PREDICTING THE LOCATION AND MOVEMENT OF CYCLONE EYE BY WIND DIRECTION USING AI & ML
PREDICTING THE LOCATION AND MOVEMENT OF CYCLONE EYE BY WIND DIRECTION USING AI & ML
Authors:
Batta Srikanth1, B Nithin Reddy2, Mrs.M.Priyanga3, Dr.P.Dhivya4
1234Department of Computer Science,
Dr. M.G.R. Educational and Research Institute, India
Abstract—Tropical cyclones are some of the most destructive forces of nature and a serious threat to life, infrastructure, and coastal ecosystems across the world. The devastating impacts of tropical cyclones are the result of extreme winds, heavy rainfall, and the inundation caused by storm surges. Accurately predicting where the eye of a cyclone will be located is critical for successful implementation of early warning systems, evacuation plans, and efficient disaster management. The objective of this project is to develop an AI and ML based framework for predicting both the location and movement of a cyclone’s eye by using wind speed and wind direction as primary indicators of cyclone dynamics. The system will incorporate historical (retrospective) and real-time meteorological datasets, including but not limited to atmospheric pressure, wind direction and magnitude, and geospatial coordinates sourced from a variety of reliable sources (e.g., IBTrACS, ERA5 reanalysis data). Baseline methodologies utilize supervised regression models (i.e., Random Forest regression and Gradient Boosting) as reference models for the analyses, while deep learning models (i.e., ConvLSTM, Transformative sequence models) are employed to capture complex spatial and temporal patterns of wind-flow properties that affect the movements of the cyclone. Using the predicted eye locations of the cyclone, the coordinates will be defined by the latitude and longitude of the cyclone's movement at each future time step. Additionally, data preprocessing methods will be applied to aid in the accuracy of prediction (i.e., normalization, spatio-temporal feature engineering). The performance measures will be calculated using Root Mean Square Error (RMSE) and Haversine distance error for measuring spatial accuracy of the predicted location. The proposed framework is designed to assist with near-real-time tracking of the cyclone and provide support for disaster response agencies.
Keywords—Tropical Cyclone Eye Prediction, Wind Direction Analysis, Wind Speed Modeling, Machine Learning, Deep Learning, ConvLSTM, Transformer Networks, ERA5 Reanalysis Data, IBTrACS Dataset, Spatiotemporal Forecasting, Haversine Distance Error, Disaster Management Systems.