Analysis of trends in water quality of buffalo pound lake

Abstract

In this study various methods were used to understand the temporal behaviour of specific water quality variables of Buffalo Pound Lake in Saskatchewan. Various techniques were employed in trend determination and exponential smoothing techniques were employed in the modelling of fourteen water quality variables. Most of the variables were serially correlated, making the determination of the magnitude of trends unattainable at the 10% level of significance. The only variables with a statistically significant magnitude of trend were available nitrogen (−0.0066 mg/l.yr) and pH (−0.0072/yr). Upward trends in available phosphorus and silica, downward trends in chlorophyll α and sulfate were significant (p < 0.10), according to the Spearman Rank test. Temperature, dissolved oxygen, total dissolved solids, total suspended solids (upward) and colour were not significant (p > 0.10) using the Spearman Rank correlation coefficient. The Spearman Rank correlation coefficient for total phosphorus, and total suspended solids and colour (p < 0.18), suggest that these variables may be reporting the correct direction of the trend at least 82% of the time. Determination of time series models for each variable using exponential smoothing techniques yielded a variety of responses. Adequate models were obtained for temperature, dissolved oxygen, available nitrogen and total suspended solids based on the evaluation of forecast residuals. Models for total dissolved solids, pH and sulfate appeared on the borderline in their ability to predict effectively. No exponential smoothing models could be developed for available phosphorus and fecal coliforms. © 2003 Taylor & Francis Group, LLC.