Snow density is a fundamental physical property of snowpacks used in many aspects of snow research. As an integral component in the remote sensing of snow water equivalent and parameterisation of snow models, snow density may be used to describe many important features of snowpack behaviour. The present study draws on a significant dataset of snow density and climate observations from the United States, Australia and the former Soviet Union and uses regression-based techniques to identify the dominant climatological drivers for snow densification rates, characterise densification rate variability and estimate spring snow densities from more readily available climate data. Total winter precipitation was shown to be the most prominent driver of snow densification rates, with mean air temperature and melt-refreeze events also found to be locally significant. Densification rate variance is very high at Australian sites, very low throughout the former Soviet Union and between these extremes throughout much of the US. Spring snow densities were estimated using a statistical model with climate variable inputs and best results were achieved when snow types were treated differently. Given the importance of snow density information in many snow-related research disciplines, this work has implications for current methods of converting snow depths to snow water equivalent, the representation of snow dynamics in snow models and remote sensing applications globally.