudon C., Gagnon P., Amyot J.-P., Létourneau G., Jean M., Plante C., Rioux D. & Deschênes M., 2005. Historical changes in herbaceous wetland distribution induced by hydrological conditions in Lake Saint-Pierre (St. Lawrence River, Quebec, Canada). Hydrobiologia 539 (1): 205–224. [CrossRef] [Google Scholar]
  • Huuskonen H., Piironen J., Syväranta J., Eronen R., Biasi C., Kiiskinen P., Kortet R. & Vainikka A., 2019. Diet and movements of pikeperch (Sander lucioperca) in a large oligotrophic lake with an exceptionally high pikeperch yield. Ecol Freshw Fish 28 (4): 533–543. [CrossRef] [Google Scholar]
  • Imbrock F., Appenzeller A. & Eckmann R., 2005. Diel and seasonal distribution of perch in Lake Constance: a hydroacoustic study and in situ observations. J Fish Biol 49: 1–13. [Google Scholar]
  • Inskip P.D., 1982. Habitat suitability index models: northern pike. In: FWS/OBS. [Google Scholar]
  • Jacobsen L., Berg S., Baktoft H. & Skov C., 2015. Behavioural strategy of large perch Perca fluviatilis varies between a mesotrophic and a hypereutrophic lake. J Fish Biol 86 (3): 1016–1029. [CrossRef] [PubMed] [Google Scholar]
  • Jeppesen E., Mehner T., Winfield I.J., Kangur K., Sarvala J., Gerdeaux D., Rask M., Malmquist H.J., Holmgren K. & Volta P., 2012. Impacts of climate warming on the long-term dynamics of key fish species in 24 European lakes. Hydrobiologia 694 (1): 1–39. [CrossRef] [Google Scholar]
  • Kaczka L.J. & Miranda L.E., 2014. Size of age-0 crappies (Pomoxis spp.) relative to reservoir habitats and water levels. J Freshw Ecol 29 (4): 525–534. [CrossRef] [Google Scholar]
  • Kangur K., Park Y.-S., Kangur A., Kangur P. & Lek S., 2007. Patterning long-term changes of fish community in large shallow Lake Peipsi. Ecol Modell 203 (1-2): 34–44. [CrossRef] [Google Scholar]
  • Keith P., Persat H., Feunteun E. & Allardi J., 2011. Les Poissons d’eau douce de France. [Google Scholar]
  • Kobler A., Klefoth T., Wolter C., Fredrich F. & Arlinghaus R., 2008. Contrasting pike (Esox lucius L.) movement and habitat choice between summer and winter in a small lake. Hydrobiologia 601 (1): 17. [CrossRef] [Google Scholar]
  • Liermann C.R., Nilsson C., Robertson J. & Ng R., 2012. Implications of dam obstruction for global freshwater fish diversity. BioScience 62: 539–548. [CrossRef] [Google Scholar]
  • Logez M., Roy R., Tissot L. & Argillier C., 2016. Effects of water-level fluctuations on the environmental characteristics and fish-environment relationships in the littoral zone of a reservoir. Fundam Appl Limnol 189 (1): 37–49. [CrossRef] [Google Scholar]
  • McAllister D., Craig J., Davidson N., Delany S. & Seddon M., 2001. Biodiversity Impacts of Large Dams. 64 pp. [Google Scholar]
  • Morales J.M. & Ellner S.P., 2002. Scaling up animal movements in heterogeneous landscapes: the importance of behavior. Ecology 83 (8): 2240–2247. [CrossRef] [Google Scholar]
  • Nathan R., Getz W.M., Revilla E., Holyoak M., Kadmon R., Saltz D. & Smouse P.E., 2008. A movement ecology paradigm for unifying organismal movement research. Proc Natl Acad Sci 105 (49): 19052–19059. [CrossRef] [PubMed] [Google Scholar]
  • Overton J., Toner D., Policar T. & Kucharczyk D., 2015. Commercial production: factors for success and limitations in European percid fish culture. 881–890. [Google Scholar]
  • Patterson T.A., Thomas L., Wilcox C., Ovaskainen O. & Matthiopoulos J., 2008. State-space models of individual animal movement. Trends Ecol Evol 23 (2): 87–94. [CrossRef] [PubMed] [Google Scholar]
