Ice-nucleating particles (INPs) are rare particles in the atmosphere that can have disproportionately large impacts on several climate-relevant properties of clouds, including their contribution to precipitation outcomes and the cooling of the earth’s surface. Ice nucleation occurs on scales that are relatively small compared to model grid cells (down to nanometers), and thus are commonly parameterized for representation in cloud or climate models. However, critical gaps in knowledge challenge the development of ice nucleation representations. Challenges include the up to five order of magnitude span in observations of INPs at any given temperature, the numerable INP sources of varying strengths, and the fact that little is known about which properties make an aerosol an INP. Of the common INP species, desert dust is thought to be the most globally important due to its relative abundance and ice nucleating potential, whereas INPs emitted from the ocean surface are implicated in remote ocean regions far from dust sources. Gaps in INP observations near important dust and marine source regions additionally challenge efforts to advance predictive understanding of INPs. Through a synthesis of observational, laboratory and modelling techniques, this dissertation aims to develop INP instrumentation and methods, identify specific marine ice-nucleating entities, and provide observations of INPs near major dust sources. Key results include an automated instrument for measurement of immersion-mode INPs, best practices for offline INP analysis of precipitation samples, INP observations from a shipborne campaign over the Red Sea, Indian Ocean, Arabian Gulf and Mediterranean, and identities of 14 ice-nucleating microbes, at least 2 of which are high likely marine in origin. Finally, to address challenges that inhibit the implementation of improved INP representation in climate models, a data processing “pipeline” is described that will facilitate hydrometeor phase evaluation of the Department of Energy Exascale Earth Systems Atmosphere Model (EAMv1). In summary, this dissertation addresses challenges to the development and application of improved representations of ice nucleation in climate models from both the observational and modeling perspective.