Xiaoqin Lu

This study proposes SOPHIENet (TC Satellite‐Observed Physical‐Persistