Date of Award

Spring 1-1-2019

Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Albin J. Gasiewski

Second Advisor

Dejan Filipovic

Third Advisor

Julie Lundquist

Fourth Advisor

Alexander Voronovich

Fifth Advisor

Kwo-Sen Kuo


A horizontally inhomogeneous unified microwave radiative transfer (HI-UMRT) model incorporating aspherical frozen hydrometeors based on the NASA/GSFC OpenSSP database is presented to study 3-dimensional (3D) effects of horizontal inhomogeneous clouds on computed microwave radiances and facilitate satellite radiance assimilation over horizontally inhomogeneous weather conditions. HI-UMRT provides a coupled two-Stokes parameter numerical radiance solution of the 3D radiative transfer equation by embedding the existing 1D UMRT algorithm into an iterative perturbation scheme. The horizontal derivatives in radiances of lower perturbation order are treated as the source functions of the azimuthal harmonic perturbation radiative transfer equations that are readily solved by the planar-stratified 1D UMRT algorithm.

The 1D UMRT algorithm requires symmetry of the transition matrix for the discretized planar-stratified radiative transfer equation to realize numerically stable and accurate matrix operations as required by the discrete-ordinate eigenanalysis method. In this thesis, the necessary block-diagonal structure of the full Stokes matrix for randomly oriented OpenSSP aspherical hydrometeors is shown to be maintained, albeit with small asymmetric deviations which introduce small asymmetric components into the transition matrix that are negligible for most passive microwave remote sensing applications. An upper bound of the brightness temperature error calculated by neglecting the asymmetric components of the transition matrix under even extreme atmospheric conditions is shown to be small. Hence the OpenSSP hydrometeor database can be reliably used within the UMRT model.

Block-diagonal Stokes matrix elements along with other single-scattering parameters of OpenSSP hydrometeors were subsequently used in radiative simulations of multi-stream dual-polarization radiances for a simulated hurricane event to demonstrate the inherent numerical stability and utility of the extended 1D UMRT algorithm. An intercomparison of computed upwelling radiances for a multiphase distribution of aspherical OpenSSP hydrometeors versus a mass-equivalent Mie hydrometeor polydispersion for key sensing frequencies from 10 to 874 GHz shows the considerable impact of complex (versus simple spherical) hydrometeors on predicted microwave radiances.

Further, a numerical performance assessment shows that the increase in computing time for the 3D HI-UMRT model relative to the 1D UMRT model is moderate since (i) the computationally efficient UMRT engine is applied only to the perturbation equations with non-trivial solutions, and (ii) the layer parameters for the 1D solution are reused for all higher perturbation orders. Numerical simulations using HI-UMRT based on 3D cloud profiles simulated by the WRF numerical weather model illustrate the convergence of the iterative perturbation series. An intercomparison of top-of-atmosphere brightness temperature images for HI-UMRT versus the planar-stratified UMRT model illustrates the considerable impact of cloud horizontal inhomogeneities on computed upwelling microwave radiances.

The microwave radiances simulated using UMRT at 118 and 183 GHz based on the Orbital Micro Systems Inc. Global Earth Monitoring System (GEMS) CubeSat constellation concept have been used in an all-weather microwave data assimilation scheme to facilitate precipitation locking of hydrometeor state variables in severe weather. The capability of first frame precipitation locking can be achieved based on constrained extended Kalman filtering (XKF), statistical estimation of a flow-dependent background error covariance matrix, and appropriate update of state variables using nonlinear iterative method. Preliminary simulation results demonstrate the potential for assimilating both thermodynamic and hydrometeor variables in first-frame locking iterations.