In this paper the Quantile estimates for various recurrence periods of peak floods, at Kotri Barrage on Indus River, are obtained by Gumbel and GEV (Generalized Extreme Value) Distributions using three parameter estimation methods. A comparison between distributions is made on the basis of various goodness of fit tests showing that Gumbel is better than GEV. The efficiency of three estimation methods for Gumbel is compared using SEs (Standard Errors) showing that MLM (Maximum Likelihood Method) is the best efficient. PWM (Power Weighted Moment) is better than MOM (Method of Moments) for GEV, on the basis of lower quantiles and by goodness of fit tests. It is recommended that the Gumbel distribution be used in flood frequency analysis on Indus river at kotri Barrage. The Quantile estimate for 100-years recurrence intervals, by Gumbel distribution, with MLM is 0.975 million cusecs which is 0.1 million cusecs (i.e. about 10%) higher than its designed capacity. Hence, it is suggested that authorities should be vigilant and take appropriate measures to pass the above predicted higher magnitude of flood safely from the Barrage without damaging its gigantic structure.