Measuring woody vegetation cover (WVC) is an essential part of ecological monitoring in rangelands. While continental-scale products that map vegetation cover from satellite remote sensing are available, it is uncertain whether these products are well suited for analyses at the sub-catchment and property scale. We evaluated a novel approach to regional-scale mapping of WVC in the southern rangelands of Western Australia, with focus on two ex-pastoral properties. We used random forest regression and a segmented linear regression trained on high resolution satellite imagery and applied to Landsat imagery, and validated them using an independent, high-resolution and multi-year dataset. We compared the performance of our models to that of a broadly used continental-scale fractional cover (FC) dataset available for Australia. We tested the sensitivity of our locally calibrated WVC models to monitor vegetation recovery following chain-clearing in 1973 compared with an adjacent un-cleared patch. Both locally calibrated models were five times more accurate at estimating WVC than the national-scale FC data. The cleared and uncleared plots showed expected trends in WVC (increasing and homeostatic, respectively) using both locally trained models. This study demonstrated that a segmented linear model based on a spectral index can reliably predict WVC at a pastoral-property scale over a multi-annual period. In contrast, continental scale products, while valid to track trends in WVC at larger scales, should be used with caution when guiding local management. The construction of a national product that consolidates regionally-calibrated models of WVC would provide more accurate local information and aggregated national trends.