The conventional and L-moment method of flood frequency analysis are applied inrndetermining flood magnitude of defined return periods by selecting the best-fit theoreticalrnprobability distribution. The most important part of the distribution is the tail as far as extremernflooding phenomena. These phenomena of floods are of utmost concern in water resourcesrndevelopment and management. In most cases the central part of the theoretical distribution fitsrnsatisfactorily with the empirical points. In FFA the objective is to estimate flood magnitudern(Q) corresponding to any specified recurrence interval of (T) years. The estimation isrncomplicated due to lack of a physical basis for determining the form of the underlying floodrnfrequency distribution and the necessity of evaluating flood event for return periods thatrnexceed the observation period (Leulseged, 2002).rnRegionalization is generally accepted term to explain the transfer of information about floodrnpeak at one catchment derived from other catchments with similar characteristics. Thernadvantage of such procedure is particularly great in the estimation of frequencies for higherrnflood magnitudes with limited at site data do exist and inference in the tail of probabilityrndistributions makes the stabilization of the estimators difficult. It is quite clear thatrnregionalization is most viable way of improving flood quantile estimation. Although therernremain researchable topics in development and application of regionalization method. Thernperformance of regional Extreme value and LN distributions are found to be highlyrnsatisfactory and can be widely applied in this paper. More attention is given to at-siternhomogeneity test to group stations in the upper Awash sub-basin (u/s koka) in to two regionsrnafter checking them for the consistency and independency testes and estimation of standardrnerror. The upper Awash sub-basin (u/s of koka) has 10-selected gauged stations consisting ofrnstream flow record varying from 12 to 37 years, out of which 6 stations are found in the upperrnregion and 4 of the stations are found in the lower region. An Extreme value EV1, GEV andrnLognormal LN2, LN3 distributions are selected as the best fit distribution for the stations inrnthe sub-basin. The sub-basin has been divided into two regions the upper and lower region,rnthe upper region covers 18.46% of the sub-basin and the lower region covers 81.54% of thernsub-basin.