The effect of water level changes on the density of newly hatched European whitefish (<i>Coregonus lavaretus</i> (L.)) larvae in unregulated and regulated lakes

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ The effect of water level changes on the density of newly hatched European whitefish (<i>Coregonus lavaretus</i> (L.)) larvae in unregulated and regulated lakes © EDP Sciences, 2024 Published version Väänänen, Tuula; Marjomäki, Timo J.; Ranta, Tomi; Karjalainen, Juha Väänänen, T., Marjomäki, T. J., Ranta, T., & Karjalainen, J. (2024). The effect of water level changes on the density of newly hatched European whitefish (Coregonus lavaretus (L.)) larvae in unregulated and regulated lakes. International Journal of Limnology, 60, Article 23. https://doi.org/10.1051/limn/2024019 2024 RESEARCH ARTICLE The effect of water level changes on the density of newly hatched European whitefish (Coregonus lavaretus (L.)) larvae in unregulated and regulated lakes Tuula Väänänen 1,* , Timo J. Marjomäki 1 , Tomi Ranta 2 and Juha Karjalainen 1 1 University of Jyväskylä, Department of Biological and Environmental Science, PO BOX 35 40014 University of Jyväskylä, Finland 2 Fisheries Centre of Häme, Kustaankuja 2 17200 Vääksy, Finland Received: 17 January 2024; Accepted: 26 August 2024 Abstract –The effect of the water levels during the period from spawning to hatching of whitefish (Coregonus lavaretus (L.)) on the density of newly hatched larvae was examined from a 23-year time series. The density of larvae was estimated in four lakes with contrasting water level regulation regimes in Central Finland using depth zone stratified bongo and tube net sampling in 2000–2022 and 2008–2022. In the regulated Tehinselkä, larval density was also modeled as a function of the whitefish spawning stock, vendace stock indices, and water levels. Larger water level maximum drops during winter were observed in three regulated lakes (41–68 cm) than in the unregulated lake (35cm). The average larval densities were highest in the unregulated Lake Southern Konnevesi and regulated Tehinselkä, with densities >20 individuals ha 1 . The regulated lakes Ruotsalainen and Puula had lower average densities, <10 individuals ha 1 . Significant synchrony in the inter-annual variation in the density time series was observed between Tehinselkä and S. Konnevesi as well as Tehinselkä and Ruotsalainen. None of studied water level variables were associated with larval density in any lake. Thus, the small-scale regulation in these Finnish lakes did not show any direct effects on the production of whitefish larvae. Our analysis did not provide information on the potential effects of water level regulation on later life. For example, it did not cover the impact on whitefish food resources or the abundance of competitors and predators regulating whitefish growth and mortality in the juvenile stage. Keywords: Coregoninae / egg incubation / fish larvae / spawning stock indices / water level regulation 1 Introduction Several natural factors, such as changes in food quantity and quality, predation, and competition within and between species affect the variation in the abundance and structure of fish populations. In addition, the impact of human activity on fish populations is undeniable, and these effects can be either direct or indirect (Eckmann, 2013;Senlu et al., 2022). An anthropogenic impact on the hydro-morphological and biological factors of lakes and rivers is the regulation of water level for energy production in hydroelectric power plants (Vuorio et al., 2015;Kuriqi et al., 2021). The effects of water level regulation are case-specific and complex, influenced by factors such as amplitude, timing, frequency, and rate of change in water level, as well as morphometry, geology, and biotic community of the lake (Hirsch et al., 2017). European whitefish (Coregonus lavaretus (L.)) is a polymorphic fish species, with its various ecomorphs differing from each other morphologically and ecologically, for example for their morphological measurements, diet, trophic and thermal niches, and spawning habitat (Svärdson, 1979; Kahilainen and Østbye, 2006;Kahilainen et al., 2014; Bitz-Thorsen et al., 2020). European whitefish can be littoral or profundal spawners and their spawning depth is known to vary from a depth of about half a meter up to 200 m (Eckmann and Rösch, 1998;Valkeajärvi et al., 2001). European whitefish spawn in autumn, and their eggs incubate over the winter at the bottom of the lake for five to seven months in boreal regions (Karjalainen et al., 2015). Particularly in lakes with long incubation periods, various factors can impact the development and survival of the whitefish eggs, including the trophic state of the lake, sediment quality, oxygen concentration, and egg predation (Müller, 1992,C. sp.; Mills et al., 2002, C. clupeaformis (Mitchill); Wahl and Löffler, 2009, C. lavaretus;Karjalainen