Investigation of drizzle onset in liquid clouds using ground based active and passive remote sensing instruments

One of the major challenges of climate prediction is a correct representation of the interactions among aerosols, clouds and precipitation. Aerosols have a strong impact on the life cycle of boundary layer clouds, which are known to significantly influence the energy available to the Earth-Atmosphere system. Specifically, drizzle formation in low-level clouds, which has been shown to depend on aerosol concentration (second indirect aerosol effect), determines cloud life time. In models, the transition from liquid cloud to precipitation must be parameterized by the so-called autoconversion process. Different parameterizations of autoconversion have been developed, whereby the corresponding transition rates differ of up to one order of magnitude. Even observations of this microphysical process are very challenging. Satellite observations have been exploited in the past to evaluate different autoconversion schemes but one of the main reasons for the encountered differences between models and observations was the poor representation of the vertical cloud structure in the satellite observations. In this context, ground-based cloud observations present a unique tool to provide observational constraints for model parameterization development by exploiting their highly temporally and spatially resolved profiling capability. In recent years, new ground-based techniques exploiting higher moments of the cloud radar Doppler spectrum (the skewness, in particular) have been successfully applied for the detection of drizzle onset in maritime clouds. In this thesis, a new, extended ground-based dataset for continental liquid clouds is exploited in order to assess the potential for early drizzle detection. For this purpose, ground-based observations of liquid water path and of the cloud radar Doppler moments reflectivity, mean Doppler velocity, spectral width and skewness have been synergetically exploited. It has been found that skewness detects drizzle formation at an earlier stage than the other radar moments. The different observational variables have been used for the development of a drizzle probability index (DI) to improve currently available drizzle classification schemes, i.e. Cloudnet. The DI represents the probability of each cloud radar bin to contain drizzle. In comparison to the Cloudnet classification, case studies show that the DI detects earlier stages of drizzle formation and eliminates falsely detected, inconsistent time-height drizzle structures. However, due to the presence of turbulence, the DI sometimes falsely attribute drizzle to a pixel. In order to understand how turbulence can impact radar Doppler measurements and also in order to optimize the radar measurement settings for the purpose of drizzle detection, sensitivity studies on integration time, spectral resolution and radar antenna beam width have been conducted using raw radar data and a forward radar simulator. It has been found that integration times no longer than 2 seconds should be used for drizzle detection and that the spectral resolution obtained with the fast Fourier transform (FFT) using 256 FFT points resolves the characteristics of the Doppler spectrum with sufficient accuracy. Also, simulations showed that smaller beam widths are beneficial for drizzle detection and that turbulence is responsible for an increase of spectral width and a reduction of observed skewness values. Finally, a microphysical interpretation of the skewness signal is provided by comparing the simulations of drizzle formation from a 1D steady-state binned microphysical model to observations. The forward simulated vertical profiles of skewness based on the modeled cloud drop and drizzle size distributions strongly depend on the applied autoconversion parameterization. A validation of the different schemes indicates that the scheme from Seifert et al., (2010) best matches the observations of reflectivity and skewness. The comparison also suggests that the modeled autoconversion rates tend to produce large drizzle too fast and too early for continental liquid clouds. This first model comparison thus demonstrates that ground-based cloud radar observations, particularly skewness, can be used for testing autoconversion parameterizations. The dataset and the results of this work constitute a unique basis for evaluating model outputs, e.g. in a next step the results of large eddy simulations, and for carrying out additional process studies to refine for example the drizzle detection criterion. Also, this data set could be exploited for future validations of satellite products, e.g. of EarthCARE. This thesis hence shows how ground-based cloud radar observations can be optimally exploited to better understand the autoconversion process and also represents an important step forward in bringing observations of drizzle and modeling together