Determinants of Trophic Cascade Strength in Freshwater Ecosystems: a Global Analysis

Home , Cascade effect (ecology), Freshwater ecosystem

Article type : Articles

Journal：Ecology Manuscript type: Articles

Running head: Trophic cascade strength

Determinants of trophic cascade strength in freshwater ecosystems: a global analysis

Haojie Su1,2, Yuhao Feng2, Jianfeng Chen3, Jun Chen1, Suhui Ma2, Jingyun Fang2, Ping Xie1,4,5

1 Donghu Experimental Station of Lake Ecosystems, State Key Laboratory of Freshwater Ecology and Biotechnology, Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan 430072, China

2 Institute of Ecology, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process, which may lead to differences between this version and the Version of Record. Please cite this article as doi:

4 Institute for Ecological Research and Pollution Control of Plateau Lakes, School of Ecology and Environmental Science, Yunnan University, Kunming, 650091 China

5 Corresponding Author. Mailing address: Institute of Hydrobiology, Chinese Academy of Sciences, 7# Donghu South Road, Wuhan 430072, China. E-mail: [email protected]

Manuscript received 2 October 2020; revised 24 January 2021; accepted 22 February 2021. Accepted Article

Top-down cascade effects are among the most important mechanisms underlying community structure and abundance dynamics in aquatic and terrestrial ecosystems worldwide. A current challenge is understanding the factors controlling trophic cascade strength under global environmental changes. Here, we synthesized 161 global sites to analyze how multiple factors influence consumer-resource interactions with fish in freshwater ecosystems. Fish have a profound negative effect on zooplankton and water clarity but positive effects on primary producers and water nutrients. Furthermore, fish trophic levels can modify the strength of trophic cascades, but an even number of food chain length does not have a negative effect on primary producers in real ecosystems. Eutrophication, warming and predator abundance strengthen the trophic cascade effects on phytoplankton, suggesting that top-down control will be increasingly important under future global environmental changes. We found no influence or even an increasing trophic cascade strength (e.g., phytoplankton) with increasing latitude, which does not support the widespread view that the trophic cascade strength increases closer to the equator. With increasing temporal and spatial scales, the experimental duration has an accumulative effect, while the experimental size is not associated with the trophic cascade strength. Taken together, eutrophication, warming, temporal scale and predator trophic level and abundance are pivotal to understanding the impacts of multiple environmental factors on the trophic cascade strength. Future studies should stress the possible synergistic effect of multiple factors on the food web structure and dynamics.

Keywords

Trophic cascade, meta-analysis, global change, trophic level, eutrophication, warming Accepted Article

This article is protected by copyright. All rights reserved Introduction Trophic cascades, the effects of predators that propagate downward through food webs across multiple trophic levels, are crucially important for us to understand the structure and dynamics of ecosystems (Ripple et al. 2016). In freshwater habitats, manipulations of piscivorous fish have been used in lake restoration to improve water quality, as predation by fish can result in strong cascading interactions in lower trophic levels and environments (Hansson et al. 1998). Recent studies have shown that such trophic interactions can be modified by anthropogenic activities, e.g., overharvesting, species invasion, pollution and habitat loss, which ultimately alter the functioning and service provisioning of ecosystems (Estes et al. 2011, Jackson et al. 2001). For instance, overfishing can alter species composition, skew food webs and even trigger ecosystem shifts to an alternative state under extreme conditions (Daskalov et al. 2007, Jackson et al. 2001, Scheffer et al. 2005). In addition, the strength of trophic cascades is also affected by ecosystem type, productivity, initial perturbation magnitude, environmental conditions, biological characteristics (e.g., body size and metabolic rate) and response variables (Borer et al. 2005, Jia et al. 2018, Pace et al. 1999, Strong 1992). Furthermore, the dynamics of food webs depend on not only the strength of top-down forcing but also the bottom-up effects of nutrient resources. It has been suggested that fish can accelerate nutrient cycling by nutrient excretion (Attayde and Hansson 2001, Vanni and Layne 1997) and resuspension of sediment (Scheffer et al. 2003). Consequently, clarifying the net cascading effects of predators and the associated determinant factors in the context of global human pressures is practically important for ecosystem management.

Ecological theory has suggested that trophic cascades on primary producers depend on the food chain length, with odd numbers of trophic levels increasing standing crops of plants, while even numbers decrease the abundance of producers due to the flourishing primary consumers (Fretwell 1977, Ripple et al. 2016). For instance, in phytoplankton-zooplankton-zooplanktivorous fish-piscivorous fish food chains, predation on zooplanktivorous fish releases the predation pressure Accepted Article

Hansson et al. 1998). However, in real-world ecosystems, as the food diet of fish in nature is complex, fish rarely have exactly even or odd trophic levels, which may obscure the top-down responses (Attayde et al. 2010). Some studies suggest that omnivory may weaken the strength of trophic cascades due to the greater complexity of food webs (Fagan 1997, Polis and Strong 1996). Evidence also shows that omnivory benefits primary producers and exerts strong cascades because of preferential feeding on herbivores rather than on plants (Ho and Pennings 2008). Nevertheless, in freshwater ecosystems, omnivorous filter-feeding fish can consume primary producers directly and thus have been used to control algal blooms and improve water quality in some subtropical eutrophic waters (Zhang et al. 2008). Therefore, our current knowledge of how the strength of trophic cascades varies with trophic levels (e.g., herbivorous, zooplanktivorous and piscivorous) in natural freshwater ecosystems remains incomplete.

Another major challenge in understanding top-down control is determining whether trophic cascade effects will be amplified or weakened by anthropogenic pressures, e.g., eutrophication and climate warming (Piovia-Scott et al. 2017, Rosenblatt and Schmitz 2016). It is well known that eutrophication-induced high resource availability and quality can support more consumers, which will increase the cascades by promoting the consumption rates of herbivores (Leibold 1989). A meta-analysis of coastal wetlands showed that eutrophication enhanced plant-herbivore interactions by increasing the foliar nitrogen concentration (He and Silliman 2015). However, eutrophication can also affect the community structure and dynamics of primary producers (Su et al., 2019), which may alter the food composition of consumers and consequently decrease the strength of trophic cascade effects. Nutrient enrichment can decouple predator-prey interactions by reducing food web efficiency (Davis et al. 2010), e.g., facilitating predator-resistant algae with toxicity and low edibility (Heisler et al. 2008). In brief, the strength of trophic cascades may differ among ecosystems depending on their background nutrient levels. Accepted Article

