Data Fusion (Amy Braverman, organizer)

Brian J. Smith (University of Iowa)
Mary Kathryn Cowles (University of Iowa)
Fusing Point-Referenced Radon Data with Areal Uranium Data Arising from a Common Spatial Process

Saturday 10:30-11:00, Fountain III

Abstract:

Because exposure to radon gas in buildings is a likely risk factor for lung cancer, estimation of residential radon levels is an important public health endeavor. Radon originates from uranium, and therefore data on the geographical distribution of uranium in the earth's surface may inform about radon levels. We fit a Bayesian geostatistical model that appropriately combines areal data on uranium with measurements of indoor and outdoor radon in the state of Iowa, thereby obtaining more accurate and precise estimation of the geographic distribution of average residential radon levels than would be possible using radon data alone.


Linda J. Young (Department of Statistics, University of Florida)
Carol A. Gotway (National Center for Environmental Health, Centers for Disease Control and Prevention)
The Effects of Change of Support on Data Fusion

Saturday 11:00-11:30, Fountain III

Abstract:

The widespread availability of digital spatial data and the capabilities of geographic information systems make it possible to easily synthesize spatial data from a variety of sources. More often than not, data have been collected at different scales, and each of the scales may be different from the one of interest. Geographic information systems effortlessly handle these types of problems through raster and geoprocessing operations based on proportional allocation and centroid smoothing techniques. However, there are many statistical issues associated with combining such disparate data. We discuss the role that the support of spatial data plays in this process and how different linkage methods can lead to different inference.


Thomas Bengtsson (Bell Labs, Lucent Technologies)
Sample Based Data Fusion in Remote Sensing

Saturday 11:30-12:00, Fountain III

Abstract:

We explore sample based data fusion in the context of blending information collected from two different remote sensing platforms. The ensemble Kalman filter, which was recently advanced to assimilate (computer) model output and weather observations for numerical weather prediction, is explored as a tool to fuse data observed at different spatial and temporal resolutions. We take a Bayesian approach and specify the remote sensing data conditionally on the unobserved field of interest. To deal with real-time requirements, linear estimation theory is used to generate random samples of the latent field given the data. Several challenges arise within this approach, e.g.: specification of the spatio-temporal smoothness of the unobserved field, estimating error variances to determine the relative contribution of each platform, addressing issues of spatial support, and devising computationally feasible algorithms that are sequential both in time and space.