Disentangling the contrasting influences of local ecosystem properties on soil temperatures is critical to understanding drivers of spatial variability in ground thermal conditions and permafrost distribution. In this study, we investigate the influence of vegetation, snow, and soil conditions on ground surface temperatures at two subarctic research basins in coastal Labrador. Eighteen ground surface temperature loggers were deployed at each basin using a stratified random deployment protocol based on land cover and snow thickness strata obtained from summer and winter uncrewed aerial vehicle surveys. Site-specific field information was also collected at each logger location in both summer and winter to contextualize ground surface temperature variability. At the southern field site (Pinware River Hills), derived microclimate indices had statistically significant associations with ground temperatures, whereas at the northern site (Nain Bay Hills), ecosystem properties including snow thickness, were important predictors of temperature variability. Ensemble model simulations of permafrost probability indicate that only 10% of loggers had more than 50% probability of having permafrost, but all of these loggers were found at Nain Bay Hills in low-snow tundra and wetland ecotypes. This research will inform regional permafrost distribution modelling and our understanding of the microclimate drivers or ground temperature variability in northern regions.