Timing determines zooplankton community responses to multiple stressors

Human activities and climate change cause abiotic factors to fluctuate through time, sometimes passing thresholds for organismal reproduction and survival. Multiple stressors can independently or interactively impact organisms; however, few studies have examined how they interact when they overlap spatially but occur asynchronously. Fluctuations in salinity have been found in freshwater habitats worldwide. Meanwhile, heatwaves have become more frequent and extreme. High salinity pulses and heatwaves are often decoupled in time but can still collectively impact freshwater zooplankton. The time intervals between them, during which population growth and community recovery could happen, can influence combined effects, but no one has examined these effects. We conducted a mesocosm experiment to examine how different recovery times (0‐, 3‐, 6‐week) between salt treatment and heatwave exposure influence their combined effects. We hypothesized that antagonistic effects would appear when having short recovery time, because previous study found that similar species were affected by the two stressors, but effects would become additive with longer recovery time since fully recovered communities would respond to heatwave similar to undisturbed communities. Our findings showed that, when combined, the two‐stressor joint impacts changed from antagonistic to additive with increased recovery time between stressors. Surprisingly, full compositional recovery was not achieved despite a recovery period that was long enough for population growth, suggesting legacy effects from earlier treatment. The recovery was mainly driven by small organisms, such as rotifers and small cladocerans. As a result, communities recovering from previous salt exposure responded differently to heatwaves than undisturbed communities, leading to similar zooplankton communities regardless of the recovery time between stressors. Our research bolsters the understanding and management of multiple‐stressor issues by revealing that prior exposure to one stressor has long‐lasting impacts on community recovery that can lead to unexpected joint effects of multiple stressors.


