Statistical Analysis of Precipitation and CrossComparison of Stochastic and Artificial Neural Network Models for a Short-Term Rainfall Monthly Forecast in Atalanti Gauge Station (Central – Eastern Greece)

The objective of this paper aims at modeling the precipitation data from Atalanti rain gauge station in CentralEastern Greece through an Artificial Neural Network (ANN) with multi-layer perceptron and a Seasonal Auto-Regressive Integrated Moving Average (SARIMA) model which can simulate systems characterized by complicated physical processes. First, a thorough statistical analysis of the rainfall timeseries and its residuals as well was carried out so as to figure out its main statistics before proceeding to further analysis. The use of stochastic methods, introduced by Box and Jenkins, has found wide application for fitting and forecasting the monthly rainfall timeseries which may be useful in decision making as well as risk management and water resources usage optimization. Also, the ANN model has the ability of identifying non-linear relationships bet