FOSSILS, NOT MOLECULES, PROVIDE DATES AND LOCATIONS: TOMATILLOS, ASIAN OAKS, AFRICAN CYCADS, AND MORE FROM THE OUTSTANDINGLY PRESERVED, EARLY EOCENE VOLCANIC RAINFORESTS OF PATAGONIA

Deep-time events are dated using radioisotopic analyses of mineral grains. However, much recent work hypothesizes the absolute ages of clade divergences using dated fossils as temporal calibration points on DNA-based phylogenies, in combination with estimated substitution rates. This “timetree" approach assumes that accurate chronological information about ancient events can be extracted from living tissues. Because new “calibrations” enter the literature slowly, it is particularly instructive when a large amount of fossil data appears in a short time from a single outstanding deposit. This is the case for the Laguna del Hunco caldera-lake flora of Chubut, Patagonia, Argentina, which is producing many temporally and geographically novel occurrences from a stratigraphic section that contains an ash bed dated to 52.2 +/- 0.22 Ma (⁴⁰Ar-³⁹Ar, sanidines) and two paleomagnetic reversals. Several striking new occurrences are added here. First, a tomatillo-like fruit (cf. Physalis) with the typical inflated calyx syndrome is the first fossil larger than a seed, or from the Southern Hemisphere, of the tomato family (Solanaceae). Second, acorns and abundant leaves of the Asian oak genus Castanopsis are the first Gondwanan occurrence of Fagaceae. Dramatically, these fossils show that the association of oaks with "southern" conifers (well represented at the fossil site) that dominates many montane rainforests in southeast Asia and New Guinea is likewise ancient and Gondwanan. Third, cycad fronds appear to represent the first fossil evidence of the African genus Encephalartos. A summary of these and other recently discovered taxa from Laguna del Hunco vs. their prior molecular age estimates shows large errors. Accurate molecular estimates are only likely if very closely related taxa were already known as fossils. This example illustrates the core conundrum of most molecular dating approaches: the results can only be considered reliable when validated against well-understood fossils, which are then used to improve the next analyses. Because very large errors will remain likely even with many calibrations (e.g., H. Sauquet et al. 2012, Syst. Biol.), a timetree is credible only to the extent that fossils validate the ages of each node.