Estimating peak flow in ungauged basins is a task frequently encounterd in the design and planning of hydraulic and water resources engineering. Regionalization is a way to deal with this issue. The regional formula for peak flows was established using gauged flows and basin topographic characterstics inorder to estimate the peak flows in ungauged areas within the homogeneous region. At first, flood frequency analysis of 9 gauged stations in the study area were analysed based on the best fit Probability whighted Moments (PWM) and L-Moment Ratio Diagram (L-MRD) statistical parametr values. The best fit statistical parametres which fits with the study area (Lake Tana basin) were found to be General Logistic (GLOG),General Pareto (GPAR) and Pearson (P3) type. Secondly, Regional regression function of peak flow with respect to different return periods were developed with topographic inputs of basin area (A), longest stream length (LL), mean slope(S), mean elevation(E), elongation ratio(Er), circularity ratio(Cr) and drainage densities (DD). For regeression model, values of catchment characterstics (predictors) computed by (GIS and archydro tools), were used as independent variable and statistical parametres computed by ( flood frequency analyis) were used as dependent variable with respect to return periods. Moreover,the coefficent of regeression function is found to have high correlation with return periods. With the regerssion model, values of peak flows for ungauged catchments were estimated with some errors and performance measures. Thirdly, the performance of the model was evaluated with six evaluation methods (criterias) suchas Coefficents of Determination (R2 in %), Nash-Sutcliffe Efficiency (NSE),Mean Error (ME),Mean relative error (MRE),Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). The peak flow values, which was estimated by the model, are compared with the values actually obtained by statistical probability distribution method,and the overall result was good in some evaluation criterias and lower return periods. The over all results of model evaluation measures are found to be coefficents of determination(R2) ranges from 87% to 97.7%, Nash-Sutcliffe Efficiency (NSE) ranges from 0.88 to 0.97, Mean Error (ME) ranges from -11.6 to 4.5, Mean Relative Error (MRE) ranges from 10.1 to 30.8,Mean Absolute Error (MAE) ranges from 15 to 47 and Root Mean Square Error (RMSE) ranges from 25 to 67. As return period increases, the performance of the model was reduced with all evaluation criterias. At last, validation of the model was also done based on external data (Garno was used for validation for this study).