et al., 2021,C. lavaretus). The littoral habitat and biota and littoral spawning fish species and morphs in general, seem to be most vulnerable and sensitive to water level regulation (Carmignani and Roy, 2017; *Corresponding author: tuula.m.vaananen@jyu.fi Special issue - Biology and Management of Coregonid Fishes - 2023 Guest editors: Orlane Anneville, Chloé Goulon, Juha Karjalainen, Jean Guillard, Jared T. Myers and Jason Stockwell Int. J. Lim. 2024, 60, 23 ©EDP Sciences, 2024 https://doi.org/10.1051/limn/2024019 Available online at: www.limnology-journal.org This 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. Hirsch et al., 2017). Strong water level regulation has been associated with poor reproduction in whitefish populations (Sutela and Mutenia, 2001,C. lavaretus;Sutela et al., 2002, C. peled (Gmelin); Mills et al., 2002, C. clupeaformis; Winfield et al., 2004,C. lavaretus). Often the water level drops in regulated lakes towards spring, and the eggs of autumnspawning salmonids can be exposed to dehydration and freezing (Brabrand et al., 2002). The effect of thick ice alone reaching the bottom in the shallow littoral area of the lake, without any decrease in water level, has been estimated to cause 5% egg destruction of littoral spawning European whitefish in Lake Päijänne (Valkeajärvi et al., 2001), where whitefish spawn in the shallow littoral zone near the shoreline, and the maximum egg density has been observed at a depth of 0.75 m. The magnitude of the decrease in water levels during the winter and the spawning depth are essential factors in determining the magnitude of the effect of the water level on the survival of whitefish eggs. For example, in Lake Päijänne, a 40 cm drop in water level during winter has been estimated to typically cause 30% (20–40%) mortality of eggs of the littoral zone spawning European whitefish (Valkeajärvi et al., 2001). Additionally, low water levels in late April due to regulation, and wave action after early ice-off have been found to negatively impact European whitefish recruitment (Linløkken and Sandlund, 2016). Approximately one-third of the total area of all Finnish lakes (more than 300 lakes), is water level regulated or covered by various water level regulatory projects (waterinfo.fi4.12. 2023). Water level regulation requires a permit, and the permit holder must monitor the effects of regulation on fish stocks and biota in general (Water Act 587/2011). Our study is part of the long-term monitoring stipulated by the water level regulation permit for Lake Päijänne and Lake Konnivesi–Ruotsalainen in Central Finland (ISY, 2002;VHO, 2004;KHO, 2006; Väänänen et al., 2022). Coregonine fish are the most important target species for inland fishing in Finland. In Tehinselkä basin in Lake Päijänne, there are three different morphs of European whitefish: a large sparsely rakered morph (LSR), a small sparsely rakered morph (SSR), and a densely rakered (DR) morph (Valkeajärvi, 1987; Valkeajärvi et al., 2012;Kahilainen and Østbye, 2006). A negative correlation between vendace (Coregonus albula (L.)) and whitefish stock indices has been observed in Lake Päijänne (Valkeajärvi et al., 2012). Competition with vendace is known to negatively affect the growth rate of European whitefish (Raitaniemi et al., 1999;Valkeajärvi et al., 2012), potentially increasing predation mortality and decreasing population fecundity among whitefish. Vendace has the potential to eat whitefish eggs (Berezina et al., 2024) and newly hatched coregonid larvae (Urpanen et al., 2012). This suggests that vendace may, to some extent, affect the production of whitefish eggs and larvae. The negative effect on growth can be caused by food competition between species with similar zooplankton diets (Svärdson, 1979;Sandlund et al., 2013). Dense vendace stocks may decrease the amount of zooplankton in the European whitefish diet and force plankton feeder whitefish to change their diet and eat more benthic food (Svärdson, 1979; Salojärvi, 1992;Pulkkinen et al., 1999). Usually, the diet of European whitefish is associated with habitat, but they can be flexible in their choice of food (Zubova et al., 2023). For example, four different European whitefish morphs in Lake Kuetsjarvi show low specialization of food and use the most abundantresourcesinthelake(Zubovaetal.,2023).Oneofthose is even more versatile than other whitefish morphs and effectively uses both macrozoobenthos, zooplankton, and fish. However, European whitefish morphs with similar trophic position, morphology, diet, and habitat use as vendace are more sensitive to competition by vendace (Sandlund et al., 2013). This study aimed to investigate the effect of the actual water level and its change during the period from spawning in late autumn to hatching in spring on the density of newly hatched European whitefish larvae in three regulated lakes and one unregulated