Analysis of Mathematical Models for Rainfall Prediction Using Seasonal Rainfall Data: A Case Study for Tamil Nadu, India

Rainfall prediction is a major issue in climatological studies because it is connected with human life and the economy. Unpredicted and extreme rainfall events is a major concern in mitigation of natural disasters such as flood, drought, etc. which is a common phenomenon encountered in many locations across the world. In this context, periodical rainfall data is analyzed using various types of Holt-Winters method for the Tamil Nadu region, India. Historical monsoon data is analyzed from 2004 to 2018. Usually throughout the year (0–365), the northeast and southwest monsoon period will get a quantifiable amount of rainfall. Rainfall precipitation for Tamil Nadu range: 431-470.7 mm in northeast monsoon and 174-828.8 mm in the southwest monsoon. Rainfall precipitation events indicate seasonality behavior. So, the exponential smoothing method is implemented in demand forecasting of rainfall perception for the Tamil Nadu region. Various types of Holt-Winters methodologies were implemented in demand forecasting of rainfall data analysis. A comparative study has been made in this research and an effective model has been identified with minimum Root Mean Square Error (RMSE). The results indicate Holt-Winter's Multiplicative method with optimized smoothing parameter (MHW OSP) gives better results (RMSE- 96.308%) than other techniques. This methodology can be implemented for demand forecasting of rainfall data in other parts of the world. Keywords. Rainfall Prediction, Time-series Analysis, Forecasting, Exponential smoothing, Holt Winter's model.