Marine Boundary Layer Cloud and Drizzle Properties from Ground-Based Observations over the Azores

Abstract

Marine boundary layer (MBL) clouds cover vast areas of the subtropical and midlatitude oceans and impose strong cooling effects onto the underlying surface, making them a key component in earth’s radiation budget. It is, however, challenging to simulate MBL clouds realistically in global climate models (GCMs). GCMs disagree substantially in the signs and magnitudes of cloud feedback for the regimes of subtropical MBL clouds, and suffer from the so-called “too few, too bright” problem. In addition, GCMs simulate too frequent but too light MBL precipitation compared with observations. To understand MBL cloud and drizzle processes, the Clouds, Aerosol, and Precipitation in the Marine Boundary Layer (CAP-MBL) field campaign was conducted at the Graciosa Island over the Azores from April 2009 to September 2010, which was later established as a fixed site since September 2013 by the US Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) program. There are three objectives for this dissertation. First, innovative methods are developed to simultaneously retrieve cloud and drizzle microphysical property profiles using a combination of ARM ground-based cloud radar, ceilometer, and microwave radiometer observations. These retrieved microphysical properties, such as cloud and drizzle particle size (rc and rd), their number concentration (Nc and Nd) and liquid water content (LWCc and LWCd), are validated with collocated aircraft in-situ measurements and the retrievals agree well with in-situ measurements both in timeseries and in vertical variations. Treating the aircraft in-situ measurements as truth, the estimated median retrieval errors are ~15% for rc, ~35% for Nc, ~30% for LWCc, and rd, and ~50% for Nd and LWCd, which are compariable with existing retrieval techniques in literature. The retrieval methods are then applied to all available ARM ENA observations with more than four years of data record. The derived climatology shows annual mean drizzle frequency in single-layered MBL clouds of ~55%, with ~70% in winter and ~45% in summer. The rd retrievals are 3-6 times greater than their rc values, and Nd and LWCd are three and one order(s) of magnitude lower than their cloud counterparts. The rc and LWCc increase from the cloud base to 75% of cloud layer by condensational growth, while the rd increases from the cloud top downwards the cloud base from the collision-coalescence process. LWCc in non-drizzling clouds can be greater than those in drizzling clouds, especially near the cloud top, where drizzle drops usually form. To investigate the impact of environmental forcing on drizzle formation, the drizzling MBL clouds are classified into two types. The type I clouds can last for more than five hours before intense drizzle occur while the type II clouds are characterized by mesoscale convection cellular (MCC) structures with drizzle occur every two to four hours. By analyzing the wind profiles (direction and speed) as well as boundary layer stability, it is found that either directional or speed shear is required to promote drizzle production in the type I clouds while the boundary layer instability is the main driver in drizzle formation in the type II clouds. Both the wind shear and boundary layer instability can generate turbulence and enhance the collision-coalescence process between cloud droplets, and/or drizzle drops, resulting in rapid drizzle formation. In the last part of this dissertation, the ground-based observations and retrievals are used to evaluate the so-called autoconversion and accretion enhancement factors (Eauto and Eaccr) in GCM warm-rain parameterizations. Eauto and Eaccr are used to account for sub-grid variations in cloud and drizzle properties but are often prescribed in GCMs. Comparing the Eauto and Eaccr values calculated from surface retrievals and the prescribed ones in a representative microphysics scheme, we find that a smaller Eauto and larger Eaccr should be used in GCMs to simulate reasonable precipitation frequency and intensity. This result partially explains the “too frequent too light” problem in GCM precipitation simulations.