Semi-arid rangelands in the western United States offer crucial ecosystem services and exhibit dynamic responses to climate change. Monitoring land surface phenology (LSP) in semi-arid rangelands using multi-scale satellite observations provides valuable insights to enhance management and sustainability efforts. This study investigates the multi-scale LSP metrics across fine (3 m), moderate (30 m), and coarse (450 m) resolutions, derived from PlanetScope, Harmonized Landsat and Sentinel-2 (HLS), and Visible Infrared Imaging Radiometer Suite (VIIRS). After detecting phenometrics from 3-m PlanetScope data, we analyzed phenological heterogeneity within 30-m and 450-m pixels. We then quantified the influence of land cover fractions (grasses, shrubs, and bare grounds) on phenological heterogeneity. Finally, we assessed the scaling effects and agreements among multiscale phenometrics under different climate conditions (normal and drought years). Results reveal that the phenological heterogeneity varies significantly across scales, phenometrics, and years, which is particularly higher in senescence phase, coarser pixels, and drought conditions. The heterogeneity is also asymmetrically influenced by the fraction of grasses and shrubs, where grass fraction has a greater impact on greenup onset particularly in the drought year, while shrub fraction is more influential on dormancy onset mostly in the normal year. Because of scaling effects, a phenological event at a coarser pixel is not detectable until the given event has occurred in a certain percentage of finer pixels that is inconsistent among phenometrics and climate conditions. Moreover, the agreement of phenometrics detected from three different scales also varies with phenological events and climate conditions with an average absolute difference ranging from 9 to 31 days. These findings underscore the importance of considering phenological heterogeneity and scaling effects in semi-arid rangelands.