This article is protected by copyright. All rights reserved Warming may also alter the sensitivity of food webs to trophic cascade effects. As the respiration rate of heterotrophs is more sensitive than the photosynthetic rate of autotrophs to temperature (Allen et al. 2005), the greater metabolic rate of fish under warmer conditions may result in higher feeding rates and stronger top-down effects on lower trophic levels. Alternatively, increasing temperature may negatively affect organismal fitness and limit the ingestion rate (Deutsch et al. 2008), which could in turn weaken the strength of the trophic cascade. In addition, warming, as well as fish predation, often decreases the body size structure of herbivorous zooplankton (Brooks and Dodson 1965, Forster et al. 2012, Daufresne et al. 2009, Yvon-Durocher et al. 2011), which may lead to weaker top-down effects on primary producers (Delong et al. 2015). It has been shown that fish manipulation in (sub)tropical lakes is less efficient than in temperate lakes due to the small grazers in the former (Jeppesen et al. 2005). Moreover, warming tends to facilitate cyanobacterial blooms, which reduce the food quality of zooplankton and thereby reduce top-down effects (Paerl and Huisman 2008). Warming may also facilitate bottom-up forcing by promoting eutrophication, e.g., elevated nutrient loading due to extreme weather, increased nutrient mineralization, evaporation and nutrient release from sediments (Gudasz et al. 2010, Jeppesen et al. 2009). In addition, elevated temperature may cause fish to eat more carbon-rich resources and excrete excess nutrients to maintain increased respiration (Hessen and Anderson 2008). Fish at different trophic positions may vary in their sensitivity to temperature (Vasseur and McCann 2005), which leads to unbalanced cascading effects on lower trophic levels. A previous experimental study suggested that phytoplankton, and thereby algal blooms, will benefit from climate warming in aquatic ecosystems with odd numbers of trophic levels but not in those with even numbers of trophic levels (Hansson et al. 2013). Therefore, temperature-dependent nutrient availability and consumer-resource species interactions can affect the algal biomass.

Fish stocking in natural freshwater ecosystems is a widespread and long-term human activity, which lead to alterations in ecosystem functions and services globally (Eby et al. 2006, Hammerschlag et al. 2019, Su et al. 2020). In this study, based on a global meta-analysis of 2202 Accepted Article

This article is protected by copyright. All rights reserved individual observations from 161 different sites (Fig. 1, also see Methods), we aimed to clarify the net trophic cascade effects of fish stocking on lower trophic levels and water quality and to quantify how the trophic level, eutrophication and warming affect the strength of trophic cascades. In addition, we evaluated whether the strength of top-down cascade effects varies with latitude, as current knowledge on the role of latitudinal gradients in determining the strength of top-down effects is inconclusive (i.e., intensification (Pennings and Silliman 2005, Baskett and Schemske 2018, Schemske et al., 2009), weakening (He and Silliman 2015) and no change (Andrew and Hughes 2005, Moles et al. 2011, Poore et al. 2012) with decreasing latitude have been reported). We also tested whether predator abundance (fish density), experimental size (volume) and experimental temporal scale (experimental duration) are associated with top-down cascade effects.

Materials and Methods

Data source and collection

Peer-reviewed journal articles related to fish manipulations were searched using the Web of Science and China National Knowledge Infrastructure (CNKI, http://www.cnki.net) search engines (1900-2020). The keywords used for the literature search were fish and exp* and (zooplankton or phytoplankton or alga*). Furthermore, previously published reviews and meta-analyses of fish manipulations in freshwater ecosystems were also selected for additional study (Bell et al. 2003, Brett and Goldman 1996, Zhang et al. 2008). We then screened the studies based on the following criteria: (i) at least one of the response variables TN, TP, water clarity (represented by Secchi depth, SD), chlorophyll a (Chl a), total phytoplankton (biomass or abundance, the same below), total zooplankton, cladocerans, copepods, and rotifers had to be reported; (ii) both fish and fishless treatments had to be started simultaneously, with fish manipulations for the purpose of restoration in lakes excluded; and (iii) experiments with only freshwater fish were included. Accepted Article

This article is protected by copyright. All rights reserved According to these criteria, we compiled 2202 individual observations at 161 different sites from 206 published papers (for detailed references please see Appendix S1). These study sites were mainly distributed in North America, Europe and East Asia (Fig. 1). For each individual observation, the location of the experiment, volume of the experiment (m3), species of experimental fish, duration of the experiment (day), fish stocking density (g m-3) and water temperature (℃) were also extracted. If information on fish density was given in individual numbers and body length, we calculated the body mass based on the corresponding length-weight relationships provided in FishBase (https://www.fishbase.cn/search.php). The trophic position of the manipulated fish species, which was evaluated based on diet composition, was also obtained from this website. The mean trophic level was used if the fish treatments included more than one species. We divided the fish species into three groups based on trophic level, namely, 2.0-2.5, 2.5-3.5, and 3.5-4.5, to represent herbivorous fish, zooplanktivorous fish and piscivorous fish, respectively. As fish may have contrasting feeding habits in different periods of ontogeny, we considered the juveniles of piscivorous fish as zooplanktivores when the literatures clearly reported it. If individual observations of water temperature were not provided, the average water temperature of the experiment was used. In addition, information on zooplankton and phytoplankton was collected as biomass and abundance separately.

Data were extracted directly from tables and text or indirectly extracted from figures using the GetData Graph Digitizer (version 2.25). The mean values, standard deviations and sample sizes of each response variable in the control and treatment groups were extracted from the publications. If the standard deviation was not reported, it was calculated by multiplying the standard error by the square root of the sample size.

Statistical analysis

The natural log-transformed response ratio (LRR) was used to measure the effect size (Hedges et al. 1999) as follows: Accepted Article

This article is protected by copyright. All rights reserved LRR = ln(Xt/Xc) where Xt is the mean treatment value and Xc is the mean value in the control. The variance (υ) of each individual effect size was calculated as follows:

2 2 푆푡 푆푐 υ = 2＋ 2 푛푡푋푡 푛푐푋푐

where nt and nc are the sample sizes of the treatment and control groups, respectively, and St and Sc are the standard deviations of the treatment and control groups, respectively. For each observation, we calculated the effect size using the “escalc” function in the R package “metafor” (Viechtbauer 2010).