| INTRODUC TI ON
Natural ecosystems usually experience multiple stressors that occur simultaneously or asynchronously.Although there has been increasing research on multiple stressors, we have a limited understanding of their interactive effects when they overlap spatially but occur asynchronously, particularly at the community level (Gunderson et al., 2016;Jackson et al., 2021;Schäfer et al., 2023).
Stressor intensity varies through time.For example, lake nitrogen levels experienced significant temporal fluctuation due to extreme floods (Minor et al., 2014); and summer chloride concentrations in streams fluctuated temporally due to precipitation events (Lawson & Jackson, 2021).These temporal fluctuations in intensity can cause stressors to be decoupled in time (e.g., asynchronous thermal and pH stresses in Kline et al., 2015 and low-salinity and elevated temperature stresses in Agrawal & Jurgens, 2023).
The combined effects of multiple stressors can deviate from the sum of their individual effects (i.e., additive), and become greater (i.e., synergistic) or less (i.e., antagonistic) than the additive effects, making it hard to predict community responses simply based on existing single-stressor research (Jackson et al., 2021;Vinebrooke et al., 2004).
Previous exposure to a stressor usually has legacy effects that influence population, community, and ecosystem responses to subsequent stressors, through mechanisms such as acclimation, selection, and species sorting that could happen during the time intervals between stressors (Jackson et al., 2021).This knowledge gap reduces our ability to predict the interactive effects of stressors with temporal variation in intensity and restricts our ability to prioritize management initiatives for environmental conservation and restoration (Bruder et al., 2019;Côté et al., 2016;Schäfer et al., 2023).
Stressors can apply strong selection pressure to a population, with some individuals experiencing mortality or reduced birth rates, leading to evolutionary changes (Ribeiro & Lopes, 2013).These changes could be beneficial or deleterious to the population's performance under subsequent stressful conditions (Jansen et al., 2011).
Stressors can also change the species composition of a community, due to interspecific variation in tolerance to stressors, with stress-resistant species increasing in dominance (Gunderson, 2000;Sousa, 1984).These remaining species may be tolerant or sensitive to a subsequent novel stressor, affecting the combined effects of the two stressors (Vinebrooke et al., 2004).When stressors are decoupled in time, there could be a stress-free period (i.e., recovery period) between them, during which populations and communities may shift to the undisturbed condition or a new state (Gunderson, 2000;Hillebrand & Kunze, 2020).Changes in community composition can then influence community response to subsequent exposure to a stressor.For instance, Jurburg et al. (2017) suggested that the legacy effect of a heat treatment on soil bacterial community composition lasted for at least 25 days after the exposure, and they predicted that communities might be more vulnerable to a novel stressor during this period.However, the effects of asynchronously occurring multiple stressors, as well as the influence of recovery time between stressors, have been rarely examined at the community level.
Agricultural activities, mining effluents, residential run-off, and winter road salt application, have increased the salinity of freshwater lakes on a global scale (Cunillera-Montcusí et al., 2022;Hopkins et al., 2022;Roesel, 2022).Like many other chemical stressors (e.g., insecticide, Mohr et al., 2012;nutrients, Minor et al., 2014), salinity levels vary throughout the year, becoming a pulse stressor (Kansman, 2015; see Figure S1A for examples of specific conductance temporal fluctuations in lakes during spring and summer, Martin et al., n.d.;U.S. Geological Survey, 2023).Salt can be transported into freshwater habitats through direct flushing and by entering groundwater, causing high salinity pulses in lakes and ponds.
Surface flow associated with melting snow and rainfall during spring and early summer can flush de-icing salts into nearby waterbodies, resulting in weeks-to months-long increases in salinity (e.g., about 200-400 mg chloride (Cl − )/L increases in Novotny et al., 2008).
Water may later flow out of the lake and/or get diluted by summer precipitation and inflows of freshwater, which reduces its salt concentration (Lawson & Jackson, 2021;Novotny et al., 2008;Rosenberry et al., 1999).Salt can also leach through soil into groundwater and raise groundwater salinity (Howard & Beck, 1993), and, during dry periods in summer, saline groundwater discharge can deliver notable loads of salt into the freshwater habitats, temporarily elevating salt concentrations (Howard & Haynes, 1993;Williams, 1999).Additionally, summer evaporation can exacerbate the situation by decreasing the water level and concentrating ions (Jeppesen et al., 2015;Nielsen & Brock, 2009).
Previous studies have shown that increased salt concentration can compromise the growth, reproduction, and survival of organisms (Ghazy et al., 2009;Walker et al., 2023).Consequently, impacts on community structure, shifts in species composition, and alteration of food web structure also occur (Castillo et al., 2018;Moffett et al., 2023;van Meter et al., 2011).As community composition largely determines a system's resilience to stressors (e.g., in soil bacterial community, Jurburg et al., 2017;in phytoplankton community, Fugère et al., 2020), salt-induced changes in communities may affect their responses to a subsequent stressor.While there have been considerable efforts to investigate salt effects on freshwater communities, little is known about the legacy effects of exposure and the recovery capacity of communities.These are influential in determining community structure and function.Lacking this knowledge limits our understanding of how temporal salinity fluctuation affects community resistance and resilience in a changing environment.
Meanwhile, extreme events associated with climate change, such as heatwaves, occur globally and their frequency and intensity have been increasing markedly over recent decades (Woolway et al., 2021(Woolway et al., , 2022)).Additionally, the IPCC sixth report (2021) demonstrated that heatwaves are likely to increase in frequency and intensity in the future in most regions.Heatwaves usually occur briefly in spring and summer and, despite having a short duration, can be destructive to organisms and ecosystems (Stillman, 2019;Woolway et al., 2021).Acute heat stress can alter protein structures, disrupt energy generation processes, and induce cellular deterioration (Klumpen et al., 2017;Stillman, 2019).Moreover, a previous experiment by Sun and Arnott (2022) illustrated that a heatwave (increasing water temperature by about 5.2°C for 3 days) decreased zooplankton community abundance and biomass and shifted species composition.Heatwaves can also interact with other perturbations; however, existing studies have largely addressed constant warming and inadequate attention has been paid to the joint effects of this extreme event and additional stressors (Polazzo et al., 2022;Stoks et al., 2014).
Salinity stress and temperature extremes (i.e., heatwaves) usually occur as asynchronous stressors because of the temporal fluctuations of salinity in freshwater bodies and increasingly frequent heatwave events.For instance, data from U.S. Geological Survey (2023) suggests that high specific conductance and extreme water temperatures often overlap spatially but are decoupled in time (Figure S1B).The time intervals between them range from days to months, depending on when the increases in specific conductance and temperature happen.Nevertheless, past research has mainly focused on the interaction between simultaneously occurring, sustained high salinity and long-term warming (e.g., Lin et al., 2017;McClymont et al., 2023), and we lack knowledge of how pulses of salt stress and extreme heatwave events interactively affect freshwater communities (Polazzo et al., 2022;Stoks et al., 2014), despite the frequent occurrence of these stressors.Sun and Arnott (2022) found that elevated salinity and heatwaves affected similar species, thus applying them simultaneously resulted in a less-than-additive (i.e., antagonistic) effect on the zooplankton community.However, it remains unclear how asynchronous, interacting stressors affect communities' responses.To our knowledge, no one has examined the joint effects of salinity and temperature extremes when they overlap spatially but are decoupled in time.If there is recovery time between the stressors, organisms surviving salt exposure may be able to grow and reproduce, causing the joint effects to change.
To address these knowledge gaps, we conducted a mesocosm experiment using freshwater zooplankton to explore the communitylevel response to asynchronously occurring elevated salinity and heatwaves, as well as the impacts of recovery time between stressors on the joint effects.We used zooplankton as our focal organisms as they have a pivotal position in freshwater food webs and are sensitive to environmental perturbations (Paquette et al., 2022).Changes in zooplankton communities can lead to huge ecological consequences such as algal blooms and disturbed energy flow and nutrient cycling (Goleski et al., 2010).In our study, 'recovery' was defined as changes in salt-treated zooplankton communities that happened during the recovery period.We expected antagonism between salinity stress and heatwave when having short recovery times since they cause mortality in similar species (Sun & Arnott, 2022).We hypothesized that antagonistic effects would eventually become additive with the increase of recovery time, because the recovery time is long enough for communities to recover to the undisturbed condition, and fully recovered communities would respond to the heat treatment the same way as undisturbed communities.The undisturbed condition/communities refer to the control communities at different time points during the experiment (i.e., dynamic baselines), instead of a static, historical condition.Previous studies have suggested that natural communities change over time, from seasonal to decadal, and using a shifting baseline for recovery processes takes those natural changes into account (Bull et al., 2014;Paris et al., 1993).

