Aim: To test competing hypotheses about the timing and extent of Holocene landscape opening using pollen-based quantitative land-cover estimates. Location: Dove Lake, Tasmanian Wilderness World Heritage Area, Australia. Methods: Fossil pollen data were incorporated into pollen dispersal models and corrected for differences in pollen productivity among key plant taxa. Mechanistic models (REVEALSâ€”Regional Estimates of VEgetation Abundance from Large Sites) employing different models for pollen dispersal (Gaussian plume and Lagrangian stochastic models) were evaluated and applied in the Southern Hemisphere for the first time. Results: Validation of the REVEALS model with vegetation cover data suggests an overall better performance of the Lagrangian stochastic model. Regional land-cover estimates for forest and non-forest plant taxa show persistent landscape openness throughout the Holocene (average landscape openness ~50%). Gymnoschoenus sphaerocephalus, an indicator of moorland vegetation, shows higher values during the early Holocene (11.7â€“9Â ka) and declines slightly through the mid-Holocene (9â€“4.5Â ka) during a phase of partial landscape afforestation. Rain forest cover reduced (from ~40% to ~20%) during the period between 4.2â€“3.5Â ka. Main conclusions: Pollen percentages severely under-represent landscape openness in western Tasmania and this bias has fostered an over-estimation of Holocene forest cover from pollen data. Treeless vegetation dominated Holocene landscapes of the Dove Lake area, allowing us to reject models of landscape evolution that invoke late-Holocene replacement of a rain forest-dominated landscape by moorland. Instead, we confirm a model of Late Pleistocene inheritance of open vegetation. Rapid forest decline occurred after c.Â 4Â ka, likely in response to regional moisture decline.