Summary Rain-on-snow (ROS) melt production and its contribution to water available for runoff is poorly understood. In the Pacific Northwest (PNW) of the USA, ROS drives many runoff events with turbulent energy exchanges dominating the snow energy balance (EB). While previous experimental work in the PNW (most notably the H.J. Andrews Experimental Forest (HJA)) has quantified these energy balance components for a handful of ROS events, little is known about the EB components of snowmelt at HJA on an annual basis and how the relative importance of each component changes with different time periods of analysis. Beyond the few measured events at HJA and elsewhere in the PNW, there is still a lack of understanding of the dominant components of the EB during high-frequency ROS events and how much annual snowmelt is produced during ROS events. A physically based snow energy balance model (SNOBAL) was applied to data from three climate stations in the HJA to address these questions. Measurements of all required forcing data except incoming longwave radiation were made at each site. We employ the largest ROS dataset ever amassed with SNOBAL to use the model as a learning tool to characterize the snowmelt regime in the HJA. The results show that radiation dominated the melt energy balance over the period 1996–2003 while net turbulent energy exchanges were much lower than expected. Annual variability in EB components reflected duration of snowpack (snow covered period) – where later season snowpack resulted in higher radiation as percentage of the total EB. Radiation was the largest contributor to melt during ROS. These results question the general perception of turbulent energy exchange dominance of ROS and seasonal melt in the PNW. Overall, melt from ROS events was a small percentage of annual melt – for the period 1996–2003 snow season, 80–90% of snowmelt comes from non-ROS days. These results prove the highly variable spatial and temporal controls on the snowmelt regime in the Pacific Northwest.