Initial Setup

We carried out this outdoor mesocosm experiment at the Sierra Nevada Aquatic Research Laboratory near Mammoth Lakes, California from 16 July to 14 September 2021. The experiment lasted 61 days with sampling performed weekly and increased to every 4 days from the beginning of the heatwave to 2 weeks after the heatwave. We used 208 L cattle tanks with a randomised array with the following factors: source site (historic thermal history; four ponds), experimental DTR (low and high), and heatwave (heated and non-heated). We used a fully factorial design (16 treatments) with 4 replicates for each treatment for a total of 64 tanks. We covered all tanks with 60% shade cloth, filled them with water from nearby Convict Creek, and added water weekly to counter evaporation. Stream water was not filtered as no zooplankton have been detected in the filtrate of other mesocosm experiments at the same site (Symons et al. 2021). To establish the four different source site communities, we collected zooplankton from four pond sites with a similar average temperature (17°C–19.1°C), but varying DTRs (3.7°C, 5.9°C, 9.1°C, and 10.7°C). Pond site temperature data was collected over 5 weeks (12 July to 16 August 2020) with temperature loggers every 5 min (HOBO pendant, Onset, Massachusetts, USA). Loggers were attached to a rock to keep them in the benthic zone and placed within 1 m of the shoreline. We collected zooplankton from the source sites with horizontal tows thrown to the center of the pond from the shoreline with a 63 μm mesh conical net. The ambient densities of each source site were achieved by filtering the total volume of the tanks for each treatment (16 tanks * 208 L = 3328 L) and distributing zooplankton evenly to the tanks. We mixed zooplankton in large containers separated by site and inoculated the tanks with equal aliquots, which were achieved by adding the same volume of inoculum to each tank. We started by distributing inoculum with 1 L bottles and then switched to 100 mL bottles as the inoculum volume decreased. Each tank received zooplankton from one source site, and all tanks were inoculated the same day the zooplankton were collected (16 July 2021). The source sites, in order of increasing historic thermal variability, were: Box Pond, Hidden Pond, PD Pond, and Shell Pond. We preserved four replicates of plankton inoculum for each treatment in 70% ethanol for later enumeration. We added nutrients on the same day at average nutrient levels from the ponds from the previous summer (66.9 N μmol/L and 0.51 P μmol/L). We added 42 temperature loggers attached to a small piece of PVC pipe at the bottom of the tank to a random subset of the tanks to track temperature (HOBO pendant, Onset, Massachusetts, USA).

Thermal Variability and Heatwave Treatments

We created low- and high-thermal variability treatments through insulation. We placed each 208 L tank inside one 1135 L tank. We filled the outside tanks to a consistent level with water to create the insulated, low-thermal variability treatment, and left the other 1135 L tanks empty to create the non-insulated, high-thermal variability treatment.

To create the heatwave treatment, we used 300 W aquarium heaters for 3.5 days. During the heatwave, all heated tanks were set to a consistent half insulation level of water in the outer tanks to create the same heatwave for both thermal variability treatments. Maintaining variability treatments during the heatwave would have caused a more severe heatwave in the high-experimental thermal variability tanks (i.e., greater maximum temperatures), which would have limited the ability to distinguish the effect of insulation (hypothesized to have an acclimation effect) from the effect of the higher maximum temperatures during the heatwave. Due to electrical constraints at the research site, we separated the heatwave treatment into two time points: heatwave A (11–15 August 2021, n = 19) and heatwave B (15–19 August 2021, n = 13). Tanks were haphazardly placed into the two heatwave blocks.

Water and Community Sampling

We sampled the tanks 10 times total. All tanks were sampled for water chemistry, zooplankton, and chlorophyll a at each sampling event. We began sampling 4 days after inoculation and then weekly until the beginning of the heatwave treatments. The day before the heatwave, we began sampling every 4 days to capture the start and end of the heatwave treatment and the subsequent community responses. After 2 weeks, we returned to the weekly sampling schedule until the end of the experiment. We sampled water chemistry (pH, dissolved oxygen [DO], specific conductivity, and total dissolved solids [TDS]) with an EXO3 sonde probe (YSI, Yellow Springs, Ohio, USA). We sampled chlorophyll-a, a proxy for phytoplankton biomass, by analysing tank water in a Trilogy Laboratory Fluorometer with a Chlorophyll In Vivo module (TurnerDesigns, San Jose, California, USA). To sample zooplankton, we gently agitated the tank water with a plastic pipe and took a 2 L sample (1% of tank volume) with a Van Dorn sampler, filtered through 63 μm mesh, and preserved with 70% ethanol. For the final sample, we took 3 Van Dorn samples for a total of 6 L for each tank. All zooplankton samples were counted and identified to the lowest practical taxonomic resolution (usually species for crustaceans, and genus for rotifers) using a Leica dissecting microscope, identifying all individuals in each sample. We sampled nutrients (total nitrogen [TN] and total phosphorus [TP]) by filtering 20 mL of water through GF/F filters and analysing the filtrate on a Seal Analytical AA3 HR Nutrient Autoanalyser (Norderstedt, Germany).

Physiological Measurements

We measured CTmax and metabolic rate of adult Leptodiaptomus signicauda and Daphnia dentifera, which made up 23.2% and 15.1%, respectively, of the zooplankton biomass across all sampling times. We were able to get the most data on these species because they were visibly conspicuous and highly abundant. Due to time constraints, CTmax and metabolic rate were only sampled before and after heatwave A, excluding heatwave B from these measurements. Additionally, we excluded tanks with the second highest historic thermal variability treatment (source site PD, DTR = 9.1°C) for metabolic rate measurements. All historic thermal variability treatments were included for CTmax.

