A Neural Network Approach to Estimate Tropical Cyclone Heat Potential in the Indian Ocean

The tropical cyclone heat potential (TCHP) or the available upper ocean thermal energy is one of the critical factors in controlling the intensity of cyclones. Given the devastating impacts Indian Ocean cyclones could bring (e.g., the “killer cyclone” Nargis in 2008, which caused more than 130000 deaths), there is a pressing need to obtain reliable and more accurate TCHP estimates over the Indian Ocean to improve the cyclone track and intensity predictions. Using more than 25000 in situ subsurface temperature profiles during 1997-2007, this research explores the possibility of developing an artificial neural network (ANN) model to derive TCHP in the Indian Ocean using satellite-derived sea surface height anomalies, sea surface temperature, and climatological depth of 26°C isotherm. The estimations have been validated using more than 8000 independent in situ profiles during 2008-2009. The root-mean-square error and the scatter index of this validation data sets are 14.6 kJ/cm² and 0.2, respectively. Comparison of the estimations from a two-layer reduced gravity model and from a multiple regression method confirms the superiority of the ANN approach over other methods.