Effects of missing seasonal data on estimates of period means of dry and wet deposition

The current study uses resampling to investigate the impacts of cyclic seasonal behavior on 1- and 5-year period means composed from seasonal mean values in the presence of missing data. This is an empirical study using complete years of seasonal monitoring data collected in the eastern US and extracted from the clean air status and trends network (CASTNET) dry and the National Atmospheric Deposition Program/National Trends Network (NADP/NTN) wet deposition data archives. Estimators of period means with missing seasonal data are determined using means of the non-missing values as estimates of the missing data. Estimates are evaluated in terms of 95% inclusion intervals (e.g., estimates are within ±X% of the true value ⩾95% of the time). For dry deposition, missing transition seasons (i.e., spring or fall) usually yield estimates of annual means that are within ±20% of the true annual mean ⩾95% of the time. Missing summers or winters usually have larger impacts on estimates of annual means of dry deposited species than missing transition seasons. A missing summer has the largest impact on estimates of annual means of dry deposition for all constituents, except SO₂, where winter is especially important. For wet deposition, a missing season yields estimates of annual means that are within ±30% of the true annual mean ⩾95% of the time. A missing summer has the largest impact on estimates of annual means of wet deposition for all constituents, except NH₄⁺, where spring and fall are important. A strategy requiring at least 3 years of seasonal representation for three seasons with the fourth season having at least two seasonal values, yields estimates of wet deposition that are within ±17% of the true 5-year means ⩾95% of the time for all species. Corresponding confidence statements for dry deposition results are considerably stronger, with estimates that are within ±10% of the true 5-year mean ⩾95% of the time.