  • Pekcan-Hekim Z., Horppila J., Nurminen L. & Niemistö J., 2005. Diel changes in habitat preference and diet of perch (Perca fluviatilis), roach (Rutilus rutilus) and white bream (Abramis björkna). Archiv für Hydrobiologie Special Issues Advances in Limnology, 59: 173–187. [Google Scholar]
  • Pierce C.L., Rasmussen J.B. & Leggett W.C., 1994. Littoral fish communities in southern quebec lakes: relationships with limnological and prey resource variables. Can J Fish Aquat Sci 51 (5): 1128–1138. [CrossRef] [Google Scholar]
  • Pierce R.B., 2012. Northern pike: ecology, conservation, and management history. Environ Hist 18 (3): 650–651. [Google Scholar]
  • Plichard L., Forcellini M., Le Coarer Y., Capra H., Carrel G., Ecochard R. & Lamouroux N., 2020. Predictive models of fish microhabitat selection in multiple sites accounting for abundance overdispersion. River Res Appl 36 (7):1056- 1075 [Google Scholar]
  • Poulet N., Lek S. & Argillier C., 2005. Pikeperch habitat use within a canal network in spring. J Fish Biol 67 (5): 1460–1474. [CrossRef] [Google Scholar]
  • Prchalová M., Kubečka J., Cech M.F., Drastík J.V., Hohausova E.M., Peterka J.J.R., Tuser M.M. & Vasek M., 2009. The effect of depth, distance from dam and habitat on spatial distribution of fish in an artificial reservoir. Ecol Freshw Fish 18: 247–260. [CrossRef] [Google Scholar]
  • Prchalová M., Kubečka J., Vašek M., Peterka J., Sedá J., Jůza T., Říha M., Jarolím O., Tušer M., Kratochvíl M., Čech M., Draštík V., Frouzová J. & Hohausová E., 2008. Distribution patterns of fishes in a canyon-shaped reservoir. J Fish Biol 73 (1): 54–78. [CrossRef] [Google Scholar]
  • Radinger J., Matern S., Klefoth T., Wolter C., Feldhege F., Monk C.T. & Arlinghaus R., 2023. Ecosystem-based management outperforms species-focused stocking for enhancing fish populations. Science 379 (6635): 946–951. [CrossRef] [PubMed] [Google Scholar]
  • Railsback S.F., Lamberson R.H., Harvey B.C. & Duffy W.E., 1999. Movement rules for individual-based models of stream fish. Ecol Modell 123 (2): 73–89. [CrossRef] [Google Scholar]
  • Rehage J.S., Blanchard J.R., Boucek R.E., Lorenz J.J. & Robinson M., 2016. Knocking back invasions: variable resistance and resilience to multiple cold spells in native vs. nonnative fishes. Ecosphere 7 (6): e01268. [CrossRef] [Google Scholar]
  • Říha M., Rabaneda-Bueno R., Jarić I., Souza A.T., Vejřík L., Draštík V., Blabolil P., Holubová M., Jůza T., Gjelland K.Ø., Rychtecký P., Sajdlová Z., Kočvara L., Tušer M., Prchalová M., Seďa J. & Peterka J., 2022. Seasonal habitat use of three predatory fishes in a freshwater ecosystem. Hydrobiologia 849 (15): 3351–3371. [CrossRef] [Google Scholar]
  • Rosenfeld J.S., 2017. Developing flow-ecology relationships: implications of nonlinear biological responses for water management. Freshw Biol 62 (8): 1305–1324. [CrossRef] [Google Scholar]
  • Roy R., 2014. Distribution spatiale et activité des poissons en milieu lacustre : impacts des facteurs environnementaux à partir d’une approche multi-échelle. Application à la retenue des Bariousses. Thèse de doctorat, Aix Marseille université. 224 pp. [Google Scholar]
  • Roy R., Beguin J., Argillier C., Tissot L., Smith F., Smedbol S. & De Oliveira E., 2014. Testing the VEMCO Positioning System: spatial distribution of the probability of location and the positioning error in a reservoir. Anim Biotelemetry 2 (1): 1–7. [CrossRef] [Google Scholar]
  • Roy R., Tissot L. & Argillier C., 2021. Environmental drivers of fish spatial distribution and activity in a reservoir with water level fluctuations. Hydroécol Appl 21: 25–46. [CrossRef] [EDP Sciences] [Google Scholar]
  • Schmieder K., 2004. European lake shores in danger − concepts for a sustainable development. Limnologica 34: 3–14. [Google Scholar]
