Abstract: Whether precipitation falls as rain or snow is crucial for snow accumulation and the processes that follow. Land surface models use precipitation phase partitioning methods to estimate snowfall, but widely used methods based only on near-surface meteorology appear to have a performance ceiling. To address this, I developed EnergyPhase, a scheme based on atmospheric melting and refreezing energy, which improves rain-snow classification and snowfall estimates across North American stations. Here, I implement EnergyPhase in Noah-MP and compare it with existing partitioning schemes in offline simulations driven by CONUS404. Preliminary results for water year 2013 show differences in accumulated peak SWE from about -80 to 50 mm among schemes, with larger impacts in the Pacific Northwest and eastern United States. Higher elevations generally show larger spread in peak SWE across schemes, with the largest differences in the eastern United States occurring around 1500-2000 m. EnergyPhase produces both higher and lower SWE than the default Jordan scheme depending on region. Interestingly, changes in snow season length do not always follow changes in peak SWE. These early results suggest that improved snowfall partitioning does not translate into a simple, uniform SWE response. Instead, the impact depends on how snowfall timing, accumulation, melt, and surface energy fluxes interact through the season. Ongoing work will examine detailed snowmelt processes, surface fluxes, hydrologic responses, and regional mean behavior to better understand how precipitation phase partitioning affects snowpack evolution in Noah-MP.