Quantifying the Impact of Hydrometric Network Reductions on Regional Streamflow Prediction in Northern Canada

Abstract

Hydrometric monitoring in northern Canada, as in much of the country, was reduced during the mid-1990s in response to fiscal pressures on the federal government. This study uses a combination of canonical correlation and multiple regression analysis to quantify the influence of discontinuing these gauges on prediction of ungauged streamflows in the region. Thirty-four stations in the Mackenzie Basin measuring streamflow from catchments in the Yukon, the Northwest Territories, northern Alberta and northern British Columbia were selected for the analysis. Twelve of the 34 stations are now closed. The difference in the predictive capability, as determined by a jack-knife procedure, between using only the active stations versus using both active and discontinued stations is assumed to represent the impacts of reducing the network size. When the 12 closed stations were not used in the analysis, extrapolation error increased by ~16% for all flow regimes. Uncertainty, as expressed by the confidence interval, remained the same for the average streamflow regime but doubled for the high flow regime. Changes in climate and the landscape will eventually result in the data from the discontinued stations being inappropriate for use in engineering design, resulting in the estimated increased error and uncertainty in streamflow prediction reported here. This uncertainty is likely to be dealt with by over-designing structures, a costly measure in northern Canada. Conversely, the risks associated with under-design could increase. © 2007, Taylor & Francis Group, LLC.