Natural disasters threatening and endangering human communities has resulted in the study and research of such disasters through the related sciences and present methods of forecasting their behavior with time and place and also from a qualification and quantity viewpoint. To this end, numerous methods for the determination of the maximum flood in various return periods has been made available which can be refered to as flood FREQUENCY analysis methods. One of these methods is the regional flood FREQUENCY analysis in which instead of using the data from a single station, it considers the data and characteristics of a group of similar stations. In the case under the research this method uses L-Moments and Index Flood in North, Razavi and South Khorasan water basins and MATLAB software. Maximum annual flood statistics were used from 68 Hydrometric stations with minimum and maximum statistical periods of 6 and 39 years. Using Cluster analysis the region under study was divided to 7 partitions. Discordance test has conducted and only one station in region C was found as discordance station. Because of knowing the homogeneity of the regions, the parameter of Kappa distribution were estimated and with using the simulation method of Monte Carlo with 500 times, the homogeneity measure was tested in 7 regions. Using homogeneity test all regions was found homogen. Using goodness-of-fit measure z and Kolmogrove-Smirnov the Log normal 3 parameters distribution were selected for two regions of A and B, GEV for C, Generalized Pareto for D and E, Generalized logistic for F and Pearson III for G. Besides, GEV distribution was found appropriate for all of the regions, only their parameters are different in any regions. For estimating of index flood a logarithmic model has found for each region with 4 variables of area, height, average slop and form factor. Using of these models, the index flood can be estimated in each region and it can be used for standardize the statistics of maximum flood values.