lake. More detailed data for the Tehinselkä basin in regulated Lake Päijänne enabled the inclusion of the effects of the abundance indices of the spawning population of the whitefish and the possible competing species, vendace, on the whitefish larval density in addition to the variables related to the water level. We hypothesized that the European whitefish larval density is associated with 1a) the annual wintertime water level change (level drop being negative change), 1b) the water level in spawning time in autumn, or 1c) the minimum water level during the egg incubation period. If the water level is low in autumn there could be less suitable spawning habitat available. The low water level in spring and especially the large drop in water level during the egg incubation period could reduce egg survival as they get exposed to ice compression. We also hypothesized that 2) lakes experiencing a large negative wintertime water level change on average also exhibit lower larval densities on average. This could be attributed to the chronic negative impact of regulation on the average level of survival of eggs from spawning to hatching, consequently influencing average spawning population size and average annual egg production. Our third hypothesis concerns the more intensively studied Tehinselkä basin in Lake Päijänne, where we assume that 3a) European whitefish spawning stock abundance, 3b) vendace abundance, and/or 3c) water level in autumn or spring, or wintertime water level change affects whitefish egg survival and larval density. In addition to the direct positive effect, the effect of European whitefish spawning stock abundance on larval density may include compensatory density-dependent mortality. The vendace population may have a negative or positive effect on whitefish egg survival and larval density. The hypothesized negative effect is the deterioration of the quality of gametes due to food competition between vendace and the European whitefish spawning population in summer/autumn. The hypothesized positive effect is that a higher abundance of the vendace spawning population results in a higher density of coregonids (vendace þwhitefish) eggs in autumn, reducing predation mortality of European whitefish eggs and newly hatched larvae and leading to higher larval density in spring. 2 Materials and method 2.1 Study lakes We studied four oligotrophic boreal lakes in the Kymijoki drainage system in Central Finland. Lake Päijänne (study area Tehinselkä basin, a 25 350 ha central part of a 108 201 ha lake, Valkeajärvi et al., 2012) and Lake Ruotsalainen (study area 4 650 ha central part of a 7 407 ha lake, Väänänen et al., 2022), Page 2 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 are the most strongly regulated lakes. Lake Puula (study area a ca. 20 000 ha central part of a 32 911 ha lake, Marjomäki and Huolila, 2001), is moderately regulated, and Lake Southern Konnevesi (12 105 ha) is an unregulated lake (Fig. 1,Tab. 1). According to the Water Framework Directive (2000/60/EC) of the European Union, these lakes are classified as large lowhumus lakes (HERTTA, open database of Finnish Environment Institute). 2.2 Water level variables To determine the intensity of regulation and natural fluctuations of water level during the whitefish egg incubation period, lake-specific daily water level data (in centimeters above sea level, using the NN Geodetic height system in Finland, with the zero point of the water level scale situated at Katajanokka, Helsinki) were obtained from the open data service (HERTTA) of the Finnish Environment Institute. The annual wintertime water level change Dh was calculated using equation (1)(Valkeajärvi et al., 2001): Dh¼hmyhNov15 y1ð1Þ where hm y = minimum water level before the ice-off in year y (year of hatching of whitefish larvae) and h Nov15 y1 = water level on November 15 in year y1 (previous autumn, spawning year). November 15 is assumed to be the typical spawning time of European whitefish (Valkeajärvi et al., 2001). Water levels (altitude system NN, cm above the sea level) in autumn and spring varied considerably between the study years. The range of water levels in autumn (h Nov15 y1 ) was in Tehinselkä 76 cm (min. 7780 cm in 2018 and max. 7856 cm in 2012), in Ruotsalainen 26 cm (min. 7736 cm in 2013 and max. 7762 cm in 2008), in Puula 44 cm (min. 9435 cm in 2002 and max. 9479 cm in 2011) and S. Konnevesi 63 cm (min. 9481 cm in 2006 and max. 9544 cm in 2012). The range of minimum water levels in spring (hm y ) was in Tehinselkä 111 cm (min. 7750 cm in 2003 and max. 7861 cm in 2020), in Ruotsalainen 59 cm (min. 7684 cm in 2011 and max. 7743 cm in 2018), in Puula 58 cm (min. in 9421 cm in 2011 and max. 9479 cm in 2020) and S. Konnevesi 56 cm (min. 9479 cm in 2003 and max. 9535 cm in 2020). Synchrony in inter-annual variation in h Nov15 y1 and hm y was observed between all lake pairs (Pearson correlation