Multi-scale Probabilistic Modeling in Geospace Science

08/01/2014 - 9:05am
Mesa Lab Main Seminar Room

Zachary Thomas portrait

Zachary Thomas, SIParCS Intern
(The Ohio State University)

The complex physical processes which govern large and small-scale variability within Earth's ionosphere are of great scientific interest. Motivated by the need for statistical methodologies with which to make inference regarding these processes (from sparse and otherwise incomplete observations), we present a general technique for spatial modeling on the sphere. Via construction of a Markov random field model over a multiresolution geodesic grid, the procedure is an adaptation of existing multiresolution basis function expansion approaches for modeling on the plane. Predictive skill is studied using numerical model output of electric potential fields over large polar regions. We present further modifications of the procedure which are of use in making inference regarding unobservable ionospheric processes from incomplete measurements of related processes (measured by various radar and satellite instruments).  

Video Presentation