Tests of the genetic structure of empirical populations typically focus on the correlative relationships between population connectivity and geographic and/or environmental factors in landscape genetics. However, such tests may overlook or misidentify the impact of candidate factors on genetic structure, especially when connectivity patterns differ between past and present populations because of shifting environmental conditions over time. Here we account for the underlying demographic component of population connectivity associated with a temporarily dynamic landscape in tests of the factors structuring population genetic variation in an Australian lizard, Lerista lineopunctulata, from 24 nuclear loci. Correlative tests did not support significant effect from factors associated with a static contemporary landscape. However, spatially explicit demographic modeling of genetic differentiation shows that changes in environmental conditions (as estimated from paleoclimatic data) and corresponding distributional shifts from the past to present landscape significantly structures genetic variation. Results from model-based inference (i.e., from an integrative modeling approach that generates spatially explicit expectations that are tested with approximate Bayesian computation) contrasts with those from correlative analyses, highlighting the importance of expanding the landscape genetic perspective to tests the links between pattern and process, revealing how factors shape patterns of genetic variation within species.