Marine boundary layer clouds constantly cover about 25% of the ocean surface. They are a critical player in the climate system due to their strong radiative forcing and their interactions with aerosols and precipitation. However, simulating marine boundary layer clouds remain a great challenge in climate models and numerical weather prediction models. Of all of the difficulties, the variability of clouds at the scales smaller than the climate model grid, known as the sub-grid cloud variability, is the most outstanding and formidable problem. To account for sub-grid processes, clouds have to be parameterized using large-scale state variables based on physical or statistical relationships. These parameterizations are often severely constrained by both the quantity and quality of observational data. For example, in a widely used climate model, the Community Atmospheric Model (CAM5) release version 5, separate sub-grid schemes are developed for Planetary Boundary Layer, shallow and deep convection, and stratiform clouds. However, the transition of cloud regimes in nature is a gradual rather than abrupt process. The use of separated sub-grid cloud schemes results in several issues. An important example is that the current climate models often struggle to simulate the transition of stratocumulus to cumulus cloud regimes over the subtropical ocean. This limited capability of marine boundary layer clouds simulation is a major source of uncertainty in future climate change prediction.

Recently, tremendous efforts have been made to improve the representation of sub-grid cloud processes in climate models. A couple of innovative sub-grid cloud parameterization schemes have been developed and implemented recently at PNNL. This recent progress has given rise to a compelling need for a systematic evaluation of cloud simulations in climate models against observations. The proposed work is motivated by this need and features a comprehensive evaluation of cloud properties simulated in the state-of-the-art climate models using a combination of satellite observations from NASA and ground-based observations from the DOE Atmospheric Radiation Measurement Mobile Facility deployed on the northern coast of Graciosa Island in the Azores. Specific goals include:

1. The first phase is a global evaluation based on satellite observations and retrievals. Outcomes from this investigation will provide us a primary understanding of the overall skill and outstanding problems of the climate models based on different sub-grid schemes for cloud simulation.
2. The second phase is a regional evaluation based mainly on the ARM Azores ground-based observations and retrievals. Outcomes from this investigation will help to confirm the findings from the global evaluation and shed a light for improving regional climate simulations. 
3. The third phase will be focused on the evaluation of marine boundary layer clouds microphysical properties and warm rain processes using both satellite and ground-based retrievals over the ARM Azores site.

The three versions of CAM5 that will be used in this study include: 1)  the commonly available CAM5 (which is referred to as CAM-Base) in which clouds are divided into two categories- stratus and cumulus; 2) CAM5 that incorporates a new sub-grid parameterization scheme called CLUBB (CLUBB stands for Cloud Layers Unified By Binormals) that features a unified physics for shallow cumulus and stratiform clouds; and 3) Super-Parameterized CAM5 (SP-CAM5) that consists of a three-dimensional cloud-resolving model at higher resolution (4km) embedded into each grid of a conventional CAM.

Several aspects distinguish the proposed study from previous work: 1) The use both satellite and ground based observations. 2) Previous studies have largely been focused on the evaluation of bulk cloud properties, while this study will evaluate both bulk cloud properties and the microphysical properties of marine boundary layer clouds and the warm rain process. 3) In previous studies, cloud observations from different instruments are usually used separately and independently in model evaluation. In contrast, it is proposed here to use measurements from passive and active sensors in an integrated manner, which will help identify the root of identified problems in the model. The proposed work will help improve cloud simulations in PNNL’s climate models and lead to a more reliable predication of future climate change.