Assessing Detectability of Change in Low Flows in Future Climates from Stage Discharge Measurements

Abstract

This study assesses whether statistical changes in low flows could be detected should they occur in future climates. Since rating curves are a key stage in the development of a discharge record, their statistical attributes determine if changes in flows can be detected. However, since uncertainty about a rating curve is heteroscedastic (i.e., the errors are drawn from different distributions for different values of the independent variables) there is a need to use a statistical procedure that correctly allocates uncertainty. A simple statistical procedure that allows a stepwise estimate of the variance of the rating curve is used in this paper. The procedure estimates the variance components over finite intervals of a generalized function and allows isolation of seasonal measurements, in particular measurements made during winter conditions. The procedure is demonstrated for one rating curve and the method is used to determine confidence limits for low flows for both summer and winter measurements from 17 stations in south-central British Columbia. The uncertainty for low flows in the warm temperature seasons of summer and fall is low compared to the uncertainty for low flows during the cold temperatures of winter. This indicates that small changes in summer low flows will be detectable, while similar changes in winter low flows cannot be resolved.