In this study, the Huff curves were extracted using the 418 events in the four selected stations namely Tabriz, Sarab, Malekan and Heris (all of them located in the East of Urmia Lake). In the first step, the total selected storms were classified into the four distinct classes according to their rainfall durations including i) 0-2, ii) 2-6, iii) 6-12 and more than 12 hours. Then, the Huff curves of each category were plotted for the 10%, 20%, … and 90% probabilities. Analysis conducted for each of the classes, separately for the stations. Moreover, the Huff curves were plotted using the information of all events in a unit class. In this study, some statistical distributions commonly used in hydrology were utilized. The three newly defined indices namely S, I, and Q were calculated in the present study. The design storm hyetographs for each of the selected stations using the information of all events in a unit class prepared for 50% and 90% Huff curves. The mathematical models of Huff curves were extracted as the Logistic model. The model parameters were estimated for models parameters. Results indicated that for the 0-2 hours rainfall duration class, except the Sarab station, having the first quartile precipitation type, the storms of the three other stations belonged to the second quartile type. For the 2-6 hours rainfall duration class, storms of the stations Sarab and Heris are included in the first quartile and the stations of Tabriz and Malekan are included in the second and third quartile, respectively. For the storms with the duration of 6-12 hours all the sites, (except Sarab station which is known as the first quartile) the other stations are known as the third quartile. Also, in the case of a class with duration of more than 12 hours the two stations namely Tabriz and Sarab are incorporated in the first quartile. Storms type of the Heris station is known as the third quartile. Also, results showed that the vertical distance of 50% and 90% Huff curves in the first, second and third quartiles in all the stations shorten as the quartiles increases (S>I>Q). Based on the results, it was found that the Logistic model is capable to fit the mentioned curves very well for the selected station. The correlation coefficients between the observed and modeled values were found to be from 0. 978 to 0. 998.