Using Doppler lidar and radar to evaluate the representation of drizzle in models

Both climate and operational forecasting models have difficulty representing the drizzle process which is crucial in simulating stratocumulus in models and predicting their persistence and break-up. Here, we use vertically-pointing Doppler lidar and Doppler cloud radar to derive drizzle properties (including size, number concentration, liquid water flux) and the vertical air velocity. Case studies (e.g. O'Connor et al., 2005) have shown that drizzle in stratocumulus appears to be preferentially produced in the updrafts. We extend this study to >1 year's observations of low level liquid water clouds and examine the implications for the parametrization of drizzle in models.

For instance, a model is likely to assume a grid box mean vertical air velocity close to zero and may also assume that any precipitation is falling with respect to the grid box mean vertical air velocity. The observed vertical air velocity averaged on timecales that correspond to the model grid spacing will also likely be close to zero. If the drizzle is falling only in the updrafts then the apparent fall velocity is reduced. This difference (which can be interpreted as a bias) means that, all else being equal, the same drizzle falls to the ground faster in the model and would not be subject to as much evaporation as observed.

In this presentation we will show the magnitude/pdf of this bias, the impact that this could have on processes such as evaporation, whether the bias is significant enough that models should take account of it, and a possible parametrization that models can apply to correct for this effect.