A variance‐weighted mixed‐model was applied to estimate the mean effect size (LRR++) with 95% confidence intervals (CIs) of treatments. The weighted average effect sizes were significant if the 95% CIs did not contain zero. In addition, multiple comparisons (Fisher's Least Significant Difference, LSD) were used to test the differences in response variables between feeding habits. For improved interpretation, the weighted effect size was transformed back to the percentage change (%), which was evaluated as follows:

Percentage change (%) = (푒퐿푅푅 + + -1) ×100%

For each response variable, we used multilevel meta-analysis models with random effects to account for differences between studies. Specifically, a publication-level random effect was included to decrease the potential dependence of our data (Cauvy-Fraunié and Dangles 2019, Nakagawa and Santos 2012). Additionally, in the meta-data, it is common to find units that are nested in higher level clusters. For instance, in the present study, different fish species treatments are often nested in studies. Therefore, we used hierarchical random-effects meta-analysis to test the potential driving factor on the trophic cascade strength, with fish species treatment nested in studies as random effects. We first used random-effects models to assess the overall responses. We then used mixed-effects meta-regressions, including TN, TP, temperature, feeding habit of fish, absolute latitude, fish density, Accepted Article

This article is protected by copyright. All rights reserved experimental volume and experimental duration as fixed factors (moderators) for each response variable. One factor was considered to amplify the strength of trophic cascades when LRR and the regression coefficient were both significantly positive or significantly negative values. We evaluated the heterogeneity of effect sizes with the Q-statistic to determine whether the models could explain a significant amount of variation. For meta-regressions, total heterogeneity can be divided into the variance explained by the moderators (Qm) and the residual error variance (Qe). The Qm-statistic is a Wald-type test of model coefficients, and a significant Qm-statistic indicates that the moderators contribute to the heterogeneity in effect sizes (Viechtbauer 2010). To test for potential publication bias in each variable, Egger’s regression was used by regressing the effect size against their standard errors weighted by their inverse variance (Egger et al. 1997). The weighted regression slope is expected to be zero in the absence of publication bias. Trim and fill was also used to both identify and adjust for funnel plot asymmetry from publication bias (Duval and Tweedie 2000). We compared the results obtained with multilevel random effects, simple random effects and the trim-and-fill method to determine whether the results were different (Appendix S1: Table S1). The “metafor” package in R software was used to perform these analyses (Viechtbauer 2010).

To assess the relative importance of multiple factors on the trophic cascade strength, we performed multi-model inference using Akaike’s information criterion (AIC). This method conducts model selection by fitting all possible combinations based on AICs. The importance of a particular factor is expressed as the sum of the weights for the models in which the candidate factor is included. The method is widely used in meta-analyses (e.g., Feng and Zhu, 2019; Civitello et al., 2015; Crawford et al., 2019; Deng et al., 2018) because it allows the existence of variance of effect size and random effects in the model. The R package “glmulti” was used in this analysis.

Results

Meta-analysis of the whole data set revealed that fish introduction had significantly positive effects on total nitrogen (TN) (LRR++= +0.18, p < 0.001; increased by 19.6%), total phosphorus (TP) Accepted Article

71.7%), phytoplankton biomass (LRR++ = +0.60, p < 0.001; increased by 82.6%) and rotifer biomass

(LRR++ = +0.57, P < 0.001; increased by 76.1%) but negative effects on water clarity (LRR++ = -0.19, p < 0.001; decreased by 17.3%), total zooplankton biomass (LRR++ = -0.76, p < 0.001; decreased by

53.0%), cladocerans biomass (LRR++ = -1.57, p < 0.001; decreased by 79.1%), and copepods biomass

(LRR++ = -0.70, p < 0.001; decreased by 50.2%; Fig. 2 and Appendix S1: Table S2). Furthermore, fish with even trophic levels (herbivorous fish and piscivorous fish) had a significantly positive influence on primary producers (Fig. 3d,e and Appendix S1: Table S3). Zooplanktivorous fish generally led to stronger trophic cascade effects on TN, TP, water clarity and Chl a than herbivorous and piscivorous fish. Total zooplankton and copepods unexpectedly exhibited the strongest negative responses to herbivorous fish (Fig. 3f,h), suggesting that some herbivorous fish could also impose strong predation pressure on zooplankton.

The experimental background level of nutrients had a positive relationship with the LRR-phytoplankton both for biomass and abundance (Fig. 4 for TN and Appendix S1: Fig. S1 for TP), indicating that eutrophication amplified the magnitude of the trophic cascade effects on phytoplankton. For zooplankton, nutrients overall had a negative influence on the trophic cascade strength (p = 0.018 for TN and p < 0.001 for TP). Nevertheless, the trophic cascade strength among taxonomic groups showed divergent responses to eutrophication. LRR-cladoceran showed positive relationships with TN and TP, while LRR-copepods and LRR-rotifer showed negative relationships, which indicated that eutrophication amplified the trophic cascade strength on copepods but inhibited the trophic cascade strength on cladocerans and rotifers.

Mixed-effect meta-regressions revealed strong positive relationships between the experimental temperature and the magnitude of the effect sizes on TN, TP, water clarity, Chl a, biomass of total zooplankton, cladocerans and rotifers (all p < 0.001; Fig. 5 and Appendix S1: Table S4), suggesting that warming amplified the trophic cascade effects. In terms of geographic variation, there were Accepted Article

This article is protected by copyright. All rights reserved significant positive relationships between latitude and LRR-phytoplankton both for biomass (p = 0.002) and abundance (p = 0.029), whereas no significant relationships were observed for zooplankton and water quality (Appendix S1: Fig. S2).

Fish abundance significantly increased the magnitude of trophic cascade effects on TN, TP, Chl a, phytoplankton biomass and zooplankton biomass (all p < 0.001; Appendix S1: Fig. S3 and Table S5), suggesting that experiments with high fish abundance included larger changes in trophic cascades. We also assessed whether cascade effect sizes changed with increases in spatial and temporal scales, e.g., whether the effect size increased with an increase in experimental duration due to accumulative effects and decreased with increasing experimental volume due to increases in habitat heterogeneity. There were significant positive linear relationships between experimental duration and the response magnitudes of TN, TP, water clarity, Chl a and total zooplankton biomass (all p < 0.001; Appendix S1: Fig. S4). In addition, the experimental duration increased the trophic cascade magnitude on the cladoceran biomass (p = 0.013) but decreased the response magnitudes of cladoceran abundance (p < 0.001; Appendix S1: Fig. S4). The experimental size (volume) showed no relationship with the effect size of water quality, cladocerans, copepods, and rotifers, or a contrasting pattern in phytoplankton biomass vs. Chl a, suggesting that the spatial scale of experiments had little impact on the strength of trophic cascades (Appendix S1: Fig. S5 and Table S5).

Discussion

Our meta-analyses synthesized global experimental data from wide geographic regions, providing strong evidence that fish stocking imposes widespread top-down cascade effects on zooplankton, phytoplankton, water nutrients and clarity. With 2202 experimental observations from 206 studies, our work represents the most comprehensive assessment to date concerning how the effects of fish introduction on lower trophic levels and water quality are regulated by multiple factors. Specifically, fish manipulations have negative effects on zooplankton and water clarity but positive effects on phytoplankton and nutrient cycling (Fig. 6). Our synthesis showed stronger trophic cascades than Accepted Article

This article is protected by copyright. All rights reserved reported in freshwater (Brett and Goldman 1996), terrestrial (Jia et al. 2018, Schmitz et al. 2000) and marine plankton ecosystems (Shurin et al. 2002). Furthermore, the response of zooplankton varied among taxonomic groups. Cladocerans and copepods exhibited negative responses to fish stocking, while rotifers exhibited a positive response. As cladocerans generally have a larger body size and a weaker swimming ability, our results showed that cladocerans were more sensitive than copepods to fish predation.

