A STATISTICAL APPROACH TO ASSESS FACTORS AFFECTING WATER CHEMISTRY USING MONITORING DATA1

ABSTRACT: A method to partition the variation in concentrations of water chemistry parameters in a river is described. The approach consists of fitting a family of curves for each chemical parameter. Each curve indicates the response of the parameter to river flow for a particular time period or location. An analysis of covariance is then used to identify statistically significant differences between curves. Such differences result largely from two factors: (1) the discharge of effluents and (2) river flow-concentration relationships. The deviations from the fitted curves indicate month-to-month variations unrelated to river flow that are controlled by factors such as temperature-related seasonal patterns. Underlying statistical assumptions are discussed with respect to water chemistry data. The technique is applied to a data set consisting of monthly samples of 22 water chemistry parameters from the Sulphur River of Texas and Arkansas. Several patterns of response to river flow and to two effluent discharges were revealed.