MLR-based Study of the Wind-pressure Profile in Arabian Sea Cyclones

Abstract

Tropical cyclones (TCs) in the Arabian Sea have become more frequent and intense, due to climatic variability. These changes represent an increasing hazard to coastal communities and infrastructure throughout the area. Accurate calculation of cyclone wind and pressure fields is critical for cyclone forecasting, early warning systems and disaster response. Usually cyclones are studied with the help of wind-pressure models. One such model is Holland Model which utilizes a wind-pressure profile parameter B. This parameter which is a power exponent plays an essential role in shaping wind and pressure profiles. It defines that how sharply the pressure drops from the outer environment to the cyclone’s center. Although various models for estimating the wind-pressure profile parameter B exist for basins such as the Atlantic, Northwest Pacific and Bay of Bengal, no such model has been specifically developed for the Arabian Sea. This study addresses this gap by constructing and evaluating multi- linear regression (MLR) models to create such power exponents for Arabian Sea cyclones using best-track cyclone data from the IBTrACS dataset (1980–2023) which is the satellite data. Four models B_α, B_β and B_γ for wind-pressure profile parameter B are developed for Arabian Sea cyclones incorporating different combinations of meteorological variables, including maximum sustained wind, central pressure, pressure drop, radius of maximum wind (RMW) and latitude. Model’s performance is assessed using adjusted R², predicted R², RMSE and MAE. Model B_γ is the best-performing model with the predicted R² of 78%, with the lowest standard error (0.1395). We have also calculated B, the Holland’s wind-pressure profile parameter, on Arabian Sea data. The data used here is satellite data instead of aircraft data. Further, all these models including B are applied to the 20% test dataset to validate the predicted wind-pressure profile parameters. Finally, we will compare B_α, B_β and B_γ with B. For the test data B_γ shows good accuracy. Its correlation with B values is 0.7796. These models are used to generate both the wind and pressure profiles of Cyclone Tauktae. Finally, results obtained by MLR are compared with the results obtained by B.  Among all MLR models B_γ gives results closest to B model. So, the model B_γ is recommended as the most reliable MLR model for estimating the wind-pressure parameter for the Arabian Sea.