The spatial and temporal dynamics of soil temperature (TSOIL) strongly influence a wide range of biotic and abiotic processes in boreal forests. Relatively few spatial and temporal TSOIL measurements have been made in these systems; in addition, not all ecosystem models take into account the effect of changing stand structure during stand development on TSOIL dynamics. The goals of this study were to measure TSOIL and its spatiotemporal variability in a boreal forest chronosequence, develop and test a computationally simple empirical model to predict TSOIL, and quantify the effects of different approaches to deriving TSOIL on simulated ecosystem processes. TSOIL was measured hourly at six depths (0–100 cm) for 3–4 years, and sampled in a spatial grid monthly during one growing season, in a black spruce (Picea mariana (Mill.) BSP)-dominated boreal forest chronosequence. We report annual and daily TSOIL patterns, air–soil hystereses, and TSOIL changes at the freeze-thaw transition. An empirical model predicting TSOIL as of the weighted sum of past air temperatures generally accounted for 90–95% of temporal TSOIL variability at shallow depths, and 77–83% at 50–100 cm. A variety of stand structural characteristics affected the model parameters, with leaf area index (LAI) usually the most significant. Spatial TSOIL correlation, measured at depths of 2 and 10 cm, was generally constant between 5 and 30 m. The new empirical model that accounts for changes in canopy structure greatly improved the prediction of TSOIL for stands with high LAI in the Biome-BGC process model; the broader implications of this change are discussed. This model has simple data requirements – only air temperature and LAI – although the parameters given here should only be used for boreal stands with similar soil types.