Risk qualified maps of hydrogen ion concentration for the New York state area for 1966–1978

Abstract

The U.S. Geological Survey has collected and analyzed bulk precipitation samples at 22 sites in New York State from 1965 to 1979. Without the benefit of any spatial analysis, these data have previously been used with data from other networks to produce H + concentration contour (or point) maps for the eastern United States. In this study, a 240 × 320 km region in the center of the USGS New York network was divided into twelve 80km 2 blocks. For each year in the period 1966–1978, point and 80km 2 block estimates of average H + concentrations and the corresponding estimation standard deviations were calculated using the best linear unbiased estimator known as the ‘kriging’ estimator. The average H + concentration for the period was 0.042 meqℓ −1 (or 4.38 pH units). The largest difference between blocks within a given year was 0.026 meq ℓ −1 and occurred in 1974. Across years, the largest difference between blocks was 0.037 meq ℓ −1. Block estimation standard deviations (which measure the error in each block estimate) ranged from 0.005 to 0.015 meq ℓ −1, but overall the average was about 0.009 meq ℓ −1. The appropriate standard error calculation indicates that the block maps do show some evidence of differences in H + concentration within the region. A weak tendency for H + concentrations to decrease from west to east was evident for most years. For point estimates of H + concentration, the average increase in estimation standard deviation compared to block estimation standard deviations was almost 60%. Since no significant change in differences between point estimates occurred, the conclusion is that yearly point maps based on the USGS data cannot be used to support spatial differences in H + concentration over New York State. Finally, the influence of sample density on estimation standard deviations was investigated. The improvement due to a regular sampling grid (one monitor at the center of each of the 12 blocks) compared to the existing network would be an average decrease in estimation standard deviations of over 30%.