Heavy rainfall occurs twice a year in the country and lately, thousands of people are always left homeless and hundreds lose life due to floods and landslides where rivers, dams, lakes and sewages overflow enhancing the spread of corona virus in slums. Agricultural products in the farms are also destroyed by floods, affecting agricultural performance to decline as it the key driver of the economy growth. Therefore we used inter-crossed model which was the combination of autoregressive moving average and artificial neural network. Zebiak cane model was also used for selection of variables that were associated to physical processes and testing the network variables. Climate networks were found to be effective tool for more qualitative El Niño Southern Oscillation prediction, by looking at a warning of the oncoming of El Niño when a predestined network attribute surpasses some critical value and also feed forward artificial neural network structures were found to be the first performing structure in terms of normalized root mean squared error at a three month head time prediction. By adding the network variable, we came up with a twelve month lead time prediction with same skill to the predictions at lower set times.

Heavy rainfall occurs twice a year in the country and lately, thousands of people are always left homeless and hundreds lose life due to floods and landslides where rivers, dams, lakes and sewages overflow enhancing the spread of corona virus in slums. Agricultural products in the farms are also destroyed by floods, affecting agricultural performance to decline as it the key driver of the economy growth. Therefore we used inter-crossed model which was the combination of autoregressive moving average and artificial neural network. Zebiak cane model was also used for selection of variables that were associated to physical processes and testing the network variables. Climate networks were found to be effective tool for more qualitative El Niño Southern Oscillation prediction, by looking at a warning of the oncoming of El Niño when a predestined network attribute surpasses some critical value and also feed forward artificial neural network structures were found to be the first performing structure in terms of normalized root mean squared error at a three month head time prediction. By adding the network variable, we came up with a twelve month lead time prediction with same skill to the predictions at lower set times.