test, all p<0.028, Supplementary material Tab. S1). Both positive and negative wintertime water level changes (Dh) were observed during the study years in both regulated study lakes Tehinselkä and Puula and the unregulated S. Konnevesi. However, in regulated Ruotsalainen, Dh values were consistently negative in all study years (Supplementary material Fig. S1). The average Dh for all study years was the smallest, indicating the largest average drop in water level during winter, in the regulated Ruotsalainen (average 27 cm, min. 55 cm in 2012–2013, max. 3 cm in 2007–2008) and Tehinselkä (average 13 cm, min. 68 cm in 2012–2013, max. þ59 cm in 2019–2020), while moderately regulated Puula (average 5 cm, min. 41 cm in 2012–2013, max. þ31 cm in 2019–2020) and unregulated S. Konnevesi (average þ3 cm, min. 35 cm in 2012–2013, max. þ39 cm in 2019– 2020) had larger average of Dh indicating a smaller average drop in water level during winter (Supplementary material Fig. S1). Synchrony in inter-annual variation in Dh was observed between all lake pairs (Pearson correlation test, all p<0.009, Supplementary material Tab. S2). 2.3 Fieldwork and estimation of larval densities Newly hatched whitefish larvae were collected on one sampling occasion annually, typically 13 weeks after the iceoff in 2000–2022 (Ruotsalainen in 20082022) (Supplementary material Tab. S3) using depth-zone stratified bongo and tube net sampling. Identical sampling method, equipment, and stratified sampling design were used in all lakes (Karjalainen et al., 1998;Urpanen et al., 2009). Egg incubation experiments and otolith analyses in the laboratory and field observations from multiple lakes in Finland show that the hatching of the majority of coregonid larvae is triggered by the warming of the water from 12°Cto46°C(Urpanen et al., 2005;Urpanen, 2011;Karjalainen et al., 2015). Hatching occurs during the short period before and after (± 2 weeks) ice-off and the maximum density of larvae is observed about 2 weeks after the ice-off (Urpanen et al., 2005;Urpanen, 2011). The sampling plots (30 per lake) were randomly selected yearly from the study area of each lake before the year 2010 and kept fixed since 2010 (Lake Ruotsalainen since 2008). Of these, 20 were in the littoral zone and 10 in the pelagic zone. A total of 9 samples were collected from each littoral sampling plot, allocated to different depth strata (0–0.5, 0.5–1, 1–2, and Fig. 1. Locations of study lakes in Central Finland (National land survey of Finland, open topographic data). Page 3 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 2–10 m depending on the lake) and vertical sampling depths (0–0.3 m and 0.3–0.6 m) in the strata with bottom depth ≥1m. Two samples were collected from each pelagic plot (vertical sampling depths 0–0.3 m and 0.3–0.6 m). The annual total sample number was 200 samples per lake. In S. Konnevesi every two years, the pelagic zone sampling depths were 0–0.6 m and 0.7–1.3 m (Supplementary material Tab. S4). Samples from the shallowest littoral stratum (0–0.5 m bottom depth) were collected by wading with a pushed tube net. If the shore was so steep that wading was not possible, the samples were collected using a tube net kept on the side of the boat. In some cases of no sampling in 0–0.5 m due to a very steep depth profile, the density of the shallowest stratum was substituted by the density in the 0.5–1 m stratum of the same sampling plot. All other samples were collected using a bongo net attached in front of a motorboat (length of boats 5 and 9 m). The mesh size of the bongo and tube net was 500 mm. The volume of each sample was determined by a flowmeter attached to the mouth of the net. The typical volume of the tube net sample was approximately 1 m 3 , while that of the bongo net sample was approximately 100 m 3 in the littoral zone and 500 m 3 in the pelagic zone The catches were preserved in the field in an ethanol–formalin solution (Karjalainen, 1992). The species of Coregoninae fish larvae (vendace or European whitefish) was identified based on the number of myomeres, as described by Karjalainen et al. (1992), and the number of larvae in each sample was counted in the laboratory. Using the number of larvae and the sample volume, the larval densities (individuals 100 m 3 ) were calculated for each sampling stratum of the sampling plots (Urpanen et al., 2009). The density outside the sampled vertical depth zone, >0.6 m deep up to 2 m and some cases in the pelagic samples from S. Konnevesi 1.32 m), was estimated based on the larval density of samples taken from 00.3 m and 0.30.6 m depth (and some cases in the pelagic samples from S. Konnevesi 0–0.6 and 0.7–1.3 m), assuming that the rate of density reduction is constant concerning depth (Urpanen et al., 2009). It was assumed that no larvae occupy a depth deeper than 2 m from the surface. The total number of larvae for the study area in each lake