| Field experiment
We conducted a mesocosm experiment at Queen's University Biological Station (Ontario, Canada) in the summer of 2020.We used chloride (Cl − ) concentration as the indicator of salinity level since it is an anion that forms many salts.Two Cl − levels were employed: 6 mg Cl − /L (186.72 μS/cm, ambient salt level of lake water) and 350 mg Cl − /L (1187.13μS/cm, elevated salinity treatment) which represents a Cl − concentration frequently seen in North America during summer (Lawson & Jackson, 2021;U.S. Geological Survey, 2023).For the high salinity treatment, we added 103.2 g of laboratory-grade NaCl (Thermo Fisher Scientific Inc., Massachusetts, USA) to each mesocosm.NaCl was dissolved in lake water and dispersed in the treatment mesocosms.The treatment exposure time was 3 weeks which is a realistic period of temporal salinity increase in water bodies in nature (e.g., about a month in Novotny et al., 2009;about 21 days in Lawson & Jackson, 2021).This period is also long enough to induce community responses (Delaune et al., 2021).The heatwave condition was produced by raising the water temperature 6.47 ± 0.55°C above the ambient water temperature for three consecutive days (Figure S2).The definition of heatwave varies among studies and regions (Raha & Ghosh, 2020), and we chose three consecutive days as the duration because many meteorology studies on heatwaves worldwide used this exposure time (e.g., Wang & Yan, 2021;Zacharias et al., 2014).The average and maximum water temperature reached about 26.1 ± 1.72°C and 27.01 ± 1.65°C, respectively in the heated mesocosms during the heating period.The water temperatures took 12-15 h to achieve the expected level.Similar extreme water temperatures have been observed in lakes in North America and Europe, and, in some of these lakes, surface water temperatures increased by 5°C or more within a day (e.g., Sharma et al., 2015;U.S. Geological Survey, 2023;Woolway et al., 2020;Figure S1C).To heat mesocosms evenly and for safety considerations (heaters must be fully submerged), one 100 W aquarium heater (Pawfly HT-2100, Guangdong, China; Aqueon 6101/6100, Phoenix, USA) was installed at the bottom of each heated mesocosm to increase temperature.
Mesocosms (black plastic 230 L, 53 cm tall, 77 cm diameter at the top, and 64 cm diameter at the bottom) were placed outside on the ground.We filled 40 mesocosms with 180 L of water from Lake Opinicon.Lake water was filtered through a 50μm mesh to eliminate larger zooplankton and invertebrates but keep smaller phytoplankton.Phytoplankton grew in the mesocosms for 1 week before zooplankton were added.We collected zooplankton from four oligotrophic to mesotrophic lakes: South Otter, Big Salmon, Round, and Opinicon (Table S2), using 50μm mesh nets, to establish a diverse initial community.Once collected, zooplankton were mixed and then evenly dispersed among mesocosms.We did not add sediment to the mesocosms.We covered each mesocosm with a 1-mm mesh screen to prevent aerial insect colonization and reduce sunlight by about 25%.Then, we allowed zooplankton to acclimate in the mesocosms for 3 days.After 3 days, and before any treatments commenced, initial zooplankton community and chlorophyll-a (chla) samples were taken.Throughout the experiment, mesocosms were mixed weekly.Table 1 and Figure 1 show the experimental set up of each treatment.
The salt treatment was started on different days (June 14, July 6, or July 28) for early, mid, and late summer salinity scenarios.We employed a two-factor experimental design with two salinity (low/ high salt) and temperature (ambient/heated) levels for all three timing scenarios (Table 1).At the end of their 3-week salt exposure period, zooplankton were transferred to new mesocosms with low salt conditions.The transfer could cause loss of zooplankton abundance, thus, every 3 weeks (i.e., at the end of each salt exposure), zooplankton from every tank (control and treatments) were transferred into new tanks in which the treatment was kept or newly established.
New mesocosms were filled with lake water that was filtered through a 50μm mesh 1 week before we transferred zooplankton into them.
We transferred zooplankton by filtering each old mesocosm 10 times using a 50μm mesh and gently pouring filtered zooplankton into new tanks.All heatwave treatments were performed at the same time (i.e., during the last 3 days of the experiment) to ensure that heated mesocosms all reached the same temperature level.
Specific conductance was measured every 3 weeks using an inoLab® Cond 7110 conductivity meter (Weilheim, Germany; Figure S3).Every 3 weeks, we used a YSI Pro20 dissolved oxygen (DO) meter (Yellow Springs Instruments, Yellow Springs, USA) to measure the DO of all mesocosms at about 0.2 m below the water surface (Figure S4).Water temperature was tracked hourly using temperature loggers (HOBO® Pendant®, Bourne, USA) deployed at about 0.5 m above the bottom of tanks.

