This study presents a probabilistic model that partitions the precipitation phase based on hourly measurements from a network of radar-based disdrometers in eastern Canada. The network consists of 27 meteorological stations located in a boreal climate for the years 2020-2023. Precipitation phase observations showed a 2-m air temperature interval where probabilities of solid, liquid, and mixed precipitation significantly overlapped. Single-phase precipitation was also found to occur more frequently than mixed-phase precipitation. Probabilistic partitioning models were developed using random forest algorithms to classify precipitation as solid, liquid, or mixed phase and partition it accordingly between solid and liquid amounts. The models utilized different sets of near-surface hydrometeorological variables and atmospheric reanalysis data to assess potential operational applications. The study utilized a single and dual temperature threshold model, along with a psychrometric balance model as benchmarks. The single temperature threshold had the highest classification performance due to a low number of mixed-phase events, but it had a substantially worse partitioning error. The other benchmark models tended to over-predict mixed-phase precipitation to decrease partitioning error. All probabilistic models proposed improved phase classification by reproducing the observed overlapping precipitation phases based on 2-m air temperature. Regarding partitioning, the probabilistic models generally had lower partitioning errors, but misclassified precipitation was more costly and increased performance variability. Although mixed-phase prediction remains a challenge, this study establishes a basis for integrating automated phase observations into a hydrometeorological observation network and developing models capable of leveraging phase observations to decrease phase classification and partitioning error.