was calculated by multiplying the average density in each depth stratum by the volume of the stratum (Supplementary material Tab. S5). Then, the average larval density (individuals ha 1 ) in the study area was calculated by dividing the total number of larvae by the study area (Tab. 1). The 95% confidence intervals of the annual average densities in each lake were calculated by bootstrapping, involving the extraction of 5000 new random samples, with replacement, of 20 littoral and 10 pelagic plots from the annual plots. Then, plot-specific sets of observed densities were used to estimate the average larval density as described above. The 2.5% and 97.5% percentiles of these 5000 bootstrap estimates were used as the lower and upper limits of the 95% confidence interval (Karjalainen et al., 2021). During the resampling process, uncertainty regarding depth zone area data was also simulated by drawing them from statistical density functions describing our uncertainty about the true areas. Therefore, the 95% confidence intervals account for the uncertainty stemming from both the spatial variability of fish density and the uncertainty in hypsographic data. 2.4 Association between water level variables and larval density and variation in larval densities between lakes Because fish density typically does not follow a normal distribution, but more likely e.g., Poisson distribution, all average larval density (D) estimates were ln(D þ1)-transformed before conducting variance, correlation, or regression analyses. In Lake Puula in 2019, the larval density estimate was zero. Therefore, the number 1 was added to the average densities for all years and lakes, enabling a logarithmic transformation and comparability of the data. The association between annual average larval density and water level variables (Dh, h Nov15 y1 or hm y ) was analyzed using Spearman‘s rank correlation coefficient with one-tailed significance test H1: positive association. The differences between the average of the time series of larval densities both between lakes and between years were tested using analysis of variance (2 ANOVA and Tukey post hoc test) with two-tailed H1: there are differences in the averages of larval densities. In addition, the temporal variation in larval densities was studied using linear trend analysis. A decreasing linear trend in Table 1. Basic information and water quality in (A) March 2023 and (B) minimum and maximum values for the whole study period in the study lakes (Sampling depth 1 m) The water quality values used for Tehinselkä (Lake Päijänne) were from the nearest deep (over 50 m) sampling point (Central-Päijänne Judinsalonselkä) located 10 km north of Tehinselkä. Information from the Finnish Environment Institute open data service HERTTA 15.9.2023. Tehinselkä Ruotsalainen S. Konnevesi Puula Location N61°29’N61°14' N 62°36' N 61°47' E25°26' E 25°57' E 26°34' E 26°41' Lake area (ha) 108 201 7 407 12 105 32 911 Study area (ha) 25 350 4 650 12 105 20 000 Maximum depth (m) 95 56 57 69 Mean depth (m) 14.2 9.9 12.5 9.2 A) Total phosphorus (mgL 1 )6 6 7 3 B) Total phosphorus (mgL 1 )414 212 314 38 A) Colour (mg L 1 Pt) 38 25 20 27 B) Colour (mg L 1 Pt) 2045 2040 1040 1043 Page 4 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 annual larval densities could potentially indicate the negative long-term effect of regulation. The synchronicity of the interannual variation in larval density between lakes was analyzed using the Pearson correlation test. In this context, the log-transformed time series were first standardized (mean = 0, standard deviation = 1). Then, to eliminate the effect of the trend, time series were made stationary by linear regression detrending, and Pearson correlation analysis was subsequently performed on the stationarized data (regression residuals). 2.5 Combined effects of spawning stock biomass, vendace abundance, and water level variables on larval density in Tehinselkä More extensive data was available for Tehinselkä, which enabled more detailed analyses of the impact of water level variables on whitefish larval density. In Tehinselkä, the whitefish SSR morph is known to spawn in the lake while the DR morph spawns in streams and it is mainly maintained by stockings. The trawl catch per unit effort (CPUE) and catch sample data suggest a very sparse population of LSR morph in Tehinselkä (Valkeajärvi et al., 2012;Puranen and Ranta, 2017). Based on the information presented above, we assumed that the larvae in our study in Tehinselkä stem from the spawning stock of SSR morph. The majority of females (89%) of SSR morph reach sexual maturity at the age of four growing seasons ( 3þyears) (Valkeajärvi et al., 2012). The proportion of mature SSR morph in the CPUE was estimated based on trawl catch samples (Valkeajärvi et al., 2012, and personal/written communication; Puranen and Ranta, 2017, and unpublished material by Häme Fisheries