| Sample collection, processing, and counting
We mixed the mesocosms before collecting chl-a and zooplankton samples.chl-a samples were taken every 3 weeks from each mesocosm as a 500 mL grab sample collected about 0.1 m below the water surface.Samples were filtered through G4 glass-fiber filters (Fisher Scientific™, Waltham, USA), frozen at −4°C, extracted in methanol for 24 h, and analyzed with a TD-700 fluorometer (Turner Designs, San Jose, USA).
Every 3 weeks (i.e., at the end of each salt treatment), zooplankton samples were collected from all the mesocosms using an 8-cm-diameter, 2 L tube sampler.The sampler captured most of the depth of the mesocosm.For each mesocosm, we obtained a total sample of 10 L by collecting samples from five locations.Samples were then filtered through a 50 μm sieve and preserved in at least 75% ethanol.We identified zooplankton using a Leica M165 C dissecting scope and a Leica DME compound microscope (Wetzlar, Germany).Each zooplankton sample was first rinsed and diluted to 100 mL, then a series of 3 mL subsamples were taken and counted until there was no new species in three subsamples in a row.All individuals were counted in each subsample, and the body length was measured.We processed at least five subsamples for each mesocosm.Except for Daphnia pulex/pulicaria, Bosmina freyi/liederi, Alona spp., and immature copepods (identified as nauplii and cyclopoid or calanoid copepodid), all the cladocerans and copepods were identified to species.Rotifers were identified to family or genus level.
Dry biomass was calculated according to measured individual body lengths and published weight-length relationships of each species (Table S2).We identified, counted, and measured initial zooplankton samples, as well as samples collected from S-early and Control at the end of Week 3, S-mid and Control at the end of Week 6, all mesocosms at the end of Week 9, and all mesocosms at the end of the experiment (i.e., end of heatwave treatment).We tested the effects of treatments based on zooplankton abundance (individuals/L), biomass (mg/L), and richness (excluding nauplii and copepodites) of all zooplankton and the four taxonomic groups: cladoceran, copepodites and adult copepod, copepod nauplii, and rotifer.We did not find significant differences across initial zooplankton communities before treatments were applied (Table S3).

| Data analysis
All the statistical analyses were performed in R 4.1.3(R Core (2) we compared salt-treated communities with control conditions at the end of each of the three treatment applications; (3) we examined the effects of the length of recovery time by analyzing community responses at the end of the experiment; (4) we examined the individual effect of heatwaves; and (5) we examined the changes in chl-a concentration through time and among treatments.
In step (1), generalized linear models (GLMs) were employed to investigate the combined effects of salt and heat treatments on zooplankton communities and each taxonomic group.We hypothesized that antagonistic effects would occur when there is no recovery time, and the type of the joint effect would change with different recovery time lengths.We applied a Gaussian (identity link function) or Gamma (log link function) distribution in GLMs (response variables were not transformed but identity or log link function was applied to models, Table S4).Residual plots and Levene's and Shapiro-Wilk's tests were used to test for assumptions of normality and homoscedasticity.Akaike information criterion (AICc, corrected for small sample sizes) scores were used to determine the best-fit distributions (Burnham & Anderson, 2002, pp. 317-323).A smaller AICc value indicates a better fit.Since different distributions and link functions have different null models (i.e., additive vs. multiplicative), which can affect the result, we ran all the analyses using both Gaussian and Gamma distributions to examine whether the findings (e.g., the type of combined effect) would change.We found that the findings remained consistent regardless of the distribution used, thus we only reported the distribution of and results from the best-fit models.A false discovery rate correction was applied only to the analyses on taxonomic groups to correct for multiple comparisons, and we only reported significant effects of the tests if they were still significant after the correction. In Step (1), we used the data collected at the end of the experiment for the control group and all the treatments to account for the shifting composition of communities.Data were analyzed for the three recovery time scenarios separately since we did not have a full factorial design when including the timing as a factor.For the three scenarios, the models included a categorical predictor variable representing the two salinity levels ("low salt" or "high salt"), a categorical variable representing the two temperature treatments ("ambient" or "heat"), and a term for their interaction.The type of interaction, if detected, was determined using the effect size of the two-stressor treatment and SD pooled was the pooled SD (Cohen, 1988).E a was the addition of each stressor's effect size.In the multiplicative model, the effect sizes were calculated as the proportion to the Control (Schäfer & Piggott, 2018), and E m was calculated using formula: , where E s and E h are the individual effects of salt and heat treatment.A table with calculated effect sizes can be found in Table S5.
Additionally, changes in zooplankton community composition were analyzed through permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001).We used the distance matrix constructed from the Bray-Curtis-transformed community matrix of zooplankton abundance for each mesocosm as the response variable.To visualize the community composition in each treatment, principal coordinates analyses (PCoA) were performed on the Bray-Curtis-transformed abundance data of all the species (except rare species that occurred in less than three mesocosms).
Analysis of variance (ANOVA) was used to examine zooplankton responses to salt exposure at different times of summer and their recovery after salt exposure.Recovery refers to any changes that happened during the recovery period.In Step (2), we compared zooplankton communities after salt exposure and in control groups in early, mid, and late-summer.In Step (3), we compared zooplankton communities in salt-only treatments and the control group at the end of Week 9. Tukey's contrast was applied to assess treatment effects when the p-value of ANOVA was less than .05.In Step (4), the individual effect of heatwaves was examined by the GLMs in the previous step.We hypothesized that the individual treatment of elevated salinity and heatwave would both negatively affect zooplankton.Finally, in Step (5), ANOVA was applied to examine changes in chl-a concentrations through time and between treatments.

