A comparison of statistical techniques for combining modeled and observed concentrations to create high-resolution...

Air quality surfaces representing pollutant concentrations across space and time are needed for many applications, including tracking trends and relating air quality to human and ecosystem health. The spatial and temporal characteristics of these surfaces may reveal new information about the associations between emissions, pollution levels, and human exposure and health outcomes that may not have been discernable before. This paper presents four techniques, ranging from simple to complex, to statistically combine observed and modeled daily maximum 8-hr ozone concentrations for a domain covering the greater New York State area for the summer of 2001. Cross-validation results indicate that, for the domain and time period studied, the simpler techniques (additive and multiplicative bias adjustment) perform as well as or better than the more complex techniques. However, the spatial analyses of the resulting ozone concentration surfaces revealed some problems with these simpler techniques in limited areas where the model exhibits difficulty in simulating the complex features such as those observed in the New York City area.