An incremental rating curve error model is proposed to describe the sy
stematic error introduced when a rating curve is extended by methods s
uch as slope-conveyance, log-log extrapolation, or fitting to indirect
discharge estimates. Extension can introduce a systematic or highly c
orrelated error which is anchored by the more extensively measured par
t of the rating curve. A likelihood function is developed which explic
itly accounts for such error and accepts both gauged and binomial-cens
ored data. A sampling experiment based on the three-parameter generali
zed extreme value distribution was conducted to assess the performance
of maximum likelihood quantile estimators. This experiment revealed t
hat substantial, and in some cases massive, degradation in the perform
ance of quantile estimators can occur in the presence of correlated ra
ting curve error (rating error). Comparison of maximum likelihood esti
mators allowing for and ignoring rating error produced mixed results.
As rating error impact and/or information content increased, estimator
s allowing for rating error tended to perform better, and in some case
s significantly better, than estimators ignoring rating error. It is a
lso shown that in the presence of rating error, the likelihood surface
may have multiple optima that may result in nonunique solutions for h
ill-climbing search methods. Moreover, in the presence of multiple opt
ima and constraints on parameters, the likelihood surface may be poorl
y described by asymptotic approximations.