Reanalysis products provide continuous spatial coverage for a wide variety of climate variables, including snow cover, and soil temperatures. Previous research has highlighted that while sizeable biases in reanalysis soil temperature exist when snow cover is present, over mountainous Snow Telemetry (SNOTEL) stations in the western United States, errors in ERA5-Land soil temperatures are substantially smaller when snow is present. This has been attributed to offsetting errors arising from complex interactions between snow cover, elevation and soil temperatures over topographically variable terrain. Here, we examine how errors in elevation, snow cover, and snow thermal conductivity impact soil temperature estimates in three reanalysis systems (ERA5, ERA5-Land and MERRA2) and one LDAS product (FLDAS) at SNOTEL stations. The reanalysis products operate at a range of resolutions between 0.1 and 0.5 degrees, and utilize different land surface models. All products underestimate peak SWE, by up to 80%, and a bias toward peak SWE occurring 1-2 months early in MERRA2 and FLDAS leads to soil temperatures warming too early in the spring. In ERA5 and ERA5-Land, mean soil temperature biases are near 0K for DJF and 2.4K-3.5K in JJA. In FLDAS and MERRA2, mean soil temperature is biased cold in DJF (-4.8K and -1.9K, respectively) and biased warm in JJA (0.8K and 4.5K, respectively). Summer (JJA) soil temperature biases are directly related to elevation biases. For winter (DJF), the causes of soil temperature biases are complex, with differences in snow thermal conductivity and snow cover between the products a dominant factor.