Trophic cascade theory suggests that the response of primary producers is determined by food chain length, and thus algal biomass will benefit in three- but not two- or four-trophic-level systems (Fretwell 1987, Hansson et al. 1998). Our results confirmed that the trophic cascade strength on phytoplankton was significantly different in fish treatments with different trophic levels. Ecosystems manipulated by herbivorous or piscivorous fish generally had a lower algal biomass and better water quality than zooplanktivorous fish-manipulated ecosystems. However, our results showed that the food chain length is not always a determinant of the positive/negative responses of primary producers. For instance, herbivorous fish have a positive effect on TN and Chl a, suggesting that manipulation by introducing herbivorous fish cannot suppress the growth of phytoplankton. Conversely, as some herbivorous fish are mechanical filter feeders without specific food selection, these fish had negative influence on zooplankton, especially large cladocerans (Fig. 3f,g). The inconsistency between actual observations and theoretical predictions is probably due to the omnivorous diet displayed by these fish in nature (Attayde et al. 2010). As the feeding diet of experimental fish in our study included plants, detritus, zooplankton, zoobenthos and juvenile planktivorous fish, the fish trophic level was not a discrete integer, in contrast to that often used in theoretical models (Hansson et al. 2013). In addition, feeding habits are often flexible and may change with the ontogenetic process of fish (Werner and Gilliam 1984), which will further blur the cascading effects of food chain length in real-world ecosystems. Accepted Article

This article is protected by copyright. All rights reserved The relative strengths of bottom-up and top-down controls of standing producer biomass have long received attention in ecology. Previous studies have shown that top-down control is more influential in freshwater and marine habitats than in terrestrial ecosystems (Gruner et al. 2008, Shurin et al. 2002). As bottom-up and top-down forces are interdependent, the resource level may enhance or dampen the top-down control effects. Our results showed that water quality (TN, TP, and SD) were more sensitive to fish addition in relatively oligotrophic conditions (Fig. 4a,b,c). Moreover, eutrophication was the most important predictor of the top-down cascade effects on the biomass and abundance of primary phytoplankton (Appendix S1: Fig. S6 and S7). Eutrophication amplified the trophic cascades, consistent with the finding that nutrient enrichment strongly increases herbivory in salt marshes (Bertness et al. 2008, He and Silliman 2015). For zooplankton, eutrophication overall decreased the trophic cascade strength, which might be due to the decreasing visibility for predictor and high available food resources for zooplankton. Moreover, eutrophication amplified the response of copepods but dampened the responses of cladocerans and rotifers, indicating that the strength of trophic cascades in response to eutrophication varied among taxonomic groups of zooplankton.

Our results suggest that an increase in temperature significantly amplify the strength of top-down cascade effects on the lower-trophic-level biomass and nutrient cycling in freshwater ecosystems (Fig. 5). These findings have strong implications for understanding the impacts of climate warming on top-down control in the context of global environmental change. The strength of trophic cascades is also expected to change along latitudinal and climatic gradients, with stronger top-down control at lower latitudes and in warmer regions (i.e., the latitudinal biotic interaction hypothesis, LBIH) (Reynolds et al. 2018, Rodriguez-Castaneda 2013, Roslin et al. 2017). Nevertheless, caution should be used when interpreting and extending these to spatial patterns. A previous meta-analysis showed that warming strengthens top-down control in colder regions but weakens it in warmer regions (Marino et al. 2018). Another study suggested that top-down control may be site specific and modulated by the local experimental context and therefore shows no clear geographic patterns (Jia et Accepted Article

This article is protected by copyright. All rights reserved al. 2018, Moles et al. 2011). In the present study, both the biomass and abundance of phytoplankton showed stronger top-down control towards higher latitudes (Appendix S1: Fig. S2), which is in contrast to what LBIH predicts but is consistent with the empirical observations that fish manipulation is generally more efficient in temperate regions than in the (sub)tropics (Lazzaro 1997). This result suggested that the geographical pattern of top-down control was determined not only by the local temperature regime but also by other biological factors (e.g., predator abundance, body size, food quality, biodiversity and omnivory) that require further study (Marino et al. 2018). For instance, our results showed that increasing fish abundance could significantly enhance the strength of trophic cascade effects on lower trophic levels. Tropical regions often have less abundant large-sized crustaceans, more filtering omnivores, and more frequent intraguild predation, which may weaken the top-down biotic interactions. Therefore, at larger spatial scales, the strength of consumer-resource interactions is subject to a consumption-defense trade-off (Piovia-Scott et al. 2017), which is related to long-term evolutionary adaptation to local habitat characteristics.

Spatial and temporal scales are an obvious issue in the design of any experiment. Our results revealed no significant or consistent changes in top-down control with an increase in experimental volume, whereas experimental duration exhibited a strong positive relationship with the strength of trophic cascades (Appendix S1: Fig. S4 and Fig. S5). Our results do not support the previous hypothesis that a larger habitat size would result in smaller trophic cascades due to increased spatial heterogeneity (e.g., providing refuge for consumers), decreased predator search efficiency or a relatively low spatial resource subsidy (Borer et al. 2005, McCann et al. 2005, Shurin et al. 2006, Leroux and Loreau 2008). Accumulative effects were observed for nutrients, phytoplankton and total zooplankton with increasing temporal scale. However, the abundance of cladocerans showed a decreasing strength with increasing experimental duration. Indeed, fish can exert detectable effects on lower trophic levels and abiotic environmental factors in a short time (≤1 week), which has been widely found in many studies (Figueredo and Giani 2005, Gu et al. 2016, Wasserman et al. 2015). Accepted Article

This article is protected by copyright. All rights reserved Cladocerans may show evolutionary adaptations that allow them to resist the top-down effects in the long term. For instance, filter-feeding fish prefer consuming large plankton, which in turn leads to the evolution of smaller cladocerans and thus reduces the predation efficiency.

