Clouds are omnipresent in the Earth's atmosphere. Their phase composition significantly modulates their interaction with solar and terrestrial radiation, as well as precipitation formation. Particularly for clouds containing both phases, known as mixed-phase clouds, a thorough understanding of the processes governing the distribution of both liquid water and ice is imperative for their accurate representation in models, which is achieved through empirical parameterizations. Compared to liquid-phase processes, mechanisms related to ice formation have been notably understudied, with most weather prediction and global climate models still lacking descriptions of critical ice multiplication processes, capable of efficiently amplifying ice crystal concentrations at relatively warm subzero temperatures. This PhD thesis aims to investigate microphysical processes that play a crucial role in determining the ice- and liquid-phase partitioning of mixed-phase clouds. We first demonstrated the ability to accurately predict cloud droplet numbers in orographic mixed-phase clouds using cloud parcel theory, along with in-situ aerosol measurements and remotely-sensed updraft velocities. Additionally, we established a relationship that could potentially be applied to decipher cloud droplet formation regimes in virtually any type of non-precipitating boundary-layer clouds. With a primary focus on the overlooked ice-related processes, we updated the microphysics scheme of a widely-used numerical weather prediction model, to account for previously neglected ice multiplication processes. We found that the combined effect of crystal fragmentation due to collisions with seeding ice particles from above the cloud, and to a lesser extent those lofted from the snow-covered surface, significantly increased modeled ice crystal concentrations, aligning them with in-situ observations of low-level orographic clouds. These findings were further corroborated through comparisons with ground-based radar measurements in a mountainous region, underscoring the need for models to incorporate additional secondary ice processes for accurate simulations of the amount of snowfall on the ground. By coupling the model with a radar simulator, we further proposed an interpretation of complex radar signatures linked to distinct ice growth and multiplication processes. Lastly, we introduced a novel framework to represent the impact of ice multiplication in polar stratiform mixed-phase clouds, identified as the most radiatively important cloud type. This framework, developed by applying machine learning techniques to regional climate simulations, demands fewer inputs for predicting ice multiplication, making integration into large-scale models more straightforward compared to conventional secondary ice schemes. In summary, this thesis advances our understanding of microphysical processes in mixed-phase clouds from various perspectives, characterizing the droplet formation regime and identifying conditions favoring secondary ice production in polar and orographic clouds, inferring potential signatures of ice growth and multiplication from radar observations, and introducing a novel methodological tool to parameterize its impact in large-scale models. These outputs open new avenues for microphysical process descriptions in global climate models, with expected improvements in predicting precipitation patterns and the radiative properties of mixed-phase clouds on a global scale.