Values which are unusually large or small with respect to the rest of the values in dataset are called outliers. The outliers affect on estimation of parameters of distribution, evaluation of model performance, etc. The presence of outliers in rainfall data affects rainfall depth-duration-frequency (DDF) curves which provide design rainfalls for hydrologic structures. The DDF curves constructed using rainfall data containing outliers could lead to inefficient designing of hydrologic structures. This study proposed the use of a mixed Gumbel distribution for constructing rainfall DDF curves considering the effect of outliers. The results indicate that the mixed Gumbel model provides more reasonable estimates of rainfall DDF curves by reducing the considerable effect of outliers.