Abstract:Tropical cyclones (TCs) can cause significant economic damage and loss of life in coastal areas. Therefore, TC prediction has become a crucial topic in current research. In recent years, TC track prediction has progressed considerably, and intensity prediction remains a challenge due to the complex mechanism of TC structure. In this study, we propose a model for short-term intensity prediction based on adaptive weight learning (AWL-Net) for the evolution of the TC's structure as well as intensity changes, exploring the multidimensional fusion of features including TC morphology, structure, and scale. Furthermore, in addition to using satellite imageries, we construct a dataset that can more comprehensively explore the degree of TC cloud organization and structure evolution. Considering the information difference between multi-source data, a multi-branch structure is constructed and adaptive weight learning (AWL) is designed. In addition, according to the three-dimensional dynamic features of TC, 3D Convolutional Gated Recurrent (3D ConvGRU) is used to achieve feature enhancement, and then 3D Convolutional Neural Network (CNN) is used to capture and learn TC temporal and spatial features. Experiments on a sample of northwest Pacific TCs and official agency TC intensity prediction records are used to validate the effectiveness of our proposed model, and the results show that our model is able to focus well on the spatial and temporal features associated with TC intensity changes, with a root mean square error (RMSE) of 10.62 kt for the TC 24 h intensity forecast.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the challenge of short - term intensity prediction of tropical cyclones (TCs). Specifically, the authors point out that although significant progress has been made in tropical cyclone track prediction in recent years, intensity prediction remains a difficult problem because of its complex structure and being affected by multiple factors. To improve this situation, the authors propose a new model (AWL - Net) based on multi - source data fusion and Adaptive Weight Learning (AWL) to improve the prediction accuracy of short - term intensity changes of tropical cyclones. ### Problem Background 1. **Hazards of Tropical Cyclones**: - Tropical cyclones (such as hurricanes, typhoons) are often accompanied by disasters such as storm surges and floods when they make landfall, causing serious economic losses and casualties. - For example, Hurricane Katrina in 2005 led to 1,800 deaths in the New Orleans area of the United States and economic losses of more than $125 billion; Typhoon Haiyan, which made landfall in the central Philippines in 2013, had a wind speed of up to 195 miles per hour, causing more than 6,000 deaths and serious damage to infrastructure. 2. **Limitations of Existing Prediction Methods**: - **Non - real - time**: Although the existing reanalysis data are helpful in improving prediction accuracy, these data are not obtained in real - time and it is difficult to meet the needs for real - time prediction in actual operations. - **Weak physical consistency**: Most existing studies mainly focus on the relationship between atmospheric reanalysis data and intensity, while ignoring the influence of internal structure changes of tropical cyclones on their intensity. - **Single information processing**: In the process of multi - source data fusion, information from different sources is usually simply combined without fully considering the differences in the importance of each information source. ### Solutions To solve the above problems, the authors propose the following innovations: 1. **Multi - source data fusion**: - Use high - resolution satellite images to capture the evolution process of tropical cyclones, especially the changes in cloud structure and morphology. - Introduce a new feature - the absolute value variance of the angle between the infrared brightness temperature gradient vector and the center connection line (DAV), which is used to characterize the symmetry of tropical cyclone cloud organization and intensity changes. 2. **Adaptive Weight Learning (AWL)**: - Construct a two - branch structure to process satellite images and DAV Hovmöller diagrams respectively. Through adaptive weight learning, the weights of each branch are automatically adjusted, enabling the model to better learn the associations between different data sources. 3. **Spatial - temporal feature extraction**: - Use 3D convolutional gated recurrent units (3D ConvGRU) to capture spatial - temporal features, and combine 3D convolutional neural networks (3D CNN) for feature fusion to enhance the model's ability to represent the target. ### Model Evaluation Through experimental verification of tropical cyclone samples in the Northwest Pacific region, the results show that this model can effectively focus on the spatial - temporal features related to tropical cyclone intensity changes, and the root - mean - square error (RMSE) of 24 - hour intensity prediction is 10.62 knots (kt), showing good prediction performance. In summary, this paper aims to improve the accuracy and real - time performance of short - term intensity prediction of tropical cyclones by introducing multi - source data fusion and adaptive weight learning methods, thereby providing more effective support for disaster prevention and mitigation.

Short-Term Intensity Prediction of Tropical Cyclones Based on Multi-Source Data Fusion with Adaptive Weight Learning

Tropical Cyclone Intensity Prediction by Inter- and Intra-Pattern Fusion Based on Multi-Source Data

A Neural Network Framework for Fine-Grained Tropical Cyclone Intensity Prediction

Enhanced Rainfall Nowcasting of Tropical Cyclone by an Interpretable Deep Learning Model and Its Application in Real-Time Flood Forecasting

TCIP-Net: Quantifying Radial Structure Evolution for Tropical Cyclone Intensity Prediction

Predicting the Intensity of Tropical Cyclones over the Western North Pacific Using Dual-Branch Spatio-Temporal Attention Convolutional Network

Spatiotemporal fusion convolutional neural network: tropical cyclone intensity estimation from multisource remote sensing images

Estimation of Tropical Cyclone Intensity Using Multi-Platform Remote Sensing and Deep Learning with Environmental Field Information.

A Lightweight Multitask Learning Model with Adaptive Loss Balance for Tropical Cyclone Intensity and Size Estimation

A Data-Driven Approach for Regional-Scale Fine-Resolution Disaster Impact Prediction under Tropical Cyclones

Tropical Cyclone Intensity Prediction by Spectral-Temporal Dislocation and Attention-Based Networks

Tropical Cyclone Intensity Estimation Using Multidimensional Convolutional Neural Network From Multichannel Satellite Imagery

A Physics-informed Deep-learning Intensity Prediction Scheme for Tropical Cyclones over the Western North Pacific

Short-term prediction of the intensity and track of tropical cyclone via ConvLSTM model

Dual-Branched Spatio-temporal Fusion Network for Multi-horizon Tropical Cyclone Track Forecast

Ctransnet: A Modern Deep Learning Model to Estimate Tropical Cyclone Intensity

A Spatio-Temporal Graph-Guided Convolutional LSTM for Tropical Cyclones Precipitation Nowcasting.

Deep Learning Approaches in Predicting Tropical Cyclone Tracks: an Analysis Focused on the Northwest Pacific Region

ATDNNS: an Adaptive Time-Frequency Decomposition Neural Network-Based System for Tropical Cyclone Wave Height Real-Time Forecasting

Enhancing tropical cyclone intensity forecasting with explainable deep learning integrating satellite observations and numerical model outputs

TCP-Diffusion: A Multi-modal Diffusion Model for Global Tropical Cyclone Precipitation Forecasting with Change Awareness