Environmental Engineering Reference
In-Depth Information
2.12 Application of Model and Ambient Data Fusion
Techniques to Predict Current and Future Year PM2.5
Concentrations in Unmonitored Areas
Brian Timin, Karen Wesson, and James Thurman
U.S. Environmental Protection Agency
1. Introduction
States with nonattainment areas for ozone and/or PM
2.5
must attain the ambient air
quality standards by their respective attainment dates. The majority of areas use
photochemical models to estimate future year pollutant concentrations to show
how they will meet and/or maintain the standards. The EPA photochemical
modeling guidance (U.S. EPA 2007) recommends an ambient monitor based rela-
tive attainment test. Since the test examines future year concentrations at monitor
locations, it does not evaluate the likelihood of violations in unmonitored areas.
To account for unmonitored areas, the modeling guidance also recommends an
“unmonitored areas analysis” to examine future year concentrations of ozone
and/or PM
2.5
in locations without nearby monitors. For the unmonitored area
analysis, the guidance recommends using data fusion techniques to combine
spatially interpolated ambient data with gridded photochemical model output to
produce a “fused” surface of spatial fields. These fused spatial fields can be pro-
duced for both current years and also combined with future year modeling output
to produce future year fields. Because they utilize both ambient data and model
outputs, the fused spatial fields should provide an improved concentration estimate
in unmonitored areas. In this application, we attempt to examine the performance
of fused spatial fields compared to a simple ambient data interpolation technique.
2. Methods
The focus of this analysis is on annual average PM2.5 predictions. Using EPA's
Model Attainment Test Software (MATS) (Abt, 2009) we created several sets of
gridded spatial fields for a domain which covers the Eastern ~2/3 of the United
States. The domain consists of 279 × 240 grid cells, 12 km on a side. In all cases,
concentrations were interpolated to the center of each grid cell. The “basic” inter-
polation technique used for all interpolations was Voronoi Neighbor Averaging
(VNA). This technique identifies the set of monitors that are nearest to the center
