Trend analysis of weekly acid rain data, 1978-83

Abstract

There are 19 stations in the National Atmospheric Deposition Program which operated over the period 1978-83 and were subsequently incorporated into the National Trends Network in 1983. The precipitation chemistry data for these stations for this period were analyzed for trend, spatial correlation, seasonality, and relationship to precipitation volume. The intent of the analysis was to provide insights on the sources of variation in precipitation chemistry and to attempt to ascertain what statistical procedures may be most useful for ongoing analysis of the National Trends Network data. The Seasonal Kendall test was used for detection of trends in raw concentrations of dissolved constituents, pH and specific conductance, and residuals of these parameters from regression analysis. Forty-one percent of the trends detected in the raw concentrations were downtrends, 4 percent were uptrends, and 55 percent showed no trends at a = 0.2. At a more restrictive significance level of a = 0.05, 24 percent of the trends detected were downtrends, 2 percent were uptrends, and 74 percent showed no trends. The two constituents of greatest interest in terms of human generated emissions and environmental effects, sulfate and nitrate, showed only downtrends, and sulfate showed the largest decreases in concentration per year of all the ions tested. The relationship between dissolved concentration and precipitation volume was expressed as: ln(C) = a + b * ln(P) and residuals from regression analysis of the data with this model were also tested for trends. Of the 133 tests for trends in residuals, 49 percent showed downtrends, 7 percent showed uptrends and 44 percent showed no detectable trends at a = 0.2. For a = 0.05, 38 percent of the tests showed downtrends, 2 percent showed uptrends and 60 percent showed no detectable trend. Further examination of the residuals showed evidence of a seasonal pattern in the data for most constituents from most stations. For example, sulfate and nitrate residuals generally were higher in the summer, and chloride residuals generally were higher in the winter. Characteristics from the regression model above and the same regression model with a seasonal term added were compared among all of the stations for all parameters, and no evidence of geographical patterns was detected. Smoothed residual series of sodium over time for all 19 stations showed a decline in 1980, some of which were very large. For example, the concentrations at Wellston, Michigan, at the beginning of 1980 were about 7.4 times the concentrations at the end of 1980. The simultaneity of the changes over the Nation shown by these plots suggests the possibility that some trends may have been the result of changes in procedures or materials used in sample collection, shipment or analysis. Five of the 19 stations were located in North Carolina which provided an opportunity to explore the spatial relationship of the records and a few of the possibilities for spatial analysis. The cross-correlation coefficients for the sulfate residuals from the seasonal model for all possible pairs of records and the interstation distance showed a strong relationship (R^ = 0.86). The Seasonal Kendall test was applied to all five stations simultaneously, which indicated a downward trend in sulfate that was significant at p = 0.018.