| Combined effects of elevated salinity and heatwave condition
We observed antagonistic effects between salt and heat treatments on the total zooplankton abundance and biomass for 3-week and 0-week recovery treatments; the combined effects were similar or just slightly greater than the individual effect of salt or heat (GLM, p ≤ .021; Figure 2).For the 6-week recovery, no interaction was detected, and the combined effect was similar to the multiplicative effect of salt and heat and caused a 53.3% decline in abundance and a 77.8% decline in biomass (Tables S6 and S7).We found an antagonistic interaction between salt and heat treatments on species richness in the 0-week recovery treatment (GLM, p = .047),but not in other recovery scenarios where the combined effect of salt and heat was similar to the additive or multiplicative effect of individual treatments.
Interactive effects of salt and heat treatments on zooplankton community composition were only observed for the 0-week recovery time scenario (PERMANOVA, p = .040;Figure 3; Table S8).
Heat treatment shifted community composition away from the control (PERMANOVA, p ≤ .019)as did the application of salt, with salt-only 0-week (S-late) composition being further from the control.Despite the recovery of composition toward the control in the salt-only 3-week (S-mid) and 6-week (S-early) recovery scenarios, the composition of the two-stressor 0-, 3-, and 6-week recovery time scenarios (SH-late/mid/early) was similar to the salt-only 0week scenario.These four treatments were located on the opposite side of the Control on the PCoA biplots.The pattern indicates that the heat treatment did not cause further change in the composition of salt-treated communities with 0-week recovery but altered the composition of the salt-treated communities with 3-and 6-week recovery.
Combined salt and heat stress had antagonistic effects on 0week recovery nauplii abundance (GLM, p = .002)and rotifer richness (GLM, p = .004)and 3-week recovery rotifer abundance (GLM, p = .015;Figures S5-S7; Table S9).These combined treatment effects were similar to or slightly greater than the individual effects of salt or heat of the corresponding scenario.In particular, the combined treatment resulted in 69.8%, 80.9%, and 10.0% reduction in nauplii abundance, cladoceran biomass, and rotifer richness, respectively, in the 0-week recovery scenario; and 29.9% reduction in rotifer abundance the 3-week recovery scenario (Table S10).The influences were weak on copepod and rotifer abundance in all scenarios.
Biomass had a similar trend as abundance, and species richness of each taxonomic group was not much affected.

| Effects of salt treatments
The total abundance of zooplankton communities with no treatment applied (Control) was not different in early versus mid-summer, but was higher in late summer (Tukey, p ≤ .002; Figure 4A; Table S11).
The total biomass and richness of the control communities did not change much through time.Zooplankton response to salt addition depended on both the timing of addition and the response metric being considered.Adding salt in early and mid-summer did not result in significant changes to the total abundance of zooplankton compared to the Control.However, total abundance was significantly reduced in the late summer salt treatment compared to the Control (reduced by 50.7%;Tukey, p < .001; Figure 4A; Table S11).In contrast, total biomass was reduced in the salt treatment during all three exposure periods (reduced by 56.0%, 57.1%, and 65.4% in early, mid-, and late summer, respectively; Tukey, p ≤ .005; Figure 4B).Species richness declined by 33.8% and 24.6% in salt treatments during early and late-summer exposures (Tukey, p ≤ .042),but not during the midsummer exposure period (Figure 4C).Salt-treated zooplankton communities had similar total abundance, biomass, and richness values after 3 weeks of exposure, regardless of when the treatment was applied (Tukey, p ≥ .944).
Zooplankton community composition of the Control shifted through time (PERMANOVA, p = .001;Figure 4D).The community changed from being dominated by large organisms (e.g., Daphnia spp.) in early and mid-summer to smaller-sized, more abundant cladocerans (e.g., Chydorus and Alona sp.), nauplii, and rotifers in late summer.Although the composition of salt-treated communities was different among the three exposure periods (PERMANOVA, p = .042;Table S12), they were all shifted by the salt treatment toward a similar direction which is opposite to the Control.
In the Control, cladoceran abundance was similar in early and mid-summer, then increased substantially in late summer (Figure S8; Table S11).Salt treatment caused a 72.4% reduction in cladoceran abundance when it was applied in late summer (Tukey, p < .001).While salt treatment effects in early and mid-summer were not statistically  S9; Table S11).More than 56.2% and 66.7% decline in nauplii abundance and biomass were caused by salt treatment in all three timing scenarios (Tukey, p ≤ .001).
Salt treatments led to similar nauplii abundance and biomass in the communities by the end of the exposures.Copepod and rotifer abundance, biomass, and richness did not change significantly through time or among treatments (Figures S8-S10; Table S11).

| Zooplankton community recovery
To investigate zooplankton community recovery after the pulse salt treatment, we compared zooplankton responses at Week-9, when  S11   and S13).In contrast, after 3 weeks of recovery, total abundance was still 34.2%lower than the late-summer control level (Tukey, p = .014).
Zooplankton total biomass increased with the increase in recovery time but remained 53.8% and 38.5% lower than the total biomass of the control group for 3-and 6-week recovery treatments, respectively (Tukey, p ≤ .024; Figure 5B).Species richness for S-early after 6 weeks of recovery increased and was similar to Control (Figure 5C).
We found that zooplankton community composition of the S-early (after 6-week recovery) and S-mid (after 3-week recovery) moved toward the control group (Figure S11), but their compositions were still different as the abundance of some species did not increase to the control level (PERMANOVA, p = .002).
Cladoceran abundance increased after 3 and 6 weeks of recovery, compared with S-late with no recovery, and were 41.5% and 50.4% less than the Control, respectively (Figure 6; Table S13).
Cladoceran biomass did not recover as much as abundance and was still lower than the Control (Tukey, p ≤ .027; Figure S12).Both cladoceran abundance and biomass had large variations among mesocosms in S-early after recovery.Cladoceran richness in S-early also slightly increased after 6 weeks (from 1.6 to 2.4 on average; Figure S13).
Nauplii abundance and biomass both increased during the 3-and 6week time intervals.Mean rotifer abundance and biomass in S-early (post-recovery) were 81.1% and 50% higher than those in S-late, although the differences were not statistically significant.Copepod abundance, biomass, and richness did not differ among treatments.

