Long-term forecasting of tropical cyclones over Bay of Bengal using linear and non-linear statistical models

Abstract

Forecasting tropical cyclones with climate and physical variability and observed cyclonic disturbances has been developed over the years for all the ocean basins successfully and is still one of the priorities for disaster risk reduction policymaking. This study attempts to forecast seasonal cyclonic disturbances and severe cyclonic storms over the Bay of Bengal, where about 80% of the tropical cyclones of the North Indian Ocean are formed. We have used three time-series models, namely, the seasonal autoregressive integrated moving average with exogenous variables (SARIMAX) model, artificial neural network-nonlinear autoregressive with exogenous variables (ANN-NARX) model, and the hybrid model. The basic purpose of considering three different models is to improve the forecasting accuracy of tropical cyclones. We have shown that the intensification rate of the severe cyclonic storms over the Bay of Bengal has been significant and increasing over the years. Results show that the ANN-NARX model with sea surface temperature and near-surface wind speed as predictors is the best performance model for longterm forecasting of cyclonic disturbances. Hence, the distribution of cyclonic disturbances is non-linear. The correlations between observed and predicted occurrences are 0.80 and 0.85 for cyclonic disturbances and severe cyclonic storms, respectively, corroborating, by and large, the forecasting accuracies of some previous studies. The forecasting of cyclonic disturbances indicates that they will vary from 5 to 13 annually and there will be, on average, one severe cyclonic storm per year. The likelihood of occurrence of severe cyclonic storms is most significant in the post-monsoon season. This forecast till 2050 would help the scientific community and policymakers significantly for applications and good disaster risk governance.

Cyclone Amphan effects on the copepods of Ganges estuary

“…A roadmap of cyclone ecology for Indian estuaries Intensity of TCs is increasing in BoB region which is a global hotspot of TCs (Golder et al 2021). Till 2050 annual incidences of cyclonic disturbances may vary from 5 to 13 and on average there may be one severe cyclonic storm per year and that is most likely in the post-monsoon (Sen et al 2021); therefore, estuarine communities are likely to be stressed for longer time than they used to be from cyclone-mediated disruptions (Kumar et al 2017;Paul et al 2020a). That may lead to unforeseen consequences for estuarine plankton including enhanced exibility to adapt stressful conditions, shift in ecological distribution and a few species may even perish (Cheal et al 2002;Peierls et al 2003;Liu et al 2021).…”

Seasonal forecasting of tropical cyclones over the Bay of Bengal using a hybrid statistical/dynamical model

“…Several methodologies have been developed for the seasonal forecast of TC activities in different parts of the world (Elsner and Schmertmann, 1993; Chan et al ., 1998; Camargo and Barnston, 2009; Kim et al ., 2010; Balachandran and Geetha, 2012; Vecchi et al ., 2014; Camp et al ., 2015; Nath et al ., 2015, 2016; Wahiduzzaman et al ., 2019; Sen et al ., 2021) since its first attempts made by Nicholls (1979) and Gray (1984a; 1984b) for the Australian and North Atlantic regions, respectively. Currently, many institutions are issuing seasonal outlooks of TC activities for various regions (e.g.…”

“…The root mean square error (RMSE) and bias error between the predicted and observed cyclonic disturbance days over the North Indian Ocean is 5 and 0.36 days, respectively. Recent studies explored the artificial neural network models to forecast seasonal cyclone activities over the North Indian Ocean (Nath et al ., 2016; Sen et al ., 2021). The nonlinear artificial models offer a reasonable skill in forecasting the seasonal cyclone activities over the North Indian Ocean in terms of various statistical metrics (Nath et al ., 2016; Sen et al ., 2021).…”

“…Recent studies explored the artificial neural network models to forecast seasonal cyclone activities over the North Indian Ocean (Nath et al, 2016;Sen et al, 2021). The nonlinear artificial models offer a reasonable skill in forecasting the seasonal cyclone activities over the North Indian Ocean in terms of various statistical metrics (Nath et al, 2016;Sen et al, 2021). Using a statistical model, have shown that the state of the quasi-biennial oscillation has the potential to improve the seasonal forecast skill of TCs over the North Indian Ocean.…”