Abstract
Strong tropical cyclones have made a drastic effect on human life and natural environment. As large amounts of meteorological data and monitoring data continue to accumulate, traditional methods for predicting tropical cyclone tracks face numerous challenges regarding their prediction efficiency and accuracy. Deep learning methods recently have been proven to be able to learn both spatial and temporal features from a large amount of dataset and be extremely efficient and accurate for forecasting data in complex structures. In this paper, we propose a novel data-driven deep learning model to predict tropical cyclone tracks using the spatial locations and multiple meteorological factors. This model comprises a multi-dimensional feature selection layer, a CNN layer and a GRU layer. The proposed model was trained using a dataset of real-world tropical cyclones from the years 1945 to 2017. Through the comparison experiments, the results verify that the proposed model outperforms the traditional forecasting methods, including a climatologically aware forecasting technique, the Sanders Barotropic technique and a numerical weather prediction (NWP) model. In addition, the proposed model has better accuracy than some deep learning methods, including RNN, GRU, CNN, AE-RNN, CNN-RNN, and CNN-GRU without the proposed feature selection layer.
Highlights
Tropical cyclones are a type of mesoscale or synoptic warm cyclone generated on the ocean surface in tropical and subtropical regions
We proposed a novel fusion tropical cyclones track prediction method based on convolutional neural network (CNN) and Gated recurrent unit (GRU) models
The deep learning model is composed of a Convolutional Neural Network (CNN) layer that can extract feature vectors from the meteorological data and track data, and a Gated recurrent unit (GRU) layer that can learn the temporal features from the k timestamps and predict the center location of the tropical cyclone at the timestamp
Summary
Tropical cyclones are a type of mesoscale or synoptic warm cyclone generated on the ocean surface in tropical and subtropical regions. Our proposed model can effectively select the most correlated meteorological variables and time range that affect the tropical cyclones tracks from the multi-dimensional feature selection layer, which can eliminate unimportant data and improve the prediction. (1) A multi-dimensional feature selection layer is proposed in our method to choose the most correlated meteorological variables and time range to tropical cyclone tracks from the perspective of attribute correlation analysis and temporal correlation analysis. (2) The deep feature extraction is considered in our work, where a CNN layer of our proposed model tends to learn and extract implicit features of the meteorological variables and tropical cyclone tracks in a correlated time range, and a GRU layer, taking output of the CNN layer as input, tends to mine the deep temporal features. (3) The proposed model was validated using a real-world tropical cyclones dataset, and the experimental results suggest that our proposed model can achieve a more accurate prediction result than some existing traditional tropical cyclone forecasting methods and deep learning methods
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call