Determination of sediment sources in a mixed watershed within the Appalachian-St. Lawrence Lowland Regions of southern Quebec using sediment fingerprinting

摘要

This paper identifies the main sediment sources to the Beaudet Reservoir in Quebec (Canada) using sediment fingerprinting. The reservoir, which is built on the Bulstrode River and provides drinking water to Victoriaville, has decreased in capacity by 35% in the past 35 years. This study provides new data on fingerprinting in large and complex watersheds, a first in the province of Quebec. Nine sampling sites on the Bulstrode River and its three main tributaries were selected and five sampling campaigns were conducted. Samples from river bank profiles and adjacent fields, along with suspended sediments, were collected. All samples were sieved to 2 mm and analyzed for 137Caesium, 15 geochemical elements and sieved to 63 mu m for color analysis. Source classification, based on an ANOVA test to verify the independence hypothesis and iterative linear discriminant analysis to optimize the ratio of inter-group/within-group variability, resulted in four sample classes: agricultural soils, forested soils, stream bank bottom and stream bank top. A Kruskal-WallisHtest then identified 21 out of the 32 tracers withpvalue 0.05. The linear discriminant analysis led to a set of 14 tracers, namely 137Cs and 13 color coefficients with a discriminating result of 94%. That combination of 137Cs and color coefficients proved to be a cost-effective fingerprint. Based on MixSIAR modeling results, this sediment fingerprinting study has demonstrated that the main sediment sources varied within the watershed but, generally, forested soil particles dominated (33 to 49%), then agricultural soils (43 to 50%) reflecting the land use changes, followed by stream bank bottoms (82%) at the Beaudet Reservoir.