Conclusion

Our study employed a global meta-analysis to focus on the strength of trophic cascades in freshwater ecosystems. The trophic cascades in this study showed a stronger response to predator manipulation than previously reported in marine and terrestrial food webs (Jia et al. 2018, Schmitz et al. 2000, Shurin et al. 2002). The predator trophic level can significantly alter the response of lower-trophic-level organisms, water nutrients and clarity. However, no evidence in this study supports the idea that an even number chain length would result in reduced primary producers, probably due to the omnivory in nature. Water quality (TN, TP and water clarity) was more sensitive to fish manipulation in relatively oligotrophic conditions than in eutrophic conditions. Furthermore, we found that both eutrophication and warming could significantly enlarge the trophic cascade strength on primary producers, which has significant implications for predictions of food web dynamics under the future global environmental change. Our synthesized data showed that the strength of top-down cascade effects on phytoplankton increased with latitude, which does not support the widespread view that trophic cascades are generally more intense towards the equator. In addition, habitat size did not alter the strength of trophic cascades, while predator abundance and experimental duration were positively related to the magnitude of trophic cascades. Our study provides new knowledge on the importance of eutrophication, warming, latitudinal gradients, predator abundance and trophic level, experimental size and duration for trophic cascade strength in freshwater ecosystems. Further research on trophic cascades should focus on the possible synergistic, additive or antagonistic effects of multiple factors, which will require more empirical studies of multiple factors to clarify the combined effects. Accepted Article

This work was supported by the National Natural Science Foundation of China (91951110), the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB31000000) and the National Natural Science Foundation of China (31988102).

Supporting Information

Additional supporting information may be found online at: [link to be added in production]

References Allen, A., Gillooly, J. and Brown, J. 2005. Linking the global carbon cycle to individual metabolism. Functional Ecology 19(2), 202-213. Andrew, N.R. and Hughes, L. 2005. Herbivore damage along a latitudinal gradient: relative impacts of different feeding guilds. Oikos 108(1), 176-182. Attayde, J.L. and Hansson, L.A. 2001. Fish-mediated nutrient recycling and the trophic cascade in lakes. Canadian Journal of Fisheries and Aquatic Sciences 58(10), 1924-1931. Attayde, J.L., van Nes, E.H., Araujo, A.I., Corso, G. and Scheffer, M. 2010. Omnivory by planktivores stabilizes plankton dynamics, but may either promote or reduce algal biomass. Ecosystems 13(3), 410-420. Baskett, C. A., and D. W. Schemske. 2018. Latitudinal patterns of herbivore pressure in a temperate herb support the biotic interactions hypothesis. Ecology Letters 21, 578-587. Bell, T., Neill, W.E. and Schluter, D. 2003. The effect of temporal scale on the outcome of trophic cascade experiments. Oecologia 134(4), 578-586. Accepted Article

This article is protected by copyright. All rights reserved Bertness, M.D., Crain, C., Holdredge, C. and Sala, N. 2008. Eutrophication and consumer control of New England salt marsh primary productivity. Conservation Biology 22(1), 131-139. Borer, E.T., Seabloom, E.W., Shurin, J.B., Anderson, K.E., Blanchette, C.A., Broitman, B., Cooper, S.D. and Halpern, B.S. 2005. What determines the strength of a trophic cascade? Ecology 86(2), 528-537. Brett, M.T. and Goldman, C.R. 1996. A meta-analysis of the freshwater trophic cascade. Proceedings of the National Academy of Sciences 93(15), 7723-7726. Brooks, J.L. and Dodson, S.I. 1965. Predation, body size, and composition of plankton. Science 150(3692), 28-35. Cauvy-Fraunié, S. and Dangles, O. 2019. A global synthesis of biodiversity responses to glacier retreat. Nature Ecology & Evolution 3, 1675-1685. Civitello, D. J., J. Cohen, H. Fatima, N. T. Halstead, J. Liriano, T. A. McMahon, C. N. Ortega, E. L. Sauer, T. Sehgal, and S. Young. 2015. Biodiversity inhibits parasites: broad evidence for the dilution effect. Proceedings of the National Academy of Sciences of the United States of America 112, 8667-8671. Crawford, K. M., J. T. Bauer, L. S. Comita, M. B. Eppinga, D. J. Johnson, S. A. Mangan, S. A. Queenborough, A. E. Strand, K. N. Suding, and J. Umbanhowar. 2019. When and where plant‐ soil feedback may promote plant coexistence: a meta‐analysis. Ecology Letters 22,1274-1284. Daskalov, G.M., Grishin, A.N., Rodionov, S. and Mihneva, V. 2007. Trophic cascades triggered by overfishing reveal possible mechanisms of ecosystem regime shifts. Proceedings of the National Academy of Sciences of the United States of America 104(25), 10518-10523. Daufresne, M., Lengfellner, K. and Sommer, U. 2009. Global warming benefits the small in aquatic ecosystems. Proceedings of the National Academy of Sciences of the United States of America 106(31), 12788-12793. Accepted Article

This article is protected by copyright. All rights reserved Davis, J.M., Rosemond, A.D., Eggert, S.L., Cross, W.F. and Wallace, J.B. 2010. Long-term nutrient enrichment decouples predator and prey production. Proceedings of the National Academy of Sciences of the United States of America 107(1), 121-126. Delong, J.P., Gilbert, B., Shurin, J.B., Savage, V.M., Barton, B.T., Clements, C.F., Dell, A.I., Greig, H.S., Harley, C.D. and Kratina, P. 2015. The body size dependence of trophic cascades. American naturalist 185(3), 354-366. Deng, M., L. Liu, L. Jiang, W. Liu, X. Wang, S. Li, S. Yang, and B. Wang. 2018. Ecosystem scale trade-off in nitrogen acquisition pathways. Nature Ecology & Evolution 2, 1724-1734. Deutsch, C.A., Tewksbury, J.J., Huey, R.B., Sheldon, K.S., Ghalambor, C.K., Haak, D.C. and Martin, P.R. 2008. Impacts of climate warming on terrestrial ectotherms across latitude. Proceedings of the National Academy of Sciences of the United States of America 105(18), 6668-6672. Duval, S., and R. Tweedie. 2000. Trim and fill: a simple funnel‐plot–based method of testing and adjusting for publication bias in meta‐analysis. Biometrics 56, 455-463. Eby, L.A., Roach, W.J., Crowder, L.B. and Stanford, J.A. 2006. Effects of stocking-up freshwater food webs. Trends in Ecology & Evolution 21(10), 576-584. Egger, M., G. D. Smith, M. Schneider, and C. Minder. 1997. Bias in meta-analysis detected by a simple, graphical test. BMJ 315, 629-634. Eriksson, B.K., Ljunggren, L., Sandstrom, A., Johansson, G., Mattila, J., Rubach, A., Raberg, S. and Snickars, M. 2009. Declines in predatory fish promote bloom-forming macroalgae. Ecological Applications 19(8), 1975-1988. Estes, J.A., Terborgh, J., Brashares, J.S., Power, M.E., Berger, J., Bond, W.J., Carpenter, S.R., Essington, T.E., Holt, R.D. and Jackson, J.B. 2011. Trophic downgrading of planet Earth. Science 333(6040), 301-306. Fagan, W.F. 1997. Omnivory as a stabilizing feature of natural communities. The American Naturalist 150(5), 554-567. Accepted Article

