Duane Lee Rosenberg
Duane Lee Rosenberg
Research Associate III
Cooperative Institute for Research in the Atmosphere (CIRA)
CIRA/NOAA, R/GSD
325 Broadway
Boulder
CO
80305
Email
Phone
Fields of interest
Computational gas- and fluid-dynamics, nonlinear dynamics, turbulence, atmospheric & ocean dynamics, data assimilation, fundamental machine learning.
Description of scientific projects
Following is a list of possible projects from which candidate might choose:
(1) Examination of connection between scale interactions to Krylov space representations: Implement 2-level optimized and/or multigrid preconditioners for spectral element-based high-order discretization of Poisson equation in CFD and compare in the context of a turbulent flow solver. What is the connection between the Krylov subspace and the Fourier modes in fully developed turbulence (FDT)? Can this connection be exploited to develop a physics-based preconditioner for turbulent flows?
(2) Validation of shallow and deep convection parameterizations in global weather forecasting models: compare how shallow and deep convective parameterizations in NOAA models compare with large-eddy simulation (LES) and direct numerical simulation (DNS) solutions in fully unstable convective flow and stably stratified flow. Examine detailed single- and multi-point turbulent statistics in addition to more traditional anomaly correlations to seek improvements in the parameterizations.
(3) Pattern recognition for gravity waves from observations using machine learning (ML): Utilize satellite and in-situ observations for gravity wave propagation and breaking to train a ML algorithms to 'recognize' gravity wave effects in observational data for use in data assimilation model state nudging.
(4) Machine learning approaches to subgrid scale modeling: Find a proper coupling of the resolved to unresolved regions of the dynamical equations (e.g., Reynolds stresses or other closures) and train a ML algorithm with this either in steady state or in a decay scenario with the subgrid solutions. Validate by using ML-based subgrid model in a coarse resolution simulation and comparing with the comparable DNS.
(5) Validate fine-scale behaviour of low- and high-order truncation geophysical fluid dynamics (GFD) numerics: Examine detailed turbulence statistics to quantify effects of numerical truncation using a classical finite volume scheme and a high order continuous or discontinuous Galerkin scheme on grids of relevance in the weather forecasting and climate domains.
(1) Examination of connection between scale interactions to Krylov space representations: Implement 2-level optimized and/or multigrid preconditioners for spectral element-based high-order discretization of Poisson equation in CFD and compare in the context of a turbulent flow solver. What is the connection between the Krylov subspace and the Fourier modes in fully developed turbulence (FDT)? Can this connection be exploited to develop a physics-based preconditioner for turbulent flows?
(2) Validation of shallow and deep convection parameterizations in global weather forecasting models: compare how shallow and deep convective parameterizations in NOAA models compare with large-eddy simulation (LES) and direct numerical simulation (DNS) solutions in fully unstable convective flow and stably stratified flow. Examine detailed single- and multi-point turbulent statistics in addition to more traditional anomaly correlations to seek improvements in the parameterizations.
(3) Pattern recognition for gravity waves from observations using machine learning (ML): Utilize satellite and in-situ observations for gravity wave propagation and breaking to train a ML algorithms to 'recognize' gravity wave effects in observational data for use in data assimilation model state nudging.
(4) Machine learning approaches to subgrid scale modeling: Find a proper coupling of the resolved to unresolved regions of the dynamical equations (e.g., Reynolds stresses or other closures) and train a ML algorithm with this either in steady state or in a decay scenario with the subgrid solutions. Validate by using ML-based subgrid model in a coarse resolution simulation and comparing with the comparable DNS.
(5) Validate fine-scale behaviour of low- and high-order truncation geophysical fluid dynamics (GFD) numerics: Examine detailed turbulence statistics to quantify effects of numerical truncation using a classical finite volume scheme and a high order continuous or discontinuous Galerkin scheme on grids of relevance in the weather forecasting and climate domains.