Artificial Neural Network for Forecasting Monsoon Rainfall of South-West Region in Bangladesh

Abstract

Changing patterns of climate factors have become a point of discussion in recent times worldwide. Several aspects of an individual’s prosperity, like communal, financial, and ecological increment, were impacted directly or circuitously by climate change. Moreover, the Bangladeshi people’s life is extremely affected by heavy rainfall because of its geographical structure, especially in the South-West region. Hence, this paper has experimented with the monthly average monsoon data of average temperature, wind speed, relative humidity, mean sea-level pressure, cloud cover, and rainfall from 1981-2018 and predicted the precipitation of 9 meteorological stations from 2019-2028 of the South-West part of Bangladesh. The monthly average monsoon rainfall strongly correlated with relative humidity, mean sea-level pressure, and cloud cover among all the mentioned weather variables. An artificial neural network (ANN) model was formulated with a gradient descent algorithm to predict the rainfall. R2 value was also measured to see the accuracy of the model. Thereafter, the nine stations of the given region have the following order of average monsoon rainfall:Khepupara(15.22mm)>Potuakhali(14.01mm)>Bhola(11.36mm)>Barishal(10.68mm)>Mongla(10.25mm)>Khulna(9.33mm)>Satkhira(9.00mm)>Faridpur(8.67mm)>Jashore(8.64mm). The predicted and real rainfall patterns showed the same escalating or plummeting trends for each station, which justified the ANN model for predicting the monthly average monsoon rainfall of the South-West region in Bangladesh. Such a rainfall prediction can assist people of this region to be more equipped for adverse heavy rain, saving lives and decreasing infrastructure loss during the monsoon season.