Sampling and analytical procedures:
Detailed experimental methods are provided for experimental design and water chemistry in Murray and Baumann (2018) and Cross et al. (2019), and for respirometry methods in Schwemmer et al. (2020). In summary, spawning-ripe adult Menidia menidia were collected from Mumford Cove (41°19’25” N, 72°1’7” W), Groton, CT, in late spring and early summer of 2016 and 2017 and transported to the Rankin seawater laboratory at University of Connecticut’s Avery Point Campus. Adults were strip-spawned, and fertilized eggs were randomly distributed into 20-L rearing containers, which were placed into each treatment tank within 2h post-fertilization. All experimental methods were approved and conducted according to University of Connecticut Institutional Animal Care and Use Committee protocol #A14-032. Of six factorial experiments conducted in 2016 and 2017, experiments 1-4 quantified CO₂ × temperature effects and experiments 5-6 quantified CO₂ × oxygen effects. Experiment 1 used 400 and 2200 uatm as target pCO₂ levels, crossed with two temperatures: 17°C and 24°C. Experiments 2 and 3 factorially crossed three pCO₂ levels (400, 2200, and 4200 uatm) with three temperatures (17°C, 20°C, and 24°C). Experiment 4 used the same three target pCO₂ levels crossed with 24°C and 28°C. Experiments 5 and 6 exposed M. menidia early life stages to three levels of pCO₂ (400, 2200, and 4200 uatm) crossed factorially with three target levels of oxygen partial pressure (pO₂): normoxic (23 kPa), suboxic (12 kPa) and hypoxic (7.5 or 9 kPa). The pCO₂ levels were calculated based on measured pH, temperature, salinity, and total alkalinity (AT). AT samples were collected three times per experiment and measured using an endpoint titration. Based on these measurements, the pCO₂ (uatm) was calculated in CO2SYS (V2.1). In the CO₂ × temperature experiments, oxygen was maintained at ~100% air saturation (>20 kPa). For experiments 1 and 4 this was achieved with continuous bubbling and validated daily for each tank with a handheld probe. For experiments 2, 3, 5, and 6, dissolved oxygen (DO, mg L⁻¹) measurements were automatically taken twice hourly in each tank by a DO probe connected to a LabView program, which adjusted bubbling of CO₂-stripped air or nitrogen gas to maintain target oxygen levels.

Closed respirometry measurements were conducted on embryos randomly sampled from each treatment 1-3 days prior to hatch and larvae sampled on the day of hatching. Oxygen consumption rates were measured by two 24-channel SensorDish readers (SDR) and glass well plates equipped with an optical oxygen sensor spot in each well. Each 0.5-mL well received a single embryo or larva, and at least one well contained only treatment water to measure background microbial respiration. Well plates were sealed and placed in temperature-controlled water baths and dissolved oxygen (DO, mg L⁻¹) was recorded every fifteen seconds by the SDR software until DO had decreased by 3 mg L⁻¹ in at least one of the wells, for 15-60 minutes. In the case of the suboxic and hypoxic treatments from experiments 5 and 6, however, the trials lasted five minutes regardless of the DO differential, given the already low oxygen in the treatment water. At the end of each measurement period, embryos and larvae were checked for injury or death, and any other factors that might have affected oxygen consumption rates were noted.

Known Problems:
Embryos were not sampled in Experiment 2 due to logistical issues. Some fish IDs might be missing because the individual died or escaped, and was therefore not included in the dataset although it had already been assigned a number before respirometry.