Centre). The index of lake-spawning SSR whitefish spawning stock biomass (SB y1 in formula 2) was the CPUE (kg h 1 ) of mature (3þyears and older) SSR morph in a commercial trawl in August. Additionally, the index of vendace stock abundance was determined using the CPUE (kg h 1 ) of age 1þyears and older vendace, also obtained from the same data. The vendace population can have a negative or positive impact on the larval density of whitefish in year y. The factor with the hypothesized negative impact in the model is the vendace abundance index (ages 1þyears and older) in year y due to potential food competition during the growing season of the previous year y1 (then ages 0þand older) or predation on whitefish eggs (Berezina et al., 2024) in winter and newly hatched larvae (Urpanen et al., 2012) in spring y. The factor with the hypothesized positive impact in the model is the vendace abundance index (ages 1þyears and older) in year y1, assuming that a high abundance of the vendace spawning population can result in a high number of eggs in year y–1, potentially reducing predation mortality of whitefish eggs and hatching larvae in spring y and leading to high survival rate from spawning to hatching. The whitefish larval density was modeled as functions of the whitefish spawning population biomass index, three water level variables (Dh, h Nov15 y1 ,hm y ), and two vendace population density indices (years y–1 and y) to study the association between the larval density and the abovementioned factors. The water level variables were not included in any model together because of their correlation (multicollinearity). The same applies to vendace CPUE indices for years y and y–1. In all model fitting, least squares regression was used (see detailed description below). If the survival and mortality of whitefish eggs are not dependent on egg density determined by spawning biomass, then the larval density (D y ) in year y is directly proportional to the whitefish spawning stock (SB y1 ) index: Dy¼aSBy1exp eðÞ;ð2Þ lnDy¼lnaþlnSBy1þe;ð3aÞ where a= constant (density-independent) index of egg survival from spawning to hatching (a=D y /SB y1 ), e= random error, e∼N (0, s 2 ), y= year of hatching of larvae and SB y1 = SSR whitefish spawning stock biomass index (CPUE) in year y1. The index SB y1 is directly proportional to the number of eggs produced by the population (population fecundity) assuming that the sex ratio of the spawners and the fecundity relative to fish weight (eggs g 1 ) are constant from year to year. The survival of eggs can also depend compensatorily on density e.g., following the model of Ricker (1954) Dy¼aSBy1exp bSBy1  exp eðÞ;ð4Þ lnDy¼lnaþlnSBy1bSB y1þe;ð5aÞ where b= survival density dependence term. Environmental factors (X) were included in the equations (3a) and (5a) lnDy¼fSB y1  þwtfX t ðÞþ...;ð3b;5bÞ where f(X t ) is a standardization function fX t ðÞ¼Xtmean of Xt ðÞ=standard deviation of Xt ðÞð6Þ The suitability of the nested models (3a and 5a and all combinations of 3b and 5b) was compared using the F-test. Models are nested when a simpler model is a special case of a more detailed model, e.g., Model 3a is included in Model 5a,as a special case of Model 5with b= 0. In addition, Akaike (1973) Information Criterion corrected for small sample sizes (AICc) and Akaike weight (w i = the probability that the model is the best among the whole set of considered models) were calculated (equations for least squares regression models, e.g.,inGlatting et al., 2007). For every model, it was visually checked whether the model residuals (observation estimate of dependent variable) implied deviation from normal distribution, heteroskedasticity, or nonlinearity of the effect of the independent variables. No such implications were found. Statistical tests were conducted using Excel (version 2307, Microsoft Co., Redmond, WA, USA) and SPSS software (version 22.0, IBM Co., Chicago, IL, USA). The threshold for statistical significance was set at 0.05. Based on the Precautionary principle approach of the FAO Code of Conduct of Responsible Fisheries, we tried to keep the test power as Page 5 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 high as possible, and for that reason 1) one-way hypotheses were used based on prior knowledge of the factors potentially reducing whitefish egg survival, and 2) Bonferroni correction was not applied for p-values. However, when interpreting the results, it should be noted that this increases the risk of rejection of true H 0 (false positive, type I error). 