Ensemble-Based Characterization of Uncertain Environmental Features

• Activity Number: 404
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Seyed Hamed Alemohammad*+ and Dennis B. McLaughlin and Dara Entekhabi
• Companies: Massachusetts Institute of Technology and MIT and Massachusetts Institute of Technology
• Keywords: Ensemble Estimation ; Importance Sampling ; Data Assimilation ; Image Fusion

Abstract

This paper considers characterization of uncertain spatial environmental features from noisy measurements. Unlike previous studies, this method derives the error likelihood using an archive of historical measurements and provides an ensemble characterization of measurement error. The characterization process is formulated as a Bayesian sampling problem and solved with a non-parametric version of importance sampling. A novel dimensionality reduction scheme using principle component analysis is implemented to describe images in a problem-specific low-dimensional attribute space, and the importance sampling operations are formulated entirely in terms of the attribute vectors. As an example, we characterize satellite measurements of rainfall over the Earth surface and assess the performance of the method using ground truth measurements from surface weather radar. Results indicate that our ensemble estimation approach is able to provide an improved description of rainfall features by giving a posterior ensemble that is narrower than the prior. This procedure is capable of fusing static images from different noisy sensors as well.