A study of trends and variability of low flow characteristics was conducted for the Reference Hydrometric Basin Network (RHBN established in order to assess the impacts of possible climate change on Canada's water resources). Low-flow index of 7-day from 220 hydrometric stations was extracted and examined to detect trends in timing of 7-day low-flows in both summer and winter portions of the year. Also, 15 stations with dominant zero 7-day low flows were chosen and examined to detect trends in the number of zero events. Man-Kendall (MK) nonparametric trend test was applied to the time series to investigate the existence of any trend at a 0.05 significance level. In order to account for serial correlation, the variance of the S statistic was modified if the absolute value of autocorrelations exceeded an 80% confidence interval.

Statistical analysis showed no evidence of domination of positive or negative autocorrelations in timing of summer 7-day low-flows. In 15% of the studied sites, timing of summer 7-day low flow experienced significant trends with approximately equal numbers of upward and downward trends. While summer significant trends in all of the stations located in Pacific coasts had upward directions, no specific patterns were observed in other parts of the country.

It was observed that timing of winter 7-day low-flows specifically in eastern Canada was slightly dominated by negative autocorrelations. Winter 7-day low-flow in 18% of the stations showed significant trends. In 74% of the detected significant trends, winter 7-day low-flow shifted toward earlier dates. Mapped results revealed that there was a domination of significant downward trends in eastern Canada (Atlantic Provinces), a lack of significant trends in central Canada, and a more or less balance between the number of upward and downward trends in western Canada.

Significant upward/downward trends were detected in 65% of the stations studied for trend in the number of zero events; however, since the time series were highly autocorrelated, only 6% of the studied stations showed significant trends after applying modifications to account for significant serial correlations.