Application of Artificial Intelligence in Forecasting the Intensity of Tropical Cyclones over North Indian Ocean: Skill Comparison with Operational Models.

The endeavour of the present research is to develop a neuro-computing based adaptive intelligent model to observe the forecast efficiency in predicting the central pressure drop (PD) and maximum sustained wind speed (MSWS) associated with different categories of cyclones at the stage of the highest intensity over the North Indian Ocean (NIO). The cyclonic systems include the phases from deep depression to extreme severe cyclones reported over NIO. The 3-hourly data and records of cyclonic systems are collected from India Meteorological Department (IMD) Best Track dataset during the period from 1990 to 2019. Present study considers eight parameters comprising five genesis parameters; sea surface temperature, low-level vorticity, mid-tropospheric relative humidity, surface to middle troposphere equivalent potential temperature difference, inverse of wind shear and three vertical wind velocity parameter at 850, 500 and 200 hPa pressure levels. Based on different statistical analysis significant parameters are selected as the input parameter of different artificial neural network (ANN) models. Various neural network models with different architectures have been trained with the data from 1990 to 2012 to select the best forecast model over Arabian Sea. The result shows that the multi-layer perceptron (MLP) model with five input layers, one hidden layer with four nodes and one output layer is the best model for forecasting PD with 99.92% accuracy at 36 h lead time, whereas the MLP model with five input layers, one hidden layer with five nodes and one output layer is found to be the best model for forecasting MSWS with 99.78% accuracy at 48 h lead time. The results are well validated with the observations from 2013 to 2019. The forecast skill of MLP model is compared with multiple linear regression model and existing operational and numerical weather prediction models. The adaptive neural network models are trained with the data from 1990 to 2015 to forecast the PD and MSWS over BOB. The result reveals that the multilayer perceptron (MLP) model provides good accuracy at 6 and 30 h lead time in forecasting the PD. The result further shows that the MLP model is the best model for forecasting the MSWS of cyclonic systems with 60 h lead time with minimum forecast error. The model results are well validated with observations from 2016 to 2019.