Evaluation of ozone exposure indices in exposure-response modeling

In exposure-response modeling, a major concern is the numerical definition of exposure in relating crop loss to O3, yet few indices have been considered. This paper addresses research in which plant growth was regressed for soybean, wheat, cotton, corn, and sorghum against 613 numerical exposure indices using the Box-Tidwell model. When the minimum sum of squared errors criterion was used, optimum performance was not attained for any single index; however, near optimum performances were achieved by two censored cumulative indices and from a class of indices called the generalized, phenologically weighted, cumulative impact indices (GPWCIs). The top-performing GPWCIs accumulated concentrations, used sigmoid weighting schemes emphasizing O3 concentrations of 0.06 ppm (118 microg m(-3)) or higher, and had phenological weighting schemes with greatest weight occurring 20 to 40 days prior to crop maturity. These findings indicate that (1) peak concentrations are important, but lower concentrations should be included in the calculations, (2) increased plant sensitivity occurs between flowering and maturity, and (3) plants respond to cumulative exposure impact.