气候变化领域集成服务门户

The US Environmental Protection Agency is required to monitor, regulate, and set national ambient air quality standards for ozone. To investigate ozone exposure, the Environmental Protection Agency utilizes monitoring devices along with estimates of gridded ground level ozone concentration produced by a deterministic air quality model, the Community Multiscale Air Quality Model. These two sources of information enable inference regarding spatial exceedance of the National Ambient Air Quality Standard for ozone, which is given in terms of the level of the annual fourth highest ozone concentration.Here, we extend previous downscaling work to propose a spatial fourth highest extreme value downscaling model to assimilate annual fourth highest ozone concentration data at geo-coded locations with estimates at grid cell level derived from the Community Multiscale Air Quality Model model output. The resulting inference enables us to make probabilistic statements, with associated uncertainty, about the spatial variation in the chance of exceeding the standard. We apply our approach to data in the Eastern USA during years 2001-2008 and compare its predictive performance to that of downscaler models based on Gaussian processes applied to daily data. Copyright (c) 2014 John Wiley & Sons, Ltd.