  • Schulze T., Baade U., Dörner H., Eckmann R., Haertel-Borer S.S., Hölker F. & Mehner T., 2006. Response of the residential piscivorous fish community to introduction of a new predator type in a mesotrophic lake. Can J Fish Aquat Sci 63 (10): 2202–2212. [CrossRef] [Google Scholar]
  • Skov C. & Koed A., 2004. Habitat use of 0+ year pike in experimental ponds in relation to cannibalism, zooplankton, water transparency and habitat complexity. J Fish Biol 64: 448–459. [CrossRef] [Google Scholar]
  • Skov C. & Nilsson A., 2018. Biology and Ecology of Pike. CRC Press, 410 pp. [Google Scholar]
  • Strayer D.L. & Findlay S.E.G., 2010. Ecology of Freshwater Shore Zones. Aquat Sci 72 (2): 127–163. [CrossRef] [Google Scholar]
  • Świerzowski A., Godlewska M. & Półtorak T., 2000. The relationship between the spatial distribution of fish, zooplankton and other environmental parameters in the Solina reservoir, Poland. Aquat Living Resour 13 (5): 373–377. [CrossRef] [Google Scholar]
  • Thorpe J.E., 1977. Morphology, physiology, behavior and ecology of Perca flluviatilis and P. flavescens Mitchill. J Fish Board Can 34: 1504–1514. [CrossRef] [Google Scholar]
  • Van Winkle W., Jager Y. & Holcomb B., 1998. An individual-based model for sympatric populations of brown and rainbow trout for instream flow assessment: model description and calibration. Ecol Modell 10: 175–207. [CrossRef] [Google Scholar]
  • Vanderbosch D. & Galatowitsch S., 2010. An Assessment of urban lakeshore restorations in minnesota. Ecol Restor 28: 71–80. [CrossRef] [Google Scholar]
  • Vašek M., Kubečka J., Matěna J. & Seďa J., 2006. Distribution and diet of 0+ Fish within a Canyon-Shaped European Reservoir in Late Summer. Int Rev Hydrobiol 91 (2): 178–194. [CrossRef] [Google Scholar]
  • Vašek M., Prchalová M., Říha M., Blabolil P., Čech M., Draštík V., Frouzová J., Jůza T., Kratochvíl M., Muška M., Peterka J., Sajdlová Z., Šmejkal M., Tušer M., Vejřík L., Znachor P., Mrkvička T., Seďa J. & Kubečka J., 2016. Fish community response to the longitudinal environmental gradient in Czech deep-valley reservoirs: implications for ecological monitoring and management. Ecol Indic 63: 219–230. [CrossRef] [Google Scholar]
  • Vehanen T. & Lahti M., 2003. Movements and habitat use by pikeperch (Stizostedion lucioperca (L.)) in a hydropeaking reservoir. Ecol Freshw Fish 12 (3): 203–215. [CrossRef] [Google Scholar]
  • Westrelin S., Cucherousset J., Roy R., Tissot L., Santoul F. & Argillier C., 2021. Habitat partitioning among three predatory fish in a temperate reservoir. Ecol Freshw Fish 31 (1): 129–142. [Google Scholar]
  • Westrelin S., Roy R., Tissot-Rey L., Bergès L. & Argillier C., 2018. Habitat use and preference of adult perch (Perca fluviatilis L.) in a deep reservoir: variations with seasons, water levels and individuals. Hydrobiologia 809 (1): 121–139. [CrossRef] [Google Scholar]
  • Wetzel R.G., 1990. Reservoir ecosystems: conclusions and speculations. In: Reservoir limnology: ecological perspective (K.W. Thornton, B.L. Kimmel, F.E. Payne Eds.) 11 pp. [Google Scholar]
  • Wu H., Chen J., Xu J., Zeng G., Sang L., Liu Q., Yin Z., Dai J., Yin D., Liang J. & Ye S., 2019. Effects of dam construction on biodiversity: a review. J Clean Prod 221: 480–489. [CrossRef] [Google Scholar]
  • Zohary T. & Gasith A., 2014. The littoral zone. In: Lake Kinneret: Ecology and Management (T. Zohary, A. Sukenik, T. Berman, A. Nishri Eds.) pp. 517–532. [CrossRef] [Google Scholar]
  • Zohary T. & Ostrovsky I., 2011. Ecological impacts of excessive water level fluctuations in stratified freshwater lakes. Inland Waters 1: 47–59. [Google Scholar]

