The effect of temporal aggregation in models to estimate trends in sulfate deposition

This research investigates the effect of temporal aggregation in regression models used to measure long-term trends in the wet deposition of sulfate. I propose a set of generalized linear models that utilize precipitation and meteorological data collected on a variety of time scales. Specifically, I examine models that fit daily-level precipitation chemistry to daily-level meteorological covariates, weekly-level precipitation chemistry to weekly-level covariates, and weekly-level precipitation chemistry to daily-level covariates using historical data collected at daily monitoring sites, with artificial aggregation to create weekly-level data. Empirical results show that there can be small differences among the estimates of long-term trend in sulfate deposition under the three aggregation schemes, as well as a loss of precision with aggregation. Using a jackknifing procedure to obtain estimates of the standard errors of the differences in parameter estimates, I conclude that there is no significant difference in the estimation of long-term trends using weekly-level data. The estimates of the long-term trend using weekly data do, however, have consistently larger standard errors.