Rainfall forecasting model to make effective decisions on cultivation using multiple artificial neural networks – Case study Gampaha District

Rural communities in developing economies mainly depend on climate-sensitive activities such as agriculture for their livelihood and are particularly vulnerable to climate change, mainly to rainfall. Therefore, accurate forecasting of rainfall is vital for the sector of agriculture. This study aims at forecasting the rainfall for a farming village located in the Gampaha District, where the majority depends on agriculture. Artificial Neural Networks (ANN) can be identified as a prominent technique for forecasting among all forecasting techniques which was inspired by human brain neural systems. ANNs were employed in the study to build the rainfall forecasting model. The other climatic factors: temperature, humidity, pressure, wind speed, and cloud percentage were considered as affecting factors for rainfall in the model building process. Previously recorded data were collected from several stations near the study area. This study expanded to develop multiple artificial neural networks using Feed Forward Neural Networks (FFNN) for each climatic factor separately and used the forecasted values of those models as input to the rainfall forecasting model. Models were trained thousands of times by changing the model parameters and obtaining optimal parameters. The model was finalized based on the minimum mean squared error of 0.0547 and minimum normalized mean squared error of 0.378. With the involvement of many climatic factors, the proposed model leads to identify the rainfall variability which will be potentially allow farmers and others in the agricultural sector to make decisions on reducing unwanted impacts or take advantage of expected favorable climate.