Rainfall analysis and forecasting in Mumbai metropolitan region using artificial neural networks (ANN)

Irregular rainfall pattern and high surface runoff are the main reasons of water logging and subsequent flood conditions in the Mumbai metro Politian region. Also, the urban heat island effect leads to increased rainfall, both in amounts and intensity in the city. Therefore, research on rainfall patterns and their variations as well as prediction of rainfall for short term as well as long term periods is necessary in Mumbai region. In the present study, an attempt has been made to perform statistical analysis of rainfall for various duration and return periods, and analysis of rainfall trend and intensity pattern for Mumbai region. The records of the observed annual maximum daily rainfall were analyzed statistically for different rainfall return periods (2, 3, 5, 10, 15, 20, 30, 50, 70 years) during the period 1994 to 2013 for Santacruz rain gauge station using Gumbel's extreme value distribution model. The forecasted rainfall values obtained from Gumbel's model is then validated using Artificial Neural Network model. It was observed from the study that the model architecture 4-2-1 performed best with a (R²) value of 0.99 and mean absolute error of 0.04 for rainfall forecasting in Mumbai metropolitan region.