Spatial and Temporal Modelling of Water Acidity in Turkey Lakes Watershed

Abstract

<p> Acid rain continues to be a major environmental problem. Canada has been monitoring indicators of acid rain in various ecosystems since the 1970s. This project focuses on the analysis of a selected subset of data generated by the Turkey Lakes Watershed (TLW) monitoring program from 1980 to 1997. TLW consists of a series of connected lakes where 6 monitoring stations are strategically located to measure the input from an upper stream lake into a down stream lake. Segment regression models with AR(1) errors and unknown point of change are used to summarize the data. Relative likelihood based methods are applied to estimate the point of change. For pH, all the regression parameters except autocorrelation have been found to change significantly between the model segments. This was not the case for SO4 2- where a single model was found to be adequate. In addition pH has been found to have a moderate increasing trend and pronounced seasonality while SO4 2- showed a dramatic decreasing trend but little seasonality. Multivariate dimension reduction methods are used to provide an overall graphical summary of the changes in TLW water system. We also report the result of applying segment regression for the analysis of first two principal components in selected stations. The results show that the efforts of the Canadian and US governments to reduce the emission of SO2 have been successful in controlling the acid rain problem in Eastern Canada. The project ends with suggestions for various extensions of the present work.</p>