EFFECT OF INTRODUCING UNCERTAIN HISTORICAL HYDROLOGIC DATA ON QUANTILE ESTIMATION ACCURACY

摘要

When the length of systematic gauged hydrological record is not enough, historical flood information is practically useful to improve the accuracy of flood-quantile estimation. Historical flood data, which can be utilized for flood frequency analysis by the adjusted-moment method or the maximum likelihood method, often have much larger errors than systematic gauged data. Thus, it is important to evaluate the effect of the error on the precision of flood-quantile estimators. Monte Carlo simulations using the Gumbel distribution show that flood-quantile estimates contain positive bias when the standard error of historical data exceeds a certain level (in this study, approximately 1/6 of the average for the systematic gauged data). Moreover, when it exceeds 1/4 to 1/3 of the average, improvement of the accuracy cannot be confirmed for calculated flood-quantile values.