ESTIMATION. OF MISSING RAINFALL DATA THROUGH ARTIFICIAL NEURAL NETWORK (ANN

Abstract

Continuous hydrological observations are required for assessment of water resources from a catchment for planning and design. However, gaps in the hydrological data series are common due to variety of reasons. For meaningful hydrological analysis, missing values specially rainfall data in such data series are often reconstructed using statistical and deterministic methods. Depending upon the type of data and the information available, different techniques are employed to find the missing rainfall values. In this report, some of the methods commonly employed for filling missing hydrological data / rainfall data have been discussed taking the catchment of Purna river in Maharashtra. Back-propagation technique of ANN has also been used along with Arithmetic Average, Normal Ratio and Inverse Distance Methods, in present study to predict the missing rainfall values and compare the results obtained by various methods. ANN model developed for prediction of missing rainfall values is trained with different learning algorithms, learning rates, and number of neurons in its hidden layer. The aim was to develop a network which provides an optimum output. On comparison of the results, ANN was found to be the most efficient technique to predict the missing rainfall data in comparison to the traditional methods.