This article is protected by copyright. All rights reserved Feng, J., and B. Zhu. 2019. A global meta-analysis of soil respiration and its components in response to phosphorus addition. Soil Biology and Biochemistry 135, 38-47. Figueredo, C.C. and Giani, A. 2005. Ecological interactions between Nile tilapia (Oreochromis niloticus, L.) and the phytoplanktonic community of the Furnas Reservoir (Brazil). Freshwater Biology 50(8), 1391-1403. Forster, J., Hirst, A.G. and Atkinson, D. 2012. Warming-induced reductions in body size are greater in aquatic than terrestrial species. Proceedings of the National Academy of Sciences of the United States of America 109(47), 19310-19314. Fretwell, S.D. 1977. The Regulation of Plant Communities by the Food Chains Exploiting Them. Perspectives in Biology and Medicine 20(2), 169-185. Fretwell, S.D. (1987) Food-chain dynamics - the central theory of ecology. Oikos 50(3), 291-301. Gruner, D.S., Smith, J.E., Seabloom, E.W., Sandin, S.A., Ngai, J.T., Hillebrand, H., Harpole, W.S., Elser, J.J., Cleland, E.E. and Bracken, M.E. 2008. A cross-system synthesis of consumer and nutrient resource control on producer biomass. Ecology Letters 11(7), 740-755. Gu, J., Jin, H., He, H., Ning, X.Y., Yu, J.L., Tan, B.C., Jeppesen, E. and Li, K.Y. 2016. Effects of small-sized crucian carp (Carassius carassius) on the growth of submerged macrophytes: Implications for shallow lake restoration. Ecological Engineering 95, 567-573. Gudasz, C., Bastviken, D., Steger, K., Premke, K., Sobek, S. and Tranvik, L.J. 2010. Temperature-controlled organic carbon mineralization in lake sediments. Nature 466(7305), 478-481. Hammerschlag, N., Schmitz, O.J., Flecker, A.S., Lafferty, K.D., Sih, A., Atwood, T.B., Gallagher, A.J., Irschick, D.J., Skubel, R. and Cooke, S.J. 2019. Ecosystem Function and Services of Aquatic Predators in the Anthropocene. Trends in Ecology & Evolution 34(4), 369-383. Hansson, L.A., Annadotter, H., Bergman, E., Hamrin, S.F., Jeppesen, E., Kairesalo, T., Luokkanen, E., Nilsson, P.A., Sondergaard, M. and Strand, J. 1998. Biomanipulation as an application of Accepted Article

This article is protected by copyright. All rights reserved food-chain theory: Constraints, synthesis, and recommendations for temperate lakes. Ecosystems 1(6), 558-574. Hansson, L.-A., Nicolle, A., Granéli, W., Hallgren, P., Kritzberg, E., Persson, A., Björk, J., Nilsson, P.A. and Brönmark, C. 2013. Food-chain length alters community responses to global change in aquatic systems. Nature Climate Change 3(3), 228-233. He, Q. and Silliman, B.R. 2015. Biogeographic consequences of nutrient enrichment for plant-herbivore interactions in coastal wetlands. Ecology Letters 18(5), 462-471. Hedges, L.V., Gurevitch, J. and Curtis, P.S. 1999. The meta‐analysis of response ratios in experimental ecology. Ecology 80(4), 1150-1156. Heisler, J., Glibert, P.M., Burkholder, J.M., Anderson, D.M., Cochlan, W., Dennison, W.C., Dortch, Q., Gobler, C.J., Heil, C.A. and Humphries, E. 2008. Eutrophication and harmful algal blooms: a scientific consensus. Harmful Algae 8(1), 3-13. Hessen, D.O. and Anderson, T.R. 2008. Excess carbon in aquatic organisms and ecosystems: Physiological, ecological, and evolutionary implications. Limnology and Oceanography 53(4), 1685-1696. Ho, C.-K. and Pennings, S.C. 2008. Consequences of omnivory for trophic interactions on a salt marsh shrub. Ecology 89(6), 1714-1722. Jackson, J.B., Kirby, M.X., Berger, W.H., Bjorndal, K.A., Botsford, L.W., Bourque, B.J., Bradbury, R.H., Cooke, R., Erlandson, J. and Estes, J.A. 2001. Historical overfishing and the recent collapse of coastal ecosystems. Science 293(5530), 629-637. Jeppesen, E., Kronvang, B., Meerhoff, M., Søndergaard, M., Hansen, K.M., Andersen, H.E., Lauridsen, T.L., Liboriussen, L., Beklioglu, M. and Özen, A. 2009. Climate change effects on runoff, catchment phosphorus loading and lake ecological state, and potential adaptations. Journal of Environmental Quality 38(5), 1930-1941. Accepted Article

This article is protected by copyright. All rights reserved Jeppesen, E., Sondergaard, M., Mazzeo, N., Meerhoff, M., Branco, C.C., Huszar, V. and Scasso, F. 2005. Lake restoration and biomanipulation in temperate lakes: Relevance for subtropical and tropical lakes. Jia, S., Wang, X., Yuan, Z., Lin, F., Ye, J., Hao, Z. and Luskin, M.S. 2018. Global signal of top-down control of terrestrial plant communities by herbivores. Proceedings of the National Academy of Sciences of the United States of America 115(24), 6237-6242. Lazzaro, X. 1997. Do the trophic cascade hypothesis and classical biomanipulation approaches apply to tropical lakes and reservoirs? Internationale Vereinigung für theoretische und angewandte Limnologie: Verhandlungen 26:719-730. Leibold, M.A. 1989. Resource edibility and the effects of predators and productivity on the outcome of trophic interactions. The American Naturalist 134(6), 922-949. Leroux, S. J., and M. Loreau. 2008. Subsidy hypothesis and strength of trophic cascades across ecosystems. Ecology Letters 11, 1147-1156. Marino, N.d.A.C., Romero, G.Q. and Farjalla, V.F. 2018. Geographical and experimental contexts modulate the effect of warming on top‐down control: a meta‐analysis. Ecology Letters 21(3), 455-466. McCann, K. S., J. Rasmussen, and J. Umbanhowar. 2005. The dynamics of spatially coupled food webs. Ecology Letters 8, 513-523. Moles, A.T., Bonser, S.P., Poore, A.G.B., Wallis, I.R. and Foley, W.J. 2011. Assessing the evidence for latitudinal gradients in plant defence and herbivory. Functional Ecology 25(2), 380-388. Nakagawa, S. and Santos, E.S. 2012. Methodological issues and advances in biological meta-analysis. Evolutionary Ecology 26(5), 1253-1274. Pace, M.L., Cole, J.J., Carpenter, S.R. and Kitchell, J.F. 1999. Trophic cascades revealed in diverse ecosystems. Trends in Ecology & Evolution 14(12), 483. Paerl, H.W. and Huisman, J. 2008. Blooms Like It Hot. Science 320(5872), 57-58. Accepted Article

