How are Uncertainties in Ice Initiation in Mixed-Phase Clouds Most Impacting Climate Modeling? (Invited Presentation)

When the tops of stratiform or cumulus clouds become even weakly supercooled, mixed-phase clouds result globally. Appearance of the ice phase dramatically increases the process complexity, adding ice crystal formation via multiple pathways, followed by habit-dependent vapor exchange, riming, and aggregation processes that feed back on habit. I will argue that fundamental uncertainties in simulating the very first step—ice crystal formation rate—remain so great that climate model development is subject to a microphysics tuning challenge without a unique solution. At high latitudes, widespread stratiform mixed-phase clouds (structurally similar to "forgotten" mid-latitude altocumulus) are commonly horizontally contiguous thin layers with sufficient droplet number concentrations to prevent warm rain. It is generally accepted that primary ice formation in clouds without warm rain proceeds weakly via activation of ice-nucleating particles (INP) within droplets at a rate that increases exponentially with temperature. Precipitation of the rapidly growing ice and associated glaciation potential can be expected to scale roughly linearly with INP activation rate. However, differing approaches to the parameterization of INP may result in orders of magnitude differences in activation rate, and INP measurement methods may also differ by an order of magnitude. At lower latitudes, cumulus congestus to cumulonimbus cloud types are ubiquitous. Clouds of such depth and extent tend to host warm rain formation before tops rise to supercooled temperatures. In updrafts forming large drops or entraining them across a melting layer, ice abundance is observed to increase dramatically, reaching concentrations orders of magnitude greater than INP can explain, commonly attributed to secondary ice production (SIP). Resulting profusions of ice within convective updrafts in turn dominate the convective outflow properties over widespread stratiform rain areas. While many processes have been identified as potentially strong sources of SIP, a subset of recent laboratory, field, and modeling studies posit that large drop fracturing processes are dominant. However, it remains unclear whether laboratory-established fracturing rates are sufficient to explain observed ice formation rates. Cold-air outbreaks and widespread post-frontal clouds over the Southern Ocean lead to a "perfect storm" of overlapping uncertainties in ice formation rate, hosting both primary and secondary ice formation in adjacent regions. Uncertainties in subsequent ice microphysical processes that are roughly linearly dependent on ice number (net vapor deposition, riming, ice precipitation) suffer from their inherent uncertainties multiplied by the ice formation rate uncertainty. To cope with such combined mixed-phase process uncertainties in ModelE at NASA GISS, our main approach to is to prepare large-eddy simulations of well-observed Lagrangian mixed-phase case studies that are suitable for physics development in single-column model mode, followed by multi-parameter global autotuning to non-unique solutions.