Earth's tropical forests repeatedly restructured on 10(5)- to 10(7)-year timescales during the Late Paleozoic Ice Age (340-280 Ma). Although it is established that these ancient plants responded to an evolving climate, their thresholds for physiological functioning and how such thresholds led to vegetation-climate feedbacks warrants further investigation. Here, we evaluate the physiology of seven plant types that dominated Middle Pennsylvanian through early Permian tropical lowland ecosystems. We apply taxa-specific leaf morphologic and geochemical measurements, time-specific atmospheric pCO(2), pO(2), and O-2:CO2, and site-specific meteorology to Paleo-BGC, a modified version of the process-based ecosystem model Biome-BGC 4.2.1. Our study reveals four major findings. (1) The existence of pCO(2) and precipitation thresholds for loss of physiological viability that provide a mechanism for replacement of wet-adapted lycopsids and medullosans by marattialean tree ferns, which were tolerant of periodic drought, and the subsequent dominance of seasonally dry-adapted cordaitaleans and conifers. These thresholds are relevant to both vegetation turnovers within glacial-interglacial cycles of the Pennsylvanian and to major restructuring events through the later Carboniferous and early Permian coincident with intensifying paleotropical aridification. (2) Under drier conditions, the combination of higher drought tolerance and primary productivity for marattialean tree ferns, conifers, and cordaitaleans provided an ecophysiological advantage over lycopsids and medullosans. (3) Our results further indicate that, although the shift to more drought-tolerant plants in the Late Pennsylvanian and early Permian could have led to increased biomass and surface runoff, their ability to affect climate was likely limited by aridity and changes in vegetation density. (4) Finally, our modeling shows that atmospheric pCO(2) has a larger effect on physiological functioning than pO(2). These findings highlight the unique insight into extinct plant function gained by coupling empirical measurements with ecosystem modeling and, in turn, the potential to physiologically constrain paleo-vegetation shifts previously inferred from the paleobotanical record.
