Impact of Precipitation on Retrieved Warm Cloud Properties Using Visible and Near-infrared Reflectances Using Markov Chain Monte Carlo Techniques

An effective means to estimate the cloud effective radius(r_e) and optical thickness in clouds from spaceborne observations is by combining reflectances at visible and near-infrared wavelengths, which is often referred to as the bispectral method (BSM). In spite of the wide use of such methods, the credibility of the BSM is subject to various factors such as the vertical and horizontal heterogeneity of target clouds, solar zenith angle, viewing geometry and the presence of precipitation. A body of evidence derived by comparing BSM results with in situ data suggests that r_e retrieved from the BSM (e.g., from the Moderate Resolution Imaging Spectroradiometer; MODIS) is often overestimated.

This study aims at quantitatively assessing the bias and uncertainties in the BSM that are induced by the presence of drizzle and rain in marine warm-topped clouds. Based upon a Bayesian Markov Chain Monte Carlo (MCMC) methodology developed in Xu et al. (2019), profiles of cloud and rain are allowed to be derived simultaneously from radiance measurements with associated uncertainties estimated. Provided an identical set of mid-wavelength visible and near-IR reflectances, two types of MCMC experiments are performed, in which precipitation is respectively assumed to be absent or present. This procedure is conducted for multiple sets of visible and near-IR reflectances with fixed measurement uncertainty.

In the presence of rain, the BSM-derived r_e tends to be biased high, and overall retrieval uncertainties are enhanced over scenes without rain, where the BSM results compare favorably. r_e appears more likely than optical thickness to be affected by the existence of precipitation. The magnitude of influence is found to depend on where in the BSM measurement space the cloud exists. When r_e is about 20 mm, the addition of a rain droplet mode introduces an approximately 25% increase in the scaled interquartile range (IQR; a measure of retrieval uncertainty) of retrieved r_e at cloud top. The effect on IQR becomes marginal as the cloud droplet size decreases to 8 mm. Also, there has often been an implicit assumption that the bias in r_e seen in BSM data provides information on rain properties. We find this to be unfounded, since the retrieved rain microphysics derive primarily from the prior estimate, due to the lack of information from the observations.