3 Results 3.1 Variation of larval density in study lakes and comparison of larval densities between lakes and years The inter-annual variation in the larval density estimates (non-transformed data) was high in all lakes, with coefficients of variation ranging from 89% to 159% (Fig. 2 and Tab. 2). The highest annual density estimate during the study period was observed in Tehinselkä in 2017 (178 ind. ha 1 ) and the lowest in Puula in 2019, where no whitefish larvae were caught (Fig. 2). No significant linear temporal trend was observed in the log-transformed annual larval density of Tehinselkä, S. Konnevesi, and Ruotsalainen in the study years. In Puula, there was a significantly decreasing trend in whitefish larval density over time (Tab. 2 and Supplementary material Fig. S2). Significant synchrony in inter-annual variation in larval densities (indicated by a positive correlation between time series) was observed in the standardized and log(Dþ1)- transformed and stationarized larval densities between the pairs Tehinselkä–S. Konnevesi and Tehinselkä–Ruotsalainen (Tab. 3 and Fig. 2). For the entire study period, there were statistically significant differences in averages of whitefish larval density both between lakes (F= 11.655, df = 3, p<0.001) and between years (F= 2.169, df = 22, p= 0.010). Based on the Tukey test the average larval density for the entire study period was highest and at a similar level in unregulated S. Konnevesi (34 ind. ha 1 ) and regulated Tehinselkä (23 ind. ha 1 ), while in regulated Ruotsalainen and moderately regulated Puula, the average larval density was lower and at a similar level, <10 ind. ha 1 (Fig. 2,Tabs. 2 and 4). Temporally, the average larval densities (all lakes) in the time series were lowest in the year 2022, differing significantly from the years 2001, 2002, 2011, 2015, and 2017 (Fig. 2, Supplementary material Tab. S6). 3.2 Association between water level variables and larval density There was no significant association between larval density and wintertime water level change (Dh, Fig. 3), the water level in autumn (h Nov15 y1 ,Fig. 4) or water level in spring (hm y , Fig. 5) in any of the study lakes (all p>0.05, 1-tailed H 0 : positive Spearman correlation between larval density and the water level variable). Only in Lake Ruotsalainen, was the association between larval density and water level in autumn indicative significant, p= 0.05 (Fig. 4). 3.3 Effects of water level variables, spawning stock biomass of whitefish, and vendace abundance on larval density in Tehinselkä Based on the models (3a and 5a), the survival of whitefish eggs in Tehinselkä was not compensatorily dependent on the density of the spawning stock i.e., the larval density was directly proportional to the spawning stock density. The value of parameter bdid not differ significantly from zero (Supplementary material Tab. S7) and the goodness of fit of model 5a was not significantly higher than that for model Fig. 2. Average newly hatched whitefish larvae density in spring (y-axis logarithmic, vertical line = 95% confidence interval) and sliding mean (3 years) in A) unregulated Southern Konnevesi, B) moderately regulated Puula, C) regulated Tehinselkä (years 2000–2022) and D) regulated Ruotsalainen (years 2008–2022). Page 6 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 3a (F= 0.048, df1 = 1, df2 = 21, p=0.829). Based on model 3b, neither the water level variable parameters f1: Dh, h Nov15 y1 or hm y (Supplementary material Tabs. S8 and S9), nor the vendace population (age 1þand older) abundance parameters f2 (year y or year y1) (Supplementary material Tabs. S8 and S9) had a significant effect on the density of whitefish larvae (all p-values >0.351, all F-values <0.43 in comparison to model (3a). The AICc-statistics support the same conclusion; the simple one parameter (a= constant density-independent survival index) model (3a)wasmost probably (w i = 29.7%) the best of all models compared (Supplementary material Tab. S10). Thus, the most parsimonious model was ln larval densityðÞ¼ln aðÞþln whitefish spawning stock indexðÞ; where a¼2:23;95% conf:interval ¼1:093:37 ðassuming a>0Þand r2¼0:25: 4 Discussion No support was found for the hypotheses 1) that the whitefish larval density is positively associated with the annual wintertime water level change (Dh) or water level in autumn (h Nov15 y1 ) or water level in spring (hm y ), or that 2) lakes with a large average negative Dh exhibit lower larval densities on average. The observed average larval densities were the highest and at a similar level in both the regulated Tehinselkä with considerable average negative Dh and the unregulated Konnevesi with considerable average positive Dh. The lowest Table 3. Pearson correlation coefficients (r), significance levels (p) for one-tailed H1: positive correlation and number of observations (n) for the standardized ln(Dþ1)-transformed larval density time-series between the study lakes shown on grey background and those for stationarized (regression residuals) time-series on white background. Statistically significant correlation coefficients are bolded. S. Konnevesi Puula Tehinselkä Ruotsalainen S. Konnevesi r 0.027 0.482 0.033 p 0.451 0.010 0.454 n232315 Puula r 0.029 0.281 0.201 p 0.448 0.097 