| Effects of heatwaves
The heat-only treatment added during the last three consecutive days of the experiment led to collapsed total abundance (44.5% reduction) and biomass (59.3% reduction; GLM, p < .003; Figure 2; Tables S6 and S7).Heat treatment, as a single stressor, reduced the species richness by 17.9% compared to the Control (GLM, p ≤ .036).
Heating also caused copepod and rotifer abundance to drop by 15.6% and 40.7%, respectively, and rotifer richness to decrease by 26.7% (Figures S5 and S7).

| Chlorophyll-a concentration
Chlorophyll-a concentrations of all the treatments increased over time with some variation among mesocosms (ANOVA, p < .001).
Across all the mesocosms, the minimum, average, and maximum chl-a at the beginning of the experiment were 0.01, 0.07, and 0.28 μg/L, respectively, which increased to 0.10, 0.56, and 1.04 μg/L by the end of the experiment.We did not see any distinct patterns between treatments (Figure S14; Table S14).

| DISCUSS ION
Combined effects of elevated salinity and heatwaves changed from antagonistic to additive (or multiplicative) with the increase of recovery time after salt addition.Although abundance and biomass recovered after 6 weeks, zooplankton composition did not fully recover, as the recovery mainly happened in small, fastgrowing taxa, and legacy effects from previous salt treatment persisted for a long period (i.e., at least 6 weeks).As a result, heatwaves, as a subsequent stressor, reset the recovery process, leading to similar zooplankton communities regardless of the recovery time between stressors.
Our results suggested that the joint effect of salt and heatwave on zooplankton communities was less than the additive effect (i.e., antagonistic) when there was a 0 or 3-week recovery, but not in a 6-week recovery time scenario.Similar interactions were found for cladoceran and nauplii abundance and biomass in 0-or 3-week recovery, but not the 6-week recovery scenario.Our communities likely compositional changes in SH-early communities and in the nonsalt-treated communities (Figure 7).Although additive effects were observed in this scenario, the findings deviate from our initial hypothesis and demonstrate that salt pulses can have long-lasting effects on the community state that influences community's responses to multiple stressors.
Furthermore, we found that the recovering communities did not become more resistant to heatwaves than the undisturbed ones in Control.This is surprising as, in our communities, zooplankton that survived the salt stress should also be resistant to heat stress.A potential explanation is that heat-resistant individuals that survived Copepod nauplii and rotifers were also more prolific in late summer than earlier in the season.Copepodids and adult copepod metrics did not change much during the course of the experiment.These patterns in the Control were consistent with successional patterns observed in natural lakes (Paris et al., 1993;Romo, 1990).Large cladocerans, such as daphniids, that are vulnerable to salinity stress in other studies (e.g., Gonçalves et al., 2007;Hintz & Relyea, 2017), were the main contributors to salt-induced biomass decline and compositional changes in early and mid-summer in the experiment.
The substantial impact of salt on zooplankton communities in late summer was because of the reduction in nauplii, which are also known to be sensitive to elevated salinity (e.g., Hintz et al., 2017),  In most species that were influenced by our manipulations, our salt treatment reduced their abundance but did not extirpate them, and zooplankton communities recovered partially after being removed from high-salt conditions.The increase in total zooplankton abundance was more profound than that of total biomass, indicating that the recovery was to a large extent driven by small-sized organisms, including nauplii, small cladoceran, and rotifer taxa.Indeed, recovery potential usually varies among species, and smaller organisms tend to recover faster since they grow more rapidly (Gollner et al., 2017;Levy et al., 2007).
The recovery of zooplankton community composition was slow and uncertain in our experiment.Although community composition in our salt treatments shifted toward that of the undisturbed control during the recovery period, there remained differences, even measured in our experiment, the reproduction of some species (e.g., Daphnia, Ghazy et al., 2009) could be affected by salt exposure, which may have led to the slow recovery.This suggests that the ecological functioning of zooplankton, for instance as algae grazers and food for macroinvertebrates and fish, might remain compromised by the end of recovery periods.In accordance with our experiment, a meta-analysis conducted by Hillebrand and Kunze (2020), demonstrated that even though the abundance, biomass, and univariate biodiversity measures of disturbed communities often resembled controls after recovery, multivariate community composition is usually more vulnerable to carry-over effects of pulse perturbations.
Further studies are required to investigate whether a community can achieve full compositional and functional recovery and the time scale of the recovery.
Our heatwave treatment impaired total zooplankton abundance, biomass, and richness.The heat treatment also altered community composition, but the shift was not as noticeable as the salt treatment.Taking into account the critical role that zooplankton play in aquatic food webs, despite the short duration, this pulse stressor can have a substantial impact on the ecological conditions of freshwater habitats (Polazzo et al., 2023).This issue can be particularly severe for shallow lakes and ponds, which are much more abundant than deep lakes (Downing et al., 2006), as there are limited or no thermal refuges in these water bodies and even water temperature at the bottom can exceed organisms' tolerance threshold (e.g., Shinohara et al., 2023).We ended the experiment and collected the last sample immediately after the heatwave treatment, thus did not examine carry-over effects and community recovery from heatwaves, which is worth exploring in future research given the predicted intensification of heatwaves.
Our finding that heatwave conditions impeded and reset zooplankton community recovery has important practical implications.
Local management usually focuses on addressing local stressors (Brown et al., 2013).However, we suggest that additional initiatives should also be implemented to buffer the system from damage by global scale stressors (e.g., climate change).For instance, in regions that are experiencing freshwater salinization and likely to encounter heatwaves, other than reducing salt inputs into water bodies, local actions, such as increasing inflow or reducing extraction of water (to slow down the increase in water temperature) and making sure that dispersal corridors (such as river connections) remain open, can be taken to reduce potential heatwave damages on the recovering ecosystem.Climate change generates many uncertainties that challenge the efforts of ecological conservation and restoration (Wiens & Hobbs, 2015).But information on the impacts of extreme events attributed to climate change on the ecosystem recovery process remains very limited, influencing our ability to manage freshwater systems (González et al., 2021).
An important caveat of our study is that dispersal from nearby habitats and sediments did not happen in our experiment.Previous research showed that the dispersal of organisms from nearby habitats can bring new species and individuals to the system which may facilitate recovery (Bell, 2017;Binks et al., 2005) but may also negatively affect the local community through competition and predation (Sinclair & Arnott, 2018).Additionally, we did not measure pH changes through time, which is also a critical water quality variable, but evidence from a previous study (Hintz et al., 2017) showed that the subtle decrease in pH caused by our Cl − level increases would not affect freshwater biota.

