Accounting for atmospheric carbon dioxide variations in pollen-based reconstruction of past hydroclimates

Changes in atmospheric carbon dioxide (CO2) concentration directly influence the ratio of stomatal water loss to carbon uptake. This ratio (e) is a fundamental quantity for terrestrial ecosystems, as it defines the water requirement for plant growth. Statistical and analogue-based methods used to reconstruct past hydroclimate variables from fossil pollen assemblages do not take account of the effect of CO2 variations on e. Here we present a general, globally applicable method to correct for this effect. The method involves solving an equation that relates e to a climatic moisture index (MI, the ratio of mean annual precipitation to mean annual potential evapotranspiration), mean growing-season temperature, and ambient CO2. The equation is based on the least-cost optimality hypothesis, which predicts how the ratio (χ) of leaf-internal to ambient CO2 varies with vapour pressure deficit (vpd), growing-season temperature and atmospheric pressure, combined with experimental evidence on the response of χ to the CO2 level at which plants have been grown. An empirical relationship based on global climate data is used to relate vpd to MI and growing-season temperature. The solution to the equation allows past MI to be estimated from pollen-reconstructed MI, given past CO2 and temperature. This MI value can be used to estimate mean annual precipitation, accounting for the effects of orbital variations, temperature and cloud cover (inferred from MI) on potential evapotranspiration. A pollen record from semi-arid Spain that spans the last glacial interval is used to illustrate the method. Low CO2 leads to estimated MI being larger than reconstructed MI during glacial times. The CO2 effect on inferred precipitation was partly offset by increased cloud cover; nonetheless, inferred precipitation was greater than present almost throughout the glacial period. This method allows a more robust reconstruction of past hydroclimatic variations than currently available tools. © 2022 The Authors