| Popis: |
A multiple regression model is introduced to describe temporal variations in precipitation chemistry data. The model considers the effects of the annual cycle, a linear trend and precipitation-quantity simultaneously. The paper discusses the application of the model to concentrations and depositions of hydronium, ammonium, nitrate and sulfate for monthly bulk samples in The Netherlands for the period 1978–1984. Statistical conclusions about the annual cycle and the trend are hardly influenced by the choice of the dependent variable (depositions, concentrations or logarithms of concentrations). With the exception of H+ concentrations, a large part of the temporal variations was due to precipitation-quantity effects. Significant annual cycles were found for nitrate, ammonium and sulfate. There was statistical evidence of a downward trend for sulfate and nitrate. A complex, non-linear trend was observed for H+ which resulted in a significant autocorrelation of the residuals from the regression equation. Much attention is paid to the detectability of trend. For ammonium, nitrate and sulfate it is possible to discriminate small systematic changes in quite short records (a mean annual change of 4–6% in a 5-year record). This is not the case for H +, because temporal variations of this component are insufficiently explained by the systematic annual cycle and precipitation-quantity. |
