Gaps in our understanding of ice-nucleating particle sources exposed by global simulation of the UK Earth System Model

Changes in the availability of a subset of aerosol known as ice-nucleating particles (INPs) can substantially alter cloud microphysical and radiative properties. Despite very large spatial and temporal variability in INP properties, many climate models do not currently represent the link between (i) the global distribution of aerosols and INPs and (ii) primary ice production in clouds. Here we use the UK Earth System Model to simulate the global distribution of dust, marine-sourced, and black carbon INPs suitable for immersion-mode freezing of liquid cloud droplets over an annual cycle. The model captures the overall spatial and temporal distribution of measured INP concentrations, which is strongly influenced by the world's major mineral dust source regions. A negative bias in simulated versus measured INP concentrations at higher freezing temperatures points to incorrectly defined INP properties or a missing source of INPs. We find that the ability of the model to reproduce measured INP concentrations is greatly improved by representing dust as a mixture of mineralogical and organic ice-nucleating components, as present in many soils. To improve the agreement further, we define an optimized hypothetical parameterization of dust INP activity (ns(T)) as a function of temperature with a logarithmic slope of −0.175 K⁻¹, which is much shallower than existing parameterizations (e.g. −0.35 K⁻¹ for the K-feldspar data of Harrison et al., 2019). The results point to a globally important role for an organic component associated with mineral dust.