Identifying ecologically relevant reference sites is important for evaluating ecosystem recovery, but the relevance of references that are temporally static is unclear in the context of vast landscapes with disturbance and environmental contexts varying over space and time. This question is pertinent for landscapes dominated by sagebrush (Artemisia spp.) which face a suite of threats from disturbance and development but also have lengthy recovery times. Here, we applied a dynamic reference approach to studying and projecting recovery of sagebrush on former oil and gas well pads in southwestern Wyoming, USA, using over 3 decades of remote sensing data (1985–2018). We also used quantile regression to evaluate factors that may affect recovery including soils, weather, elevation, and well pad characteristics. We then created projections for percent recovery and years to recovery (relative to references and thresholds for greater sage-grouse [Centrocercus urophasianus] habitat) across the study area. This approach substantially increased our sample, and therefore inferential base, and illustrated benefits of using dynamic references and quantile regression to evaluate and project recovery of vegetation such as sagebrush. In this data release, we provide datasets used to fit models and projection maps for percent recovery and years to recovery across the study area and in areas identified as greater sage-grouse habitat.