129-65 Coaxing Climate Inference from Dicot Leaves: A Test in the Rainforest of Queensland Australia

Abstract:

Fossil dicotyledonous angiosperm leaves have been used to reconstruct paleoclimate since 1915, when botanists Irving W. Bailey and Edmund W. Sinnott recognized a correlation between the proportion of entire-margin leaves in a region’s flora and mean annual temperature (MAT). Such non-taxonomic climate reconstruction techniques rely on convergent evolution of leaf size and shape in response to climate conditions. However, they can be vulnerable to phylogenetic and biogeographic contingencies. The most basic method, leaf margin analysis (LMA) quantifies the relationship between the proportion of entire margin leaves and MAT to develop a least-squares predictive model. Simple to use and effective for fragmentary fossil material, LMA is sensitive to its training set and reconstructs only one paleoclimate variable. Improvements in computing power allowed consideration of more leaf features and more climate variables using canonical correspondence analysis in the climate leaf analysis multivariate program (CLAMP). Although CLAMP has evolved to include a global training set, its presence-absence scoring method has been criticized as difficult to reproduce. Improvements in digital image manipulation and hypotheses about the functional underpinning of leaf form-climate correlations led to digital leaf physiognomy (DiLP), which uses continuous measurements of leaf size and shape analysed with a custom R function. DiLP has shown promise, warranting a comparison between it, CLAMP and LMA. With development of these paleoclimate proxies weighted toward the northern hemisphere, calibration datasets tend to produce incongruent results for Southern Hemisphere applications, leading to multiple training set for CLAMP, and multiple LMA equations to account for biogeographic variation. A known-answer validation of these methods is overdue. We sampled leaves from 23 locations in Queensland, Australia. Samples were taken to mimic fossil assemblages, including 20-30 species growing in close physical proximity within primary rainforest. LMA, CLAMP and DiLP climate reconstructions were compared against actual climate measurements elicited from averaging 10 years of 0.5° x 0.5° gridded climate data.

© Copyright 2026 The Geological Society of America (GSA), all rights reserved.