You are here:
Space-Time Fusion Under Error in Computer Model Output: An Application to Modeling Air Quality
Berrocal, V. J., A. E. Gelfand, AND D. M. HOLLAND. Space-Time Fusion Under Error in Computer Model Output: An Application to Modeling Air Quality. Biometrics. Department of Physics, The University of Texas at Arlington, Arlington, TX, 68(1):1-12, (2012).
In the last two decades a considerable amount of research effort has been devoted to modeling air quality with public health objectives. These objectives include regulatory activities such as setting standards along with assessing the relationship between exposure to air pollutants and adverse health outcomes. Conclusions on the latter are often contrasting, highlighting the importance of good data quality for both health outcomes and environmental exposure. In this paper, we provide methods that can be used to obtain higher quality exposure data. In particular, we propose two modeling approaches to combine monitoring data with numerical model output, addressing the difference in spatial scale between the two sources of data, yielding improved prediction of exposure at point level. Extending our earlier downscaler model (Berrocal et al., 2010b), these new models are intended to address two potential concerns with the model output. One is potential spatial displacement in the computer model values assigned to a grid cell. Possibly, this output is appropriate for a displacement of the grid cell. The second recognizes that, with regard to improving predictive performance of the fusion at a location, there may be useful information in the outputs for grid cells that are neighbors of the one in which the location lies. The first model is a Gaussian Markov random field smoothed downscaler that relates monitoring station data and computer model output via the introduction of a latent Gaussian Markov random field linked to both sources of data via, respectively, a spatial linear regression model and a measurement error model. The second model is a smoothed downscaler model with spatially varying random weights defined through a latent Gaussian process and an exponential kernel function, that allows us to smooth the computer model output and derive at each site a new variable on which the monitoring station data is regressed with a spatial linear model. We applied both methods to daily ozone concentration data for the Eastern US during the summer months of June, July and August 2001, obtaining a 10-15% predictive gain in predictive mean square error over our earlier downscaler model (Berrocal et al., 2010b).
Since the National Mortality and Morbidity Air Pollution Study (NMMAPS; Samet, Dominici, Zeger, Schwartz, and Dockery 2000; Samet, Zeger, Dominici, Curriero, Coursac, Dockery, Schwartz, and Zanobetti 2000; Daniels et al. 2000) was published, an increasing number of studies have been conducted to determine an association between exposure to air pollutants and health effects: as an illustration, it is just suffcient to observe that while in the 1991-2000 decade the number of studies linking exposure to air pollutants and health effects were about 1,900, in the 2001-2010 decade the number has raised to about 6,150. Also the suite of effects on human health outcomes investigated has become more diverse: while the NMMAPS study focused mostly on the effect of air pollution on risks of mortality and morbidity in general, in the most recent years, research efforts have become more specialistic and have focalized on investigating whether exposure to air pollutants is linked to increased risks of cardiovascular diseases (e.g. Pope et al. 2002; Dominici et al. 2006), respiratory complications (e.g. Dominici et al. 2006; Braga et al. 2001), and even adverse birth outcomes (e.g. Basu et al. 2004; Maisonet et al. 2004; Sram et al. 2005; Bell et al. 2007).
URLs/Downloads:Space-Time Fusion Under Error in Computer Model Output: An Application to Modeling Air Quality Exit
HOLLAND 10-079 FINAL JOURNAL ARTICLE MEM_PAPER_JUN24_BIOM_SHORTER.PDF (PDF,NA pp, 1174 KB, about PDF)
Record Details:Record Type: DOCUMENT (JOURNAL/PEER REVIEWED JOURNAL)
Organization:U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
NATIONAL EXPOSURE RESEARCH LAB
ENVIRONMENTAL SCIENCES DIVISION
LANDSCAPE CHARACTERIZATION BRANCH (RTP)