| CON CLUS ION
Ecosystems can be disturbed collectively by local (e.g., elevated salinity) and global scale (e.g., climate change) stressors.
Identifying joint and interactive effects between multiple stressors is critical for determining the expected outcome of conservation strategies.Our findings emphasize the necessity of taking the timing of stressor exposure as well as the temporal variation of community composition into account when assessing stressor effects, particularly when using these organisms as biological indicators of impacts.We found a legacy effect of salinization on community structure and recovery processes, leading to unexpected outcomes even when multiple stressors are decoupled in time.Thus, understanding legacy effects and recovery processes is essential for ecological conservation and restoration, and more studies should be done to explore the process and time scale of Experimental set up of the Control and treatment groups.

F
Mesocosm experimental design.Solid black lines represent lowsalt condition; solid blue lines represent exposures to elevated salinity; light red curves represent ambient temperature conditions; and dark red curves represent heating conditions.The left y-axis shows chloride concentrations, and the right y-axis shows temperature levels.
combined treatment (E c ) and a predicted additive or multiplicative effect (E a or E m , respectively).Synergism means that the E c is greater than E a or E m .Antagonism means that the E c is smaller than E a or E m .We employed an additive null model (E a ) when GLMs used identity link function and a multiplicative null model (E m ) when GLMs used log link function.Cohen's d was employed as the effect size in the additive model, which was calculated using the mean and standard deviation (SD) of the data: (M control − M treatment )/SD pooled , where M control and M treatment were the mean values of the Control and each significant, there was a 63.6% and 50.2% reduction in cladoceran abundance, respectively.No difference in cladoceran biomass and richness was found over time in the Control (Figures S9 and S10).But the biomass decreased by 63.2%, 64.7%, and 70.0% when salt was added in early, mid-, and later summer, respectively (Tukey, p ≤ .003).Compared with the Control, cladoceran richness was negatively affected by salt during early-summer exposure only (61.9% reduction; Tukey, p = .009).However, despite salt being applied at different times in the season, the treatments resulted in similar abundance, biomass, and richness of cladoceran in the communities at the end of each exposure.Nauplii abundance and biomass in the Control increased in late summer (Tukey, p ≤ .001;Figures S8 and

F I G U R E 2
Boxplots showing the total abundance (a), biomass (b), and richness (c) of zooplankton communities at the end of the experiment (N = 5).Bold horizontal lines represent mean values.Salt-treated communities had different recovery time: 6 weeks (S/ SH-early); 3 weeks (S/SH-mid), or 0 week (S/SH-late)."Low salt" represents no salt was added.The Control and heat treatments are the same for each recovery time scenario.S-early had recovered for 6 weeks, S-mid had recovered for 3 weeks, and at the end of the S-late treatment.After 6 weeks of recovery, total zooplankton abundance in S-early increased and was 9.8% less than the Control in late summer (Tukey, p = .75;Figure 5A; Tables

