Reanalysis products provide spatially homogeneous coverage for a variety of climate variables, which is invaluable for regions where observational data are limited. ERA5-Land is a state of the art reanalysis system that provides high-resolution (9 km) estimates of critical land surface variables such as vertically-resolved soil temperature that is used to estimate active layer depth and permafrost extent. Recent research has shown that ERA5-Land soil temperatures show substantial warm biases over permafrost regions. Here we use mean bias subtraction (MBS), multiple linear regression (MLR) and random forest regression (RF) to perform bias correction of the ERA5-Land soil temperature product. The MLR and RF models employ 10 predictors, including soil depth, product soil temperature, information regarding the seasonal cycle of soil temperature, air temperature, vegetation, snow cover, elevation, latitude and longitude. The RF model substantially outperforms MBS and MLR over all regions and latitudes, providing an average RMSE reduction (relative to ERA5-Land soil temperature) of between 50% and 57% when the ground is snow-covered, and between 43% and 64% during the snow-free season. Most notably, the RF model reduces the RMSE over high-latitude regions north of 60N by 52% to 57% in the snow-covered season, when ERA5-Land soil temperatures exhibit an RMSE of > 4K. The bias-corrected soil temperature product provides gridded soil temperature data over the extratropical northern-hemisphere at a resolution of 5km, and will be useful for a wide range of applications, including as an initialization condition for hydrological models, and as a tool to validate model soil temperatures.