Estimation and analyzing the frequency of rainfall is essential to help in defining the policies regarding water resource management and a source of data for flood hazard mitigation. Additionally, design rainfall is widely utilized in hydraulic structures and in urban infrastructure planning. Therefore, finding the suitable probability distribution that fits the actual data is considered as the first step in design rainfall estimation. In this study, 15 probability distribution models are used to determine the best fit probability distribution in the case of annual rainfall. Kolmogorov–Smirnov (K–S) test is used to select the best fit probability, distribution model. It is found that Pearson 6, 3-parameter-Weibull and Log-Pearson 3 distribution functions gave better fits for the rainfall data for all locations. Moreover, rainfall is characterized by pronounced fluctuations, and geographic differentiation, which causes many difficulties in its accurate simulation and prediction. Therefore, Multi-Layer Perceptron Neural Networks (MLPNN) and Multiple Linear Regression (MLR) are used for predicting the annual rainfall in Sudan. To this aim, the minimum temperature (Tmin), maximum temperature (Tmax), average temperature, solar radiation, wind speed and vapor pressure were used as input variables for the models. The results indicate that MLPNN model shows potential for predicting the annual rainfall in Sudan.