The implications of fitting distributions with non-zero lower limits t
o low flow data are examined. The 3-parameter Weibull distribution is
fitted to annual minimum now series from 60 long term stations in Cana
da. The relations between the estimated lower limit and three sampling
variables (skewness, smallest, and largest observations) were investi
gated. The lower limit strongly depends on the sample skewness; it var
ies directly with the sample skewness, which in turn is highly influen
ced by the largest observation. For a given skewness, the value of the
estimated lower limit is determined by the value of the smallest obse
rvation. Therefore, the lower limit cannot be accurately determined, a
nd the resulting low flow estimates will be either too small or too hi
gh.