This article is protected by copyright. All rights reserved Pennings, S.C. and Silliman, B.R. 2005. Linking biogeography and community ecology: Latitudinal variation in plant-herbivore interaction strength. Ecology 86(9), 2310-2319. Piovia-Scott, J., Yang, L.H. and Wright, A.N. 2017. Temporal variation in trophic cascades. Annual review of ecology, evolution, and systematics 48, 281-300. Polis, G.A. and Strong, D.R. 1996. Food web complexity and community dynamics. The American Naturalist 147(5), 813-846. Poore, A. G., A. H. Campbell, R. A. Coleman, G. J. Edgar, V. Jormalainen, P. L. Reynolds, E. E. Sotka, J. J. Stachowicz, R. B. Taylor, and M. A. Vanderklift. 2012. Global patterns in the impact of marine herbivores on benthic primary producers. Ecology Letters 15, 912-922. Reynolds, P.L., Stachowicz, J.J., Hovel, K., Bostrom, C., Boyer, K., Cusson, M., Eklof, J.S., Engel, F.G., Engelen, A.H., Eriksson, B.K., Fodrie, F.J., Griffin, J.N., Hereu, C.M., Hori, M., Hanley, T.C., Ivanov, M., Jorgensen, P., Kruschel, C., Lee, K.-S., McGlathery, K., Moksnes, P.-O., Nakaoka, M., O'Connor, M.I., O'Connor, N.E., Orth, R.J., Rossi, F., Ruesink, J., Sotka, E.E., Thormar, J., Tomas, F., Unsworth, R.K.F., Whalen, M.A. and Duffy, J.E. 2018. Latitude, temperature, and habitat complexity predict predation pressure in eelgrass beds across the Northern Hemisphere. Ecology 99(1), 29-35. Ripple, W.J., Estes, J.A., Schmitz, O.J., Constant, V., Kaylor, M.J., Lenz, A., Motley, J.L., Self, K.E., Taylor, D.S. and Wolf, C. 2016. What is a trophic cascade? Trends in Ecology & Evolution 31(11), 842-849. Rodriguez-Castaneda, G. 2013. The world and its shades of green: a meta-analysis on trophic cascades across temperature and precipitation gradients. Global Ecology and Biogeography 22(1), 118-130. Rosenblatt, A.E. and Schmitz, O.J. 2016. Climate change, nutrition, and bottom-up and top-down food web processes. Trends in Ecology & Evolution 31(12), 965-975. Accepted Article

This article is protected by copyright. All rights reserved Roslin, T., Hardwick, B., Novotny, V., Petry, W.K., Andrew, N.R., Asmus, A., Barrio, I.C., Basset, Y., Boesing, A.L. and Bonebrake, T.C. 2017. Higher predation risk for insect prey at low latitudes and elevations. Science 356(6339), 742-744. Scheffer, M., Carpenter, S. and de Young, B. 2005. Cascading effects of overfishing marine systems. Trends in Ecology & Evolution 20(11), 579-581. Scheffer, M., Portielje, R. and Zambrano, L. 2003. Fish facilitate wave resuspension of sediment. Limnology and Oceanography 48(5), 1920-1926. Schemske, D. W., G. G. Mittelbach, H. V. Cornell, J. M. Sobel, and K. Roy. 2009. Is there a latitudinal gradient in the importance of biotic interactions? Annual Review of Ecology Evolution and Systematics 40, 245-269. Schmitz, O.J., Hambäck, P.A. and Beckerman, A.P. 2000. Trophic cascades in terrestrial systems: a review of the effects of carnivore removals on plants. The American Naturalist 155(2), 141-153. Shurin, J. B., D. S. Gruner, and H. Hillebrand. 2006. All wet or dried up? Real differences between aquatic and terrestrial food webs. Proceedings of the Royal Society B: Biological Sciences 273, 1-9. Shurin, J.B., Borer, E.T., Seabloom, E.W., Anderson, K., Blanchette, C.A., Broitman, B., Cooper, S.D. and Halpern, B.S. 2002. A cross-ecosystem comparison of the strength of trophic cascades. Ecology Letters 5(6), 785-791. Strong, D.R. 1992. Are trophic cascades all wet? Differentiation and donor‐control in speciose ecosystems. Ecology 73(3), 747-754. Su, H., R. Wang, Y. Feng, Y. Li, Y. Li, J. Chen, C. Xu, S. Wang, J. Fang, and P. Xie. 2020. Long-term empirical evidence, early warning signals and multiple drivers of regime shifts in a lake ecosystem. Journal of Ecology Su, H., Wu, Y., Xia, W., Yang, L., Chen, J., Han, W., Fang, J. and Xie, P. 2019. Stoichiometric mechanisms of regime shifts in freshwater ecosystem. Water Research 149, 302-310. Accepted Article

This article is protected by copyright. All rights reserved Vanni, M.J. and Layne, C.D. 1997. Nutrient recycling and herbivory as mechanisms in the ''top-down'' effect of fish on algae in lakes. Ecology 78(1), 21-40. Vasseur, D.A. and McCann, K.S. 2005. A mechanistic approach for modeling temperature-dependent consumer-resource dynamics. The American Naturalist 166(2), 184-198. Viechtbauer, W. 2010. Conducting meta-analyses in R with the metafor package. Journal of Statistical Software 36(3), 1-48. Wasserman, R.J., Vink, T.J.F., Dalu, T. and Froneman, P.W. 2015. Fish predation regimes modify benthic diatom community structures: Experimental evidence from an in situ mesocosm study. Austral Ecology 40(7), 806-815. Werner, E. E., and J. F. Gilliam. 1984. The ontogenetic niche and species interactions in size-structured populations. Annual Review of Ecology and Systematics 15, 393-425. Yvon-Durocher, G., Montoya, J.M., Trimmer, M. and Woodward, G. 2011. Warming alters the size spectrum and shifts the distribution of biomass in freshwater ecosystems. Global change biology 17(4), 1681-1694. Zhang, X., Xie, P. and Huang, X. 2008. A review of nontraditional biomanipulation. TheScientificWorldJournal 8(1), 1184-1196. Accepted Article

Figure 1 Global distribution of fish addition experiments used in this analysis. A total of 161 sites across the world were used in this study.

Figure 2 Responses of (a) TN, total nitrogen; (b) TP, total phosphorus; (c) SD, water clarity; (d) chlorophyll a, Chl a; (e) PP, phytoplankton; (f) ZP, total zooplankton; (g) Cla, cladocerans; (h) Cop, copepods; and (i) rotifers to the top-down effects of fish. Effect sizes and 95% confidence intervals for each sample are given in increasing order. Colors indicate significant negative (red), significant positive (blue) or nonsignificant (gray) effect sizes. LRR++ is the mean effect size (weighted response ratio), and s is the SE. The numbers of samples with negative and positive responses are given in the frequency distribution chart. Yellow represents biomass and green represents abundance data.

