Determining Tcrit, T50, and T95 heat tolerances
At the end of the study period, we collected random leaves from each focal individual and brought them to nearby laboratory facilities at the University of Miami. Depending on the size of the leaves, between 3 and 66 leaves were collected from each individual and used to determine the heat tolerances. Random leaflets from different leaves were sampled if species had compound leaves. Once in the lab, we used a hole punch to cut ~1.9 cm diameter disks from the leaves. We placed six leaf disks from each individual in Miracloth fabric to prevent anaerobiosis during heat treatments (Krause et al. 2010); one layer of Miracloth was placed on the abaxial leaf surface and three layers of Miracloth were placed on the adaxial leaf surface. We then placed the Miracloth-enclosed leaves into waterproof plastic bags with air removed and submerged in water baths maintained at room temperature (~23˚C), 38, 40, 42, 44, 46, 48, 50, 52, 54, or 60˚C with circulating heaters. Immediately following 15-minutes of heat treatment, we removed the leaf pieces from water baths, placed them into petri dishes lined with moist paper towels, and allowed them to recover for 24 hours at room temperature under low light (~1μmol photons m-2 s-1). Following this recovery period, we dark-adapted the leaf pieces for 20 minutes before measuring their maximum quantum yield (FV/FM) with an OS30p+handheld fluorometer (Opti-Science, Hudson, NH USA).
To estimate each species’ Tcrit and T50, we modeled the relationship of FV/FMversus treatment temperature for each plant using the ‘nls’ function in base R’s ‘stats’ package (Core 2020). We calculated Tcrit by finding the temperature where the slope of the Fv/Fm vs. temperature relationship reached 15% of its most extreme value. We calculated T50 and T95 by predicting the temperature that caused a 50% or 95% reduction in Fv/Fm compared to the control treatment as:
\(heat\ tolerance\ =\frac{log(\frac{\theta_{a}}{x}-\theta_{b})}{\theta_{c}}\)(eq. 3)
where \(\theta_{a}\) is the asymptote of the heat treatment-response variable relationship, \(\theta_{b}\) is a constant, xrepresents 50% or 95% reduction in Fv/Fm compared to control treatments, and \(\theta_{c}\) is the decay parameter. The \(\theta\) parameters were optimized and fit to the temperature-response relationship using R’s ‘nls’ function following
\(y=\frac{\theta_{a}}{1+e^{-(\theta_{b}+\theta_{c}T)}}\) (eq. 4)
where T is the heat treatment temperature (R Core Team, 2018). We generated bootstrapped means for Tcrit, T50, and T95, by randomly resampling data and fitting a new model for each species 100 times (Fig. 1b) . We present the mean bootstrapped values for Tcrit, T50, and T95.