Analysis of droughts from both the regional and probabilistic points of view is crucial in drought-prone arid and semiarid lands. In this work, the probabilistic behavior and spatial pattern of extreme hydrological droughts based on the minimum streamflow drought index values at 3-, 6-, 9- and 12-month time scales were analyzed by using L-moments in the Karkheh river basin in the southwest of Iran. The results indicated that for the 3-month time scale, the region was homogenous, while for larger time scales the region was divided into two subregions. For all time scales and subregions, the Pearson Type III and Generalized Pareto distributions provided a better fit to the data than the generalized logistic, generalized extreme value and log-normal distributions. The simulation results of the estimated regional quantiles showed that the accuracy of the quantile estimates decreased as return period increased. Furthermore, higher probability of severe droughts is expected at regional scale compared with site scale.