Figure 3 Weighted response ratio for introductions of fish with different feeding habits. The errors represent 95% confidence intervals (CIs) of the weighted response ratio. The letters a, b, and c indicate significant differences between groups with different feeding habits. Her, herbivorous; Zpl, zooplanktivorous; Pis, piscivorous; TN, water total nitrogen; TP, water total phosphorus; PP, phytoplankton abundance; Chl a, chlorophyll a; ZP, total zooplankton abundance; Cla, cladoceran abundance; and Cop, copepods abundance (the same below). Yellow represents biomass and green represents abundance data.

Figure 4 Influence of water total nitrogen on top-down cascade effects determined using a mixed-effect meta-regression. Points are drawn proportionally to their 1/SE. Significant trends (p < 0.05) are shown with solid regression lines with 95% confidence intervals, and nonsignificant trends are shown with dashed lines. (+/-)(+/-)(+/-) indicates positive/negative responses to fish introduction, Accepted Article

This article is protected by copyright. All rights reserved positive/negative relationships with increasing TN (or other factors), and strengthening/weakening of the magnitude of trophic cascades, respectively (the same below). Yellow dots represent biomass, and green dots represent abundance.

Figure 5 Influence of temperature on the strength of trophic cascades caused by fish predators.

Figure 6 Schematic diagram of trophic cascade effects and possible influential factors in freshwater ecosystems. The numbers represent the weighted percentage change, and the color represents the biomass (yellow) and abundance (green). The positive and negative effects of fish addition on lower trophic level consumers, primary producers and water quality are represented by red up arrows and black down arrows, respectively. The upper half of this figure shows how multiple factors influence the trophic cascade strength on primary producers. Red arrows represent factors that can strengthen the trophic cascade strength, while gray indicates no relationship. ZP, zooplankton biomass, Cla, cladoceran biomass; Cop, copepods biomass; PP, phytoplankton biomass; Chl a, chlorophyll a; TN, total nitrogen concentration; TP, total phosphorus concentration; SD, Secchi depth (represented as water clarity). Accepted Article

LRR = 0.179 LRR = 0.210 LRR = -0.189 (a) ++ (b) ++ (c) ++ 2.00 s = 0.047 s = 0.052 s = 0.044

1.00

0 200 348 487 100 442 491 120 393 277

Effect size ï ï −1.00 100 50 60 ï ï −2.00 TN 0 TP 0 SD 0 0 1 −1 0 1 2 −1 0 1 ï ï 1 835

LRR = -0.541 LRR = 0.602 LRR = 0.453 LRR = -0.756 LRR = -1.017 (d) ++ (e) ++ ++ (f) ++ ++ s = 0.075 s = 0.128 s = 0.183 s = 0.137 s = 0.166

40 60 306 790 30 196 399 453 160 207 250 285 65 Effect size ï 30 ï 20 20 10 ï 0 0 0 ï Chl a −1 0 1 2 3 ï PP 0 2 4 ZP −4 −2 0 ï 1 1 1 963

9.00 LRR = -1.566 LRR = -1.462 LRR = -0.697 LRR = -0.901 LRR = 0.566 LRR = 0.354 ++ ++ ++ ++ ++ ++ (g) s = 0.175 s = 0.178 (h) s = 0.164 s = 0.206 (i) s = 0.198 s = 0.170 6.00 3.00

0 20 330 55 20 188 74 15 85 201 308 70 292 103 99 136 Effect size −3.00 10 10 ï 10 ï 5 −6.00 Cla 0 Cop 0 Rotifer 0 −4 −2 0 ï −4 −2 0 2 −2 0 2 4 −9.00 ï 1 763 1 1 521 Sample number (smallest to largest effect size) a a Qm = 26.24 Qm = 36.72 Qm = 28.16 0.40 (a) (b) (c) p < 0.001 p < 0.001 p < 0.001 a a b 0.20 b b a 0 b Effect size ï −0.20 ï TN TP ï SD −0.40 ï Her Zpl Pis Her Zpl Pis Her Zpl Pis

Qm = 65.89 Qm = 36.14 Qm = 6.18 Qm = 30.43 Qm = 41.60 (d) p < 0.001 (e) p < 0.001 p < 0.001 (f) p < 0.001 p < 0.001 1.00 a a a ab a a 0.75 a ab a a a b ï a a 0.50 b ï Effect size 0.25 ï ï Chl a PP ZP 0 ï ï Her Zpl Pis Her Zpl Pis Her Zpl Pis

Qm = 79.65 Qm = 65.70 Qm = 19.58 Qm = 20.08 Qm = 14.03 Qm = 33.00 (g) (h) (i) p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 p < 0.001 0 a a a a a a a a a ab a a a −1.00 a a a b b Effect size −2.00 ï

Cla Cop Rotifer ï −3.00 ï Her Zpl Pis Her Zpl Pis Her Zpl Pis Trophic level 2.00 p < 0.001 p < 0.001 p = 0.014 (a) (b) (c) (+) (-) (-) (+) (-) (-) (-) (+) (-) 1.00 0 ï Effect size −1.00 ï TN ï TP SD −2.00 ï −2 −1 0 1 2 ï ï 2 ï ï

4.00 p = 0.722 p < 0.001 p < 0.001 p = 0.018 p = 0.168 (d) (e) (+) (+) (+) (+) (+) (+) (f) (-) (+) (-) 2.00 ï

Effect size 0 Chl a PP ï ZP −2.00 ï −2 −1 0 1 2 ï 1 ï ï

3.00 p = 0.303 p < 0.001 p = 0.107 p < 0.001 p < 0.001 p = 0.827 (g) (-) (+) (-) (h) (-) (-) (+) (i) (+) (-) (-)

0 ï Effect size −3.00 ï ï Cla Cop Rotifer −6.00 ï ï −1 0 1 2 ï ï ï 1 Log10 TN (mg L-1) 1.50 p < 0.001 p < 0.001 p < 0.001 (a) (+) (+) (+) (b) (+) (+) (+) (c) (-) (-) (+) 1.00 0.50

0 ï Effect size −0.50 ï TN ï TP SD −1.00 ï 10 15 20 25 30

p < 0.001 p = 0.013 p = 0.178 p < 0.001 p < 0.001 4.00 (d) (+) (+) (+) (e) (+) (+) (+) (f) (-) (-) (+) (-) (+) (-)

2.00

ï Effect size 0

Chl a ï PP ï ZP −2.00 10 20 30 15

3.00 p < 0.001 p = 0.176 p = 0.669 p = 0.001 p < 0.001 p = 0.076 (g) (-) (-) (+) (h) (-) (+) (-) (i) (+) (+) (+) 0

Effect size −3.00 ï ï Cla Cop Rotifer −6.00 ï ï 10 15 20 25 30 15 25 Temperature (℃)