From pollen to pixels: Towards spatially meaningful pollen-based vegetation reconstructions across southeastern Australia

Reconstruction of past vegetation underpins understanding of long-term ecosystem and societal change, helping to contextualise present-day patterns in biodiversity, ecosystem trajectories and effective landscape management practices against historical ecosystem states. This is particularly important in regions, such as southeastern Australia, where landscapes have been extensively shaped by long-term human activity including Indigenous fire management, creating cultural landscapes that can be difficult to identify through traditional palaeoecological techniques. Quantitative vegetation reconstruction models provide an opportunity to reconstruct these systems with greater precision but require robust validation with real-world data before application to past landscapes. Here, we present the first Southern Hemisphere application of the Extended Downscaling Approach (EDA), focussing on western Tasmania and southeastern Victoria. Using 86 modern pollen samples from moss and sedimentary records across seven study locations, we test whether the EDA reconstructs local-scale vegetation cover more accurately than uncorrected pollen percentages. By comparing EDA outputs with ground-surveyed vegetation cover, we find that the EDA reduces taxonomic representation biases, particularly for tree taxa. Across all study locations, EDA reconstructions differ from surveyed vegetation by <2.5% on average, compared with 4–9% for uncorrected pollen, while also showing higher Bray-Curtis similarity and Squared Chord Distance (SCD) values consistently below southeastern Australian analogue benchmarks. While this study does not assess the spatial outputs of the EDA, it demonstrates that the model can provide reliable local-scale vegetation reconstructions, supporting future spatial validation and Holocene applications within ancient cultural landscapes.