Machine learning reveals the effect of leaf temperature extremes on shifts in plant photosystem heat tolerance thresholds

Catherine Anne Pottinger, Pieter Arnold, Adrienne Nicotra , Andrea Leigh , Sonya Geange, Andrei Herdean, Lisa Danzey, Michelle Bird

Plant physiological heat tolerance thresholds can acclimate rapidly in response to changing leaf temperature, which varies considerably across microclimatic space and time. How leaf temperatures trigger shifts in these heat thresholds has not been established. We aimed to determine the influence of temporally proximal leaf temperatures (Tleaf) on leaf photosystem heat tolerance thresholds (Tcrit) for two co-occurring plant species in situ in the Australian Alps. We measured Tcrit and Tleaf over five days at 16 sites, paired by aspect (northwest, southeast) across two locations: a cold air drainage valley and a high exposed ridgeline. To investigate how Tcrit was influenced by Tleaf in the days prior, we used traditional statistical approaches (linear mixed models) and a machine learning technique. While traditional models found that Tleaf parameters explained some variation in Tcrit, machine learning identified that 85% of the variation in Tcrit was explained by both maximum and minimum leaf temperatures in the four days prior to measurement. This finding illustrates that heat tolerance acclimation is driven by exposure to not only maximum, but also minimum leaf temperatures. To uncover complex relationships between fluctuating environmental conditions and plant acclimatory responses, we recommend integrating machine learning techniques with traditional statistical methods.