F
Principal coordinates analyses biplot of all the species in each treatment at the end of the experiment.Each point represents a mesocosm community.Different color indicates different treatment.Bray-Curtistransformed abundance data were used.contained taxa and individuals that are similarly sensitive to both salt and heatwave conditions, such as Chydorus, Daphnia, Keratella, and nauplii.When there was no or short recovery time between treatments, heatwave impacts were masked by the previous salinity treatment because sensitive individuals were previously eliminated by salt stress and the community had not recovered to the undisturbed state.Thus, subsequent heat treatments did not generate further effects on the communities.Similar community composition between SH-late and S-late treatments supports this assertion.Our results are congruent with Vinebrooke et al.'s (2004) statement that when species tolerances to two stressors are positively correlated, exposure to one stressor can eliminate sensitive species and make the surviving community more tolerant to the other stressor, causing antagonism between stressors.It is important to recognize the antagonistic effect between these stressors because otherwise, seeing no additional effects can lead to an underestimation of heatwave impacts.After a 6-week recovery, the assemblage composition of recovering communities shifted toward the undisturbed condition but did not completely recover to the undisturbed state despite the long recovery period.This legacy effect influenced zooplankton responses to a subsequent heatwave condition and led to different F I G U R E 4 Total abundance (A), biomass (B), richness (C), and community composition (D) of zooplankton communities after salt exposure at different times of summer and communities in the control group at the corresponding times of summer (N = 5).In boxplots, bold horizontal lines represent mean values.Boxes with different letters are significantly different from each other in the Tukey's contrasts, and comparisons are within each plot only.Plot (D) is the principal coordinates analyses biplot of all the species in zooplankton communities.Each point represents a mesocosm community.Different color indicates different treatment.Bray-Curtistransformed abundance data were used for the analyses.Control-early = Control group in early summer; Controlmid = Control in mid-summer; Controllate = Control in late summer.
salt treatment failed to survive or reproduce during the recovery period, and, through time, populations became dominated by heatsensitive individuals.This could happen as an earlier stressor can affect a population's genetic composition and therefore response to future environmental conditions (i.e., multiple stressors differential tolerance hypothesis;Ribeiro & Lopes, 2013).A few studies have demonstrated that communities experiencing recurring exposure to the same stressor develop superior tolerance to the stressor as earlier exposure would select for resistant species and individuals (e.g.,Fugère et al., 2020).However, it remains unclear how the influence changes when the stressors are different.Human activities and climate change will likely intensify the temporal fluctuations of salinity and heatwaves, therefore, further research on the collective impacts of these temporally-fluctuating stressors, in different types of systems, is urgently required.For example, our zooplankton were collected from lakes with low salt levels (although South Otter and Opinicon lake may have received seasonal salt input), and it is unclear whether communities with different previous exposure and evolutionary histories would show different outcomes.Moreover, the chl-a levels in our mesocosms were lower than 1.1 μg/L throughout the experiment (oligotrophic condition), and the outcomes could change with higher chl-a concentrations because food quantity can influence salt toxicity(Brown & Yan, 2015) and more food for zooplankton could promote population growth and community recovery.Whereas, a previous study also showed that salinity and thermal stresses could change phytoplankton community composition, and increase biomass of some groups, such as cyanobacteria, which may be harmful for zooplankton(McClymont et al., 2023).Further research efforts are also desired on this aspect.Seasonal succession of the zooplankton community resulted in different effect sizes of salinity stress.As expected, zooplankton communities under control conditions changed throughout the season.While cladoceran biomass in control groups was consistent through time, their abundance increased in later summer because the assembly shifted from large, less abundant organisms (e.g., Daphnia spp.) in early and mid-summer to smaller-sized, more abundant cladocerans (e.g., Chydorus and Alona sp.) in late summer.
smaller-sized cladoceran.Additionally, salt treatment resulted in a shift in rotifer species composition with Keratella being replaced by more salt-tolerant Anuraeopsis as observed inSun and Arnott (2022).

F
Boxplots showing the total abundance (A), biomass (B), and richness (C) of salt-only treatments and control group zooplankton communities at the end of Week-9 (N = 5).Salt-only treatment communities had different recovery times: 6 weeks (S-early); 3 weeks (S-mid), or 0 week (S-late).Bold horizontal lines represent mean values.Boxes with different letters are significantly different from each other in the Tukey's contrasts, and comparisons are within each plot only.Our results suggest that temporal fluctuations in the abundance of salt-tolerant/sensitive species and individuals can produce different responses to salt stress.The temporal variation in responses to stressors should be considered when zooplankton are used as biological indicators that signal changes in lake and pond conditions (e.g.,Farkas et al., 2003), as impacts are associated with the successional stage of the community.Furthermore, plankton community seasonal succession tends to differ in lakes with different trophic status (e.g.,Wentzky et al., 2020) which also deserves caution.Other studies have also found seasonal variation in response to stressors.For example,Baggini et al. (2014) demonstrated that seasonal variation in species composition in the benthic macroalgal community strongly affected their response to ocean acidification.
after 6 weeks of recovery.Six weeks is enough time for substantial population growth as most zooplankton species have short growth and reproduction time (e.g., Daphnia pulicaria mature and start to reproduce in 7-10 days).For some species, the salinity treatment was not extreme enough to extirpate them, thus allowing for population re-growth of salt-sensitive species.However, some species, such as Bosmina, Chydorus, and Daphnia ambigua, did not completely recover within the course of our research.Although it was not directly F I G U R E 6 Boxplots showing the abundance of each taxonomic group in salt-only treatments and control group zooplankton communities at the end of Week 9 (N = 5).Salt-only treatment communities had different recovery times: 6 weeks (S-early); 3 weeks (Smid), or 0 week (S-late).Bold horizontal lines represent mean values.Boxes with different letters are significantly different from each other in the Tukey's contrasts, and comparisons are within each plot only.F I G U R E 7 Conceptual figure showing hypothesized and actual outcomes of the single and combined effects of elevated salinity and heatwaves with different recovery times.The state indicates zooplankton community composition.Circles with different colors represent different community compositions.UC E , undisturbed (Control) community in early summer; UC L , undisturbed (Control) community in late summer; UC M , undisturbed (Control) community in mid-summer.The numbers indicate the order of treatment application.
communities' recovery from pulse stressors.Additionally, the determination and effective management of multiple stressors effects needs to consider not only the nature of stressors but also their temporal scales and fluctuations.AUTH O R CO NTR I B UTI O N S Xinyu Sun: Conceptualization; data curation; formal analysis; investigation; methodology; visualization; writing -original draft; writing -review and editing.Shelley E. Arnott: Conceptualization; funding acquisition; investigation; supervision; writing -review and editing.