Flood frequency analysis with regional and historical information

This paper develops generalized maximum likelihood estimators which efficiently employ both at‐site historical information and systematic flood records with a regional index flood distribution. Monte Carlo experiments document the performance of these estimators in heterogeneous and homogeneous regions, with different historical record lengths, and with various perception thresholds. When carefully employed with stationary flood records and appropriate and robust statistical procedures, historical information is shown to be valuable at those sites where it is available, as well as in the estimation of the regional index flood distribution.