Spatiotemporal Interpolation and Constraint Databases for a GIS Application: Ozone in the Contiguous U.S.
Proceedings of the Dagstuhl-Seminar on Constraint Databases, Geometric Elimination and Geographic Information Systems
It is important to conduct research on the connection between air pollution such as ozone and human health on a large scale with respect to area and population. A GIS-based geostatistical approach has been used to do spatial interpolation for environmental exposure analysis in the work of (Liao et al. 2006). However the geostatistical analyst in ArcGIS could not handle the spatiotemporal interpolation, and computationally this approach is not efficient for large datasets. In our paper, using a set of spatiotemporal data with annual ozone concentration measurements in the contiguous U.S. during 1994 and 1999, we address the following challenging issues in conducting such research: spatiotemporal interpolation, comparison of spatiotemporal interpolation methods, data representation, visualization, and analysis of population exposure to ozone.
The shape function based spatiotemporal interpolation method has been used in this paper to estimate the ozone concentrations at any unmeasured location and time. Using the leave-one-out cross-validation, we compute error statistics to compare the shape function and IDW (Inverse Distance Weighting) methods. It is shown that the shape function method is better than IDW in terms of MAPE (Mean Absolute Percentage Error) and algorithm complexity. We illustrate how to use constraint databases to represent the interpolation results efficiently and accurately. For generating maps of annual ozone concentrations, we propose a new approach to select locations to interpolate and visualize: picking U.S. census block centroids as sample locations. The advantage of this approach is to generate more sample points in the areas with more intensive human activities.
In our experiment, there were about 8, 000, 000 sample points selected per year. Traditional GIS techniques are insufficient in handling such kind of spatiotemporal data. The visualization results of ozone concentration distribution at the census tract level in the contiguous U.S. from 1994 to 1999 are illustrated. We also analyze the population exposure to ozone in the year of 1999 according to diffierent ozone concentration levels following the recommendations given by the U.S. EPA on air quality. Our finding is that in the year of 1999, 9.8% total population in the contiguous U.S. has been exposed to a high risk ozone level, 78.7% to a moderate risk, and only 11.5% to a low risk.
Li, Lixin, Xingyou Zhang, Reinhard E. Piltner.
"Spatiotemporal Interpolation and Constraint Databases for a GIS Application: Ozone in the Contiguous U.S.."
Proceedings of the Dagstuhl-Seminar on Constraint Databases, Geometric Elimination and Geographic Information Systems: 5-6 Dagstuhl, Germany.