All Tables

Table I

Input parameters of the 3Plake model and values used in the simulations for the case study.

In the text
Table II

Parameter p (optimal value and range obtained by bootstrap) for which simulation and observation minimized the chosen metrics for each of the following criteria: mean hourly distance, mean daily distance and distance between density maps.

In the text

All Figures

thumbnail Fig. 1

Flow diagram of processes included in the 3Plake model. Input data are represented in green ellipses (dotted arrows indicate where they occurred). Tmax is the number of hourly time steps; Dmax is the hourly distance range; Nind is the number of individuals in the simulation; N Rep is the number of replicates of the simulation.
In the text
thumbnail Fig. 2

Yearly mean of simulated hourly distances for European perch as a function of the selectivity parameter, p, varying from 1% to 100%. The solid red line represents the simulated mean (across all time steps, all individuals and all replicates) and the dashed red lines its range over the ten replicates (very close to the mean). The solid blue line is the yearly mean of observed hourly distances over all data and the dashed blue lines are its 95% confidence interval. Each grey boxplot represents the simulated distribution of the hourly distances for the 365 × 50 × 10 × 24 h. The blue boxplot stands for observed values. The extreme values of the observed boxplot are truncated and the maximum value, 1140, is displayed. Boxplots represent extreme values and 25%, 50% and 75% quantiles.
In the text
thumbnail Fig. 3

Yearly mean of simulated daily distances for perch as a function of the values of p varying from 1% to 100%. The solid red line is the simulated mean (across all time steps, all individuals and all replicates) and the dashed red lines its range over the ten replicates (very close to the mean). The solid blue line is the yearly mean of observed daily distances over all data for Perch, and the dashed blue lines its 95% confidence interval. Each grey boxplot represents the simulated distribution of the daily distances of the 365 × 50 × 10 days. The blue boxplot stands for observed values. The extreme values of the boxplots are truncated and the maximum value is displayed on each boxplot. Boxplots represent extreme values and 25%, 50% and 75% quantiles.
In the text
thumbnail Fig. 4

Distance between simulated and observed densities for perch as a function of p. The solid blue line is the mean over the 10 replicates and the dashed blue lines the range over the 10 replicates. The null model of this distance is also represented (mean over 1,000 repetitions and 95% confidence interval corresponding to the 2.5 and 97.5 quantiles of these 1,000 values, respectively in green solid line and dashed green lines).
In the text
thumbnail Fig. 5

Difference between simulated and observed density maps (unitless) for the optimal p value for each species, (a) pike (p = 10%), (b) perch (p = 20%) and (c) pikeperch (p = 61%). Positive values indicate larger density values for simulation than for observation. Negative values indicate smaller density values for simulation than for observation. Values smaller than −4 and larger than 4 are truncated to −4 and 4 on the map. Grey color indicates blind areas where no fish could be tracked by the telemetry system. Upstream part is to the North-East and downstream part to the South-West.
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.