Cloud and precipitation particles vary in size from a few microns to a few millimeters or larger. To accurately quantify these sizes from space-borne instruments, observations from multiple…

This National Aeronautics and Space Administration (NASA) grant awarded to Morgan State University, a Historically Black College and University (HBCU), provides $159,428 to develop and apply an Expectation Maximization (EM) framework that leverages an Ensemble Smoother (ES) technique. This research aims to better understand and mitigate uncertainty in satellite-based estimates of cloud and precipitation properties by minimizing the impact of a priori assumptions on the final estimates. The EM framework will incorporate accurate physical models to simulate observations and iteratively adjust 3D atmospheric fields to maximize agreement with actual observations. This work supports NASA's broader efforts in atmospheric science, climate modeling, and remote sensing.

Cloud and precipitation particles vary in size from a few microns to a few millimeters or larger. To accurately quantify these sizes from space-borne instruments, observations from multiple instruments (e.g. radar and radiometers) are usually necessary. However, even when observations from multiple instruments are available, ambiguities in their interpretation still exist, and, to prevent the derivation of arbitrary, highly erroneous estimates, methodologies to constrain the estimates to prescribed statistical properties are necessary. While the incorporation of regularization methodologies in the estimation process significantly reduces uncertainties in the estimates, IT also makes IT difficult to determine whether the relationships resulting among the estimated properties are real or just a consequence of prescribing structures to the estimates. For example, cloud water is a main ingredient in several precipitation processes. To better understand these processes, we need global cloud water estimates coincident with liquid and solid precipitation estimates. However, the extent to which current estimates of these variables from satellite observations are driven by unambiguous signal in the observations and not a consequence of the regularization process is generally ignored in analyses and insufficiently understood. To better understand and mitigate uncertainty in satellite estimates of cloud and precipitation, properties we propose the development and application of an expectation maximization (EM) framework that minimizes the impact of a priori assumptions on the final estimates. The EM framework incorporates an ensemble smoother (ES) technique that derives cloud and precipitation estimates from multiple instrument observations. The ES technique relies on accurate physical models that simulate observations as a function of 3D atmospheric fields generated by a stochastic model. An iterative gauss-newton procedure is used to maximize the agreement between the simulated and actual observations. This is achieved by adjusting the stochastically generated 3D atmospheric fields within their prescribed uncertainties.