Long-term Arctic ground-based observations for Global Climate Model evaluation: Liquid phase occurrence within the context of cloud vertical structure

Description

The presence of supercooled liquid in clouds affects surface radiative and hydrological budgets, especially at high latitudes. Capturing these effects is crucial to properly quantifying climate sensitivity. Observationally locating and quantifying supercooled liquid is challenging and so is properly representing it in a general circulation model (GCM). Improvements could emerge from considering cloud phase in the context of cloud type. In the current study, ground-based observations from 2011 to 2016 are used to evaluate the GISS ModelE GCM over the ARM North Slope of Alaska (NSA) site. Vertically pointing lidar and mm-radar observations are combined to identify hydrometeor location and phase using the Shupe (2007) technique. Thresholds are applied to ModelE hydrometeor mixing ratios to produce comparable masks. Both data sets are compared through the Cloud Vertical Structure (CVS) approach, which roughly categorizes cloud layering using two distinct pressure limits (530 hPa, 790 hPa). CVS are also used to resample this high-resolution observational data sets to produce a scale-aware comparison with the coarse GCM output. Resampling of the observational data set is done by only considering the observations associate with the most frequent CVS over 30-min durations (with the exception of clear sky, which requires an entire 30-min period without hydrometeor detection). The NSA site is rarely hydrometeor-free over 30-min time periods (5% during the winter season). Single-layer, shallow boundary-layer clouds are the dominant CVS type (20-40%) with large occurrence during the transition seasons. While ModelE produces significant fractions of these clouds (10-30%), it mostly produces deep hydrometeor layers (~25%), which are rather infrequently observed (~5%). Liquid-only clouds are shallow and confined to the boundary layer with no strong seasonal cycle. 80% of them are single layer while the remaining are multi-layer, a behavior that is rather well captured by ModelE. For the most part, mixed-phase systems are single layer and present in the low levels. Both observations and simulations show that 15% of mixed-phase conditions are multi-layer. ModelE tends to produce too many mixed-phase deep systems at the expense of mixed-phase single-layer shallow boundary-layer clouds. This insight points to a need to reevaluate deep cloud system representation in ModelE. A more in-depth analysis should help determine if model phase occurrence errors are primarily errors