Weather forecasting and modeling using SARIMA, ANN and hybrid model for central zone of Kerala

Abstract

The forecasting of weather parameters is one of the main objectives faced by scientists all over the world. Predicting weather parameters is important because it helps control the effect of natural calamities due to climate change by taking precautionary measures to manage the harmful effects. The forecasting of weather parameters is also important in agriculture activities since various crops, from sowing to till harvesting, clearly depend upon factors like rainfall, temperature and relative humidity. The prime focus of the current study was to undergo modeling and forecasting of weather parameters such as rainfall, maximum and minimum temperature, relative humidity, cloud content and wind speed with maximum accuracy for the central zone of Kerala. The monthly weather data including rainfall, maximum and minimum temperature obtained from RARS Pattambi in Palakkad district of Kerala and data including relative humidity, cloud content and wind speed using data access viewer from the same location over 39 years (1982-2020). The methods for modeling the weather parameter are SARIMA (Seasonal Autoregressive Integrated Moving Average), ANN (Artificial Neural Network) and hybrid SARIMA-ANN models. The comparison for identifying the best model suitable for each weather parameter was selected based on mean square error (MSE), root mean square (RMSE), mean absolute error (MAE) and coefficient of determination (R^2). The model showing the least value for error and the highest value for coefficient of determination was selected as the best model. The results revealed that weather parameter showed the best performance for different models. According to the findings of the study, the ANN model is the most accurate model for projecting anticipated values of rainfall. The best model selected for forecasting minimum temperature and cloud content was the hybrid SARIMA-ANN model whereas, for maximum temperature, relative humidity and wind speed are SARIMA  , SARIMA   and SARIMA   respectively. The best-fitted model was employed for forecasting weather parameters for the next 5 years.