0.236 n23 23 15 Tehinselkä r 0.493 0.205 0.536 p 0.008 0.174 0.020 n23 23 15 Ruotsalainen r −0.168 0.204 0.441 p 0.725 0.233 0.050 n15 15 15 Table 2. Average larval density (individuals ha -1 , non-transformed data), standard deviation = s.d., median = Md and coefficient of variation = CV (standard deviation / average * 100%) and significance of linear trend (ln(D þ1)-transformed data) (p-value) of the time series in the study areas. Study area Period Average s.d. Md CV Trend p Tehinselkä 2000–2022 23 36 11 159% 0.318 Ruotsalainen 2008–2022 8 8 4 90% 0.225 S. Konnevesi 2000–2022 34 40 25 118% 0.985 Puula 2000–2022 6 6 4 89% 0.019 Table 4. Mean differences in the ln(D þ1)-transformed larval density time series between the study lakes. 2 ANOVA, multiple comparison: Tukey HSD. Mean difference risk levels (p) according to the two-tailed hypothesis (H1: average larval densities are different) and 95% confidence interval (CI). Statistically significant mean differences are in bold. Mean difference pCI S. Konnevesi Puula 1.40 <.001 0.73–2.08 Tehinselkä 0.41 0.375 0.26–1.09 Ruotsalainen 1.09 0.002 0.33–1.85 Tehinselkä Puula 0.99 0.002 0.31–1.66 Ruotsalainen 0.68 0.096 0.08–1.44 Ruotsalainen Puula 0.31 0.702 0.45–1.07 Page 7 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23 larval densities (all lakes) in the time series were in 2022. However, in the winter of 2021–2022, the Dh was not abnormally negative (large drop) but at the level of the average of the time series. There was no significant negative trend in the annual larval density of regulated Tehinselkä and Ruotsalainen in the study years. This result also did not support our hypothesis 2) that regulation causes a negative long-term effect on the reproductive efficiency of the spawning stock and a chronic negative impact on larval production. These results align with the conclusions of a previous study (Valkeajärvi et al., 2001), which suggested that egg losses due to water level changes in winter do not significantly regulate the abundance of European whitefish year classes in Tehinselkä. Also, Marttunen (1992) has estimated that the effects of regulation on the carrying capacity of the reproduction areas have a more important role on the European whitefish stocks than regulation-induced changes in egg mortality. However, in moderately regulated Puula, a decreasing trend in European whitefish larval density was observed over the time series. The reason for the decline in Puula remains unknown. No increase in trophy or potential egg predator populations has been reported in Puula. Cunningham and Dunlop (2023) observed significant declines in larval densities of lake whitefish (Coregonus clupeaformis) in Lake Huron between the historical (1977–1986) and contemporary time periods (2017–2019, 2021). They found that Dreissenid mussels have a negative association with larval density and water level in November was positively associated with higher larval densities, but the reduced water levels were not a primary factor in the substantial declines in larval densities (Cunningham and Dunlop 2023). Our results for Lake Ruotsalainen implied an indicatively positive association between the water level in November and larval density in the following spring, as Cunningham and Dunlop (2023). The positive association we found may very well be spurious but, according to the precautionary principle, warrants future monitoring. The lakes in our study are close to each other and therefore exposed to very similar variations in climatic factors that affect water level which causes strong synchrony in the variability of water level variables between all lakes. There was significant synchrony in stationarized and standardized annual larval densities only between Tehinselkä (regulated) and Konnevesi (non-regulated) in 2000–2022 and between Tehinselkä and Ruotsalainen (the largest average drop in water level) in 2008– 2022. However, as the water level variables were not associated with larval density, variations in larval density are more likely driven by some other factors. Synchronicity in the variability of populations on a scale of 100–200 km has been found in vendace (C. albula,Marjomäki et al., 2004) but it is more likely related to the synchronicity in the inter-annual variability of the regionally correlated exogenous environFig. 3. Association between wintertime water level change, Dh (minimum water level in spring water level 15.11.) and density of newly hatched whitefish larvae in spring (y-axis logarithmic) in A) unregulated Southern Konnevesi, B) moderately regulated Puula, C) regulated Tehinselkä (years 2000–2022) and D) regulated Ruotsalainen (years 2008–2022). Spearman correlation coefficients (rho) and significance level (p) according to the one-tailed hypothesis (H1: positive correlation). Negative Dh = a decrease in water level. Page 8 of 14 T. Väänänen et al.: Int. J. Lim. 2024, 60, 23