Bayesian Cloud Property Retrievals from ARM Active and Passive Measurements

Principal Investigator(s):
Derek Posselt, JIFRESSE/UCLA

Co-Investigator(s):
Pavlos Kollias, Stony Brook University
Marcus van Lier-Walqui, Columbia University / GISS

Cloud processes remain a major source of uncertainty in regional and global modeling of the Earth's climate system. Cloud microphysics and vertical motion lie at the heart of this uncertainty, as they (1) govern the rates of conversion between phases of water, (2) modulate the amount of water transported vertically from the boundary layer to all levels of the troposphere, and (3) determine the properties of particles at cloud top and in the cloud interior. The availability of long-term estimates of microphysical properties (e.g., water content, characteristic size) and in-cloud vertical motion with well-characterized uncertainties is key to our efforts to evaluate and improve the representation of clouds and precipitation in numerical models. Recent work has demonstrated that microphysical retrievals based on Bayesian methodologies have promise for producing robust estimates of cloud particle size distribution properties, and for returning quantitative measures of uncertainty in the retrievals. These methods also yield an assessment of information contained in observations from one or more different measurement platforms, and can be used to establish measurement accuracy criteria. Over the last five years, the ARM program has invested in a comprehensive suite of active and passive remote sensing instruments and is poised to begin a systematic large eddy simulation (LES) modeling activity at the Atmospheric Radiation Measurement (ARM) Climate Research Facility Southern Great Plains (SGP) site. Effective use of updated sensors and measurements, and evaluation of LES output, will require development of a robust cloud property retrieval and uncertainty analysis framework suitable for application to multiple instruments, regions, and seasons.

This project will produce retrievals of cloud particle size distributions and in-cloud vertical velocity in shallow warm clouds, and will address the need for evaluation of current and future LES of shallow cumulus at the ARM SGP site. Retrievals will have robust quantitative estimates of uncertainty, and will utilize ARM scanning and vertically pointing radar, Raman and micropulse lidar, and microwave radiometer measurements. The methodology is fully Bayesian, and utilizes a Markov chain Monte Carlo (MCMC) algorithm to produce both an optimal estimate and robust estimates of uncertainties. MCMC is well proven for characterization of cloud properties and their uncertainties, but has not been used operationally on ARM data. As such, the retrieval framework will first be tested using cloud fields obtained from LES output over SGP and Eastern North Atlantic (ENA) sites. Following evaluation using synthetic data, the new algorithm will be used to produce several-month datasets of low cloud properties over SGP and ENA sites. While other cloud property retrievals from ARM instruments exist, this new product will be unique in the following ways: (1) it utilizes information from the new scanning ARM cloud radars and draws upon LES databases, (2) it leverages the latest set of instrument forward operators, and (3) it returns robust Bayesian estimates of information content and measurement uncertainty. Quantitative estimates of warm cloud microphysical properties will serve as the first product, and as a proof of concept for future retrievals in more complex mixed- and ice-phase clouds. The retrieval approach is general, in that it can be adapted to retrieve cloud microphysical properties consistent with any assumed particle size distribution shape and any commonly used bulk or bin microphysical parameterization.

The deliverables from the first set of experiments are the following: (1) establishment of observation requirements for shallow cumulus and information to guide development of LES, and (2) an open-source Bayesian algorithm suitable for application by the ARM community to quantification of uncertainty in a range of retrieval problems. The research we propose will directly address the following goals of the Atmospheric System Research program:

  1. Develop and evaluate an algorithm that retrieves cloud microphysics in shallow cumuli clouds using the next generation of measurements at the SGP and ENA ARM sites.
  2. Determine information content of current ARM observational systems, and assess their potential use in evaluation of large eddy, cloud resolving model, and global climate model simulations.
  3. Return robust uncertainty estimates for current suites of ARM measurement platforms.