RAINFALL DISAGGREGATION MODEL USING ARTIFICIAL NEURAL NETWORKS

RAINFALL DISAGGREGATION MODEL USING ARTIFICIAL NEURAL NETWORKS

ABSTRACT

Since time immemorial, India has been experiencing the wrath of flood and droughts alike. The undeniable cause that initiates the suffering running to the tune of over Rs 1000 crores per annum just in monetary terms is “rain and the lack of it”. The monsoon pattern needs to be given a renewed look at to curb this twin problem. Our study evaluates the principle of artificial neural networks as applied to this. The nonlinear nature of the relationship, unavailability of long historical records, and the complexity of physically based models in this regard, are the essential factors that make ANNs a logical choice. Two ANN models acting in succession are suggested for the same in this paper. The first ANN model is a standard back-propagation feed forward network. It is a three layered network. The second one is a competitive learning model that uses feed back among the hidden layers to accomplish the competition. This paper introduces an alternative for the disaggregation of hourly rainfall based on artificial neural networks (ANNs).. The ANN models are evaluated by comparing their performance with the performance of two established hourly rainfall disaggregation procedures using rainfall data from a gauging station in Macquarie University, Australia.