Analysis on the occurrence of tropical cyclone in the south Pacific region using recurrent neural network with LSTM

Abstract

Weather prediction over the years has been a challenge for the meteorological centers in the South Pacific region. This paper presents Recurrent Neural Network (RNN) Architecture with Long Short Term Memory (LSTM) times-series weather data for prediction. From the gathered dataset, the Sea Surface Temperature (SST) is studied since it is known to be the foundation of the cyclone formation. This paper focuses on two scenarios. The first part is predicting upcoming SST using dataset from January 2013 to December 2017. The second part is taking out data of two different cyclones and predicting the SST for the next 14 days. Once the SST prediction is made, the predicted SST is compared with SST in the dataset for those 14 days. The main aim of this paper is to predict the SST using RNN and LSTM to anticipate the occurrence of tropical cyclones. The paper will focus on the reason for this study, a discussion of the model used, how the cyclones are formed, regarding the current threshold, the analysis of the dataset and lastly, the results from the experiment carried out.