Why Do Microphysics Schemes over-Predict Ice Size in Tropical Deep Convective Updrafts?

In situ aircraft measurements from the High Altitude Ice Crystals – High Ice Water Content (HAIC-HIWC) field campaigns show that cloud-system resolving simulations of observed cases using several variable complexity bulk and bin microphysics schemes produce too many large ice particles within tropical deep convection. This is true even after controlling for the temperature, ice water content, and vertical wind speed. We build on these results published in Stanford et al. (2017) by further examining in situ measurements at temperatures around -10°C within the context of W-band and X-band radar observations from the French Falcon 20 and the National Research Council (NRC) of Canada Convair aircraft. Differences between observed and simulated ice properties within convective updrafts are produced in the mixed phase region between 0 and -10°C. Liquid water content is very limited at -10°C, while ice concentrations exceed 100 L^-1, dominated by vapor grown particles with equivalent diameters of 300-400 µm that are being lofted from warmer temperatures. This leads to ice water contents that are frequently greater than 2 g m^-3 with radar reflectivities that are less than 30 dBZ.

A vigorous secondary ice production process appears to be operating between 0 and -10°C. Despite significant ice concentrations produced via Hallett-Mossop rime splintering in simulations, liquid persists and hydrometeor growth becomes dominated by riming in simulated updrafts, producing too many large graupel particles and insufficient numbers of much smaller vapor grown particles. Combined with particle image evidence, this suggests that additional ice multiplication mechanisms that are not parameterized are operating in sampled updrafts. Along with excessive large rimed ice, a portion of the model ice size bias is produced by inadequate parameterization of snow particle size distributions and single particle properties that may be suitable for stratiform regions but not convective regions. However, a portion of the model bias may also be related to insufficient parameterization of processes such as convective mixed phase interactions or turbulent mixing. Possible additional contributions to model ice size biases by insufficient parameterization of convective mixed phase interactions and turbulent mixing are explored through examination of the vertical evolution of updraft kinematic and microphysical properties within observational and simulation datasets.