1989 Volume 25 Issue 7 Pages 737-743
Frequency analysis of annual maximum k(=1, 2, 3)-day precipitation in the Chikuma basin is considered by using c-histogram which can be applied even if maximum entropy distribution does not exist. The application study to the hydrological data shows that the c-histogram with 30 blocks and 3 moments gives good fit to the hydrological histogram and it estimates the suitable return period for maximum value judging from the number of data. Comparing with other distributions, the c-histogram estimates the large probable hydrologic variate in many cases which means the safety side in flood-control planning and the small standard deviation where the probability density of the right tail in the hydrological data histogram is large, though its log likelihood is smaller than those of other distributions. Besides the parameter identification time of the c-histogram does not depend much not only the number of given moments but also the number of blocks, and is much shorter than that of the maximum entropy distribution. The short calculation time for the estimation of the c-histogram is to be important when it is developed from 1-variate to multivariate.
