Theme-1-4-Moderating outliers in rainfall frequency analysis.

Abstract

The high outliers in an extreme rainfall series are not be rejected as our interest lies in estimating the probable high intensity for design purposes. Maintaining the outliers in extreme series poses problem of finding a suitable distribution to describe the series. EV-II, the only befitting distribution in such cases tends to yield very high intensities for higher return periods, which is generally undesirable for practical use. As such the moderation of outlying observations seems to be an acceptable way out. In this paper, the authors have theoretically examined the effect of i) rejecting; ii) moderating outliers on the return period values of rainfall and their standard errors. It has also been analytically deduced as to what can be the extent of minimum moderation which makes an outlier to fit in a suitable choosen extreme value distribution.

The results of derivation have been illustrated by practical exercises on daily extreme rainfall series containing outliers of different magnitudes.