The microphysics of clouds over the Antarctic Peninsula – Part 2: modelling aspects within Polar WRF

Abstract. The first intercomparisons of cloud microphysics schemes implemented in the Weather Research and Forecasting (WRF) mesoscale atmospheric model (version 3.5.1) are performed on the Antarctic Peninsula using the polar version of WRF (Polar WRF) at 5 km resolution, along with comparisons to the British Antarctic Survey's aircraft measurements (presented in part 1 of this work; Lachlan-Cope et al., 2016). This study follows previous works suggesting the misrepresentation of the cloud thermodynamic phase in order to explain large radiative biases derived at the surface in Polar WRF continent-wide (at 15 km or coarser horizontal resolution) and in the Polar WRF-based operational forecast model Antarctic Mesoscale Prediction System (AMPS) over the Larsen C Ice Shelf at 5 km horizontal resolution. Five cloud microphysics schemes are investigated: the WRF single-moment five-class scheme (WSM5), the WRF double-moment six-class scheme (WDM6), the Morrison double-moment scheme, the Thompson scheme, and the Milbrandt–Yau double-moment seven-class scheme. WSM5 (used in AMPS) and WDM6 (an upgrade version of WSM5) lead to the largest biases in observed supercooled liquid phase and surface radiative biases. The schemes simulating clouds in closest agreement to the observations are the Morrison, Thompson, and Milbrandt schemes for their better average prediction of occurrences of clouds and cloud phase. Interestingly, those three schemes are also the ones allowing for significant reduction of the longwave surface radiative bias over the Larsen C Ice Shelf (eastern side of the peninsula). This is important for surface energy budget consideration with Polar WRF since the cloud radiative effect is more pronounced in the infrared over icy surfaces. Overall, the Morrison scheme compares better to the cloud observation and radiation measurements. The fact that WSM5 and WDM6 are single-moment parameterizations for the ice crystals is responsible for their lesser ability to model the supercooled liquid clouds compared to the other schemes. However, our investigation shows that all the schemes fail at simulating the supercooled liquid mass at some temperatures (altitudes) where observations show evidence of its persistence. An ice nuclei parameterization relying on both temperature and aerosol content like DeMott et al. (2010) (not currently used in WRF cloud schemes) is in best agreement with the observations, at temperatures and aerosol concentration characteristic of the Antarctic Peninsula where the primary ice production occurs (part 1), compared to parameterization only relying on the atmospheric temperature (used by the WRF cloud schemes). Overall, a realistic double-moment ice microphysics implementation is needed for the correct representation of the supercooled liquid phase in Antarctic clouds. Moreover, a more realistic ice-nucleating particle alone is not enough to improve the cloud modelling, and water vapour and temperature biases also need to be further investigated and reduced.