CTmax

We measured CTmax before and after heatwave A to determine how CTmax varied with source site (thermal history), between species, and before/after heatwave. Pre-heatwave measurements were taken over the 4 days leading up to the heatwave (7–10 August 2021) and post-heatwave measurements were taken over the 3 days after the heatwave (15–17 August 2021). To collect zooplankton to measure CTmax, we sampled 2 L of water from each tank using a Van Dorn water sampler. We gently agitated the tank water with a PVC pipe before sample collection. We filtered the sample through 63 μm mesh and immediately placed the collected zooplankton into filtered Convict Creek water in a 3 L holding tank. We combined samples from each replicate to use as a pool for each treatment, then used randomly selected zooplankton from these pooled samples to run trials before and after the heatwave. All replicates were included pre-heatwave (n = 4), as we did not know that we would not be able to heat all the tanks at the same time due to electrical limitations. Post-heatwave, only tanks that were included in Heatwave A were used for sampling (n = 1–4). For D. dentifera, we sampled 1–18 individuals per treatment (mean = 8.25 individuals), but missed the non-heated, low-experimental variability, highest historic thermal variability (shell) treatment. For L. signicauda, we sampled 2–30 individuals per treatment (mean = 11 individuals), but missed the non-heated, low-experimental variability, second highest historic thermal variability (PD) treatment. Treatments were missed when species were absent from the subsamples.

From each pooled sample, we picked out individuals of L. signicauda and D. dentifera in equal quantities when possible. These individuals were placed in individual 12 mL test tubes. A maximum of 24 test tubes were used at any single time point, across two aquaria with an identical set-up (12 tubes per aquaria, one zooplankton per tube). We filled each test tube with 10 mL of filtered stream water from nearby Convict Creek. We placed test tubes into an aquaria bath of 7 L of tap water at 20°C and left them for 5 min to acclimate before trials began. During trials, we increased water temperature by ~0.3°C every minute to 30°C, then ~0.1°C every minute at temperatures over 30°C using a 300 W heater with a max temperature of 34°C and a 50 W heater with a max temperature of 40°C. Both heaters were used together to achieve this warming rate, which was within range of other experimental warming rates (Geerts, Meester, and Stoks 2015; Kellermann and Sgrò 2018; Ribeiro, Camacho, and Navas 2012). We fit each aquarium with a water pump to homogenise water temperature. We defined CTmax as the temperature at cessation of motor function, following Hutchison's dynamic method (Lutterschmidt and Hutchison 1997).

Metabolic Rate

Using zooplankton samples collected from tanks as described above, we placed individual zooplankton from each treatment into 2.5 mL glass vials with aerated and filtered (63 μm) water originating from nearby Convict Creek. In order to limit feeding and control for food digestion costs, zooplankton were held in filtered water for 8 h before metabolic rate measurement. We sealed the glass vials with rubber stoppers and placed them into the tank from which they originated overnight. We used blank vials (filtered creek water only) to control microbial oxygen consumption. We measured the oxygen content of the water once immediately after incubation in each vial with a FireStingGO2 fibre-optic pocket oxygen meter (PyroScience, Aachen, Germany). We calculated oxygen consumption as a proxy of metabolism for each zooplankton, using the difference in oxygen concentration between vials containing zooplankton and vials containing creek water only. The metabolic rate of each individual was expressed per microgram mass of O2 and per hour (μg O2 μg−1 h−1).

Data Analysis

Community Analysis

We conducted all analyses in R version 4.2.1 (R Core Team 2022). Because we could not strictly use “pre-heatwave” and “post-heatwave” due to electrical constraints splitting the heatwave treatment into 2 timepoints, we created a variable that coded for “not heated” (including all samples pre-heatwave and those that remained unheated) or “post-heatwave” (including samples only from tanks after they had experienced a heatwave). We used this variable as our “Heatwave” variable in our analyses. DTR of the source sites (3.7°C, 5.9°C, 9.1°C, and 10.7°C) was used as the numeric variable for historic thermal variability for all analyses.

We assessed shifts in zooplankton community composition due to our treatments using permutational multivariate analysis of variance (PERMANOVA) based on Bray–Curtis dissimilarities using the “vegan” package (Dixon 2003). We used DTR × Experimental thermal history × Heatwave × Date as the model for our PERMANOVA analysis. We included date in the model to account for time. We used the vegan package to visualise compositional differences across treatments and time using non-metric multidimensional scaling (NMDS) based on Bray–Curtis dissimilarity and total abundances.

We used the “lme4” package to create linear mixed-effects (LME) models to analyse the independent and interactive effects of heatwave, historic thermal variability (DTR), and experimental thermal variability, with date and mesocosm ID as random variables to account for repeat measures, on each response variable (Bates et al. 2015). Our response variables included Shannon diversity, total abundance, abundances of each taxonomic group, total biomass, and community-weighted mean (CWM) individual length. Total biomass and CWM length were calculated using published averages of weights and lengths of zooplankton (Dumont, Van de Velde, and Dumont 1975; Jones et al. 2020; McCauley et al. 1990). We did not measure individual zooplankton lengths or weights in this experiment, other than body lengths for metabolic rate calculations. While we acknowledge that body lengths in our experiment may have differed from literature values, and that this method does not include intraspecific variation in body size, interspecific variation in zooplankton body size is usually higher than intraspecific variation (Gao et al. 2019).

CTmax and Metabolic Rate Analysis

For both CTmax and metabolic rate analyses, we only used tanks that experienced heatwave A (n = 19) due to time constraints. Because of this, we were able to use pre- versus post-heatwave for the heatwave variable in our model rather than the variable we described for the community composition model. For our LME model, the fixed effects included heatwave, DTR, experimental thermal history, species, and the two- and three-way interactions among them. We used date as a random effect. We did not include the four-way interaction between the fixed effects to simplify interpretation.