Estimates of organismal thermal tolerance are frequently used to assess physiological risk from warming, yet the assumption that these estimates are predictive of mortality has been called into question. We tested this assumption in the cold water-specialist frog, Ascaphus montanus. For seven populations, we used dynamic experimental assays to measure tadpole critical thermal maximum (CTmax) and measured mortality from chronic thermal stress for three days at different temperatures. We tested the relationship between previously–estimated population CTmax and observed mortality, as well as the strength of CTmax as a predictor of mortality compared to local stream temperatures capturing varying timescales. Populations with higher CTmax experienced significantly less mortality in the warmest temperature treatment (25℃). We also found that population CTmax outperformed stream temperature metrics as the top predictor of observed mortality. These results demonstrate a clear link between CTmax and mortality from thermal stress, contributing evidence that CTmax is a relevant metric for physiological vulnerability assessments.

We sampled tadpoles to ensure that survival was the sole contributor to fitness during the experiments. For both dynamic and static experiments (detailed below), we used handheld fish nets to collect tadpoles from streams. Tadpoles were held in stream water in insulated containers during sampling with frequent water changes to maintain temperatures. We transported tadpoles to laboratory facilities (Fort Missoula, University of Montana) for experiments via the protocol outlined in [Essner et al., 2012. Herpetol. Rev. 43, 247–249].    

Dynamic CTmax Experiments (CTmax_p)

See manuscript and references therein for full details of CTmax experiments. Briefly, we sampled 10–24 tadpoles from each of seven populations of A. montanus in Montana. We held tadpoles for three days at 8℃, a commonly-experienced temperature among populations, without food to reduce the effects of natal stream temperature and feeding. For CTmax experiments, tadpoles were moved to an experimental tank and held in individual mesh containers. They were given two minutes to habituate before temperature ramping at 0.3℃min^-1 began. CTmax was defined as the temperature at which tadpoles lost the ability to respond to tactile stimuli but fully recovered when placed in cooler temperatures. We used population median CTmax for analyses, hereafter referred to as CTmax_p, representing previously-characterized CTmax.

Static Thermal Stress Experiments

A. montanus tadpoles overwinter in their natal streams for at least one year [Brown. 1975. Comp. Biochem. Physiol 50A, 397–405]. To collect tadpoles for static thermal stress experiments, we returned two years later (3 July – 28 July, 2019) to substantially decrease the probability of sampling the same cohort. Since we permanently removed tadpoles from the population for CTmax_p experiments, we did not resample the same individual. We collected ~60 tadpoles from each population (total n= 420).

Tadpoles from each population were evenly and randomly assigned to one of five holding temperature treatments: 5℃, 10℃, 15℃, 20℃, 25℃. Temperatures were decided based on thermal regime data [Cicchino et al. In Press. Freshw. Biol]: 5℃ and 10 ℃ are commonly experienced temperatures; 15℃ and 20℃ treatments are near maximum stream temperatures; 25℃ is greater than current maximum stream temperatures but is ecologically relevant given current rates of warming [Isaak et al. 2017. Water Resour. Res. 53, 9181–9205]. Tadpoles were held in the temperature treatments for three days and fed ad libitum by placing rocks collected from their stream in the holding tanks, from which they graze algae. We maintained oxygenation using flowing water and bubblers. After three days, we counted surviving tadpoles.

Local Stream Temperatures

We characterized stream temperatures experienced by the tadpole at three timescales: 1) immediate, 2) annual, and 3) long-term/ multi-generational. To characterize the immediate and annual stream temperature metrics, we used quality-controlled [Dunham et al. 2005. USFS Tech. Rep. , 1–18] logged temperature data at 4-hour intervals from each stream (detailed in [Cicchino et al. In Press. Freshw. Biol]). Using Water Year 2018 data, we calculated “immediate thermal experience” by averaging maximum daily temperatures of the three days preceding sampling for the static mortality experiments. We calculated “annual thermal experience” by measuring the maximum temperature experienced in a year for each population. Lastly, we used modeled temperature data of 40-year averages of the mean temperature during the warmest month (August) for each stream (Isaak et al. 2016, 2017) to quantify “long-term thermal experience”. These metrics were uncorrelated with each other.

Data Analysis

Analyses were performed in R version 4.1.2 [R Core Team 2021]. Statistical significance was evaluated using α=0.05. To test differences in mortality among temperature treatments, we used a Fisher’s Exact Test for count data. Due to limited variation in mortality in other temperature treatments, we only investigated the relationship between the probability of mortality and CTmax_p in the 25℃ treatments. For this and subsequent models, we used a logistic regression with a 2-column matrix of number of deaths and number survived as the response variable. To assess the strength of CTmax_p as a predictor of mortality, we compared the performance of this model against ten other models: (1) null; each thermal experience metric modelling separately as (2–4) independent predictors, (5–7) additive predictors with CTmax_p, and (8–10) interactive predictors with CTmax_p. All models used a binomial distribution. We compared models using Akaike’s Information Criterion (Akaike, 1973) adjusted for sample size (AICc), AICc weights, and evidence ratios with the top model.

Ethics

All experimental protocols were approved by Colorado State University IACUC (16-6667AA) and the University of Montana IACUC (024-17WLDBS-042117). Collection was permitted by Montana Fish, Wildlife and Parks (Permit 2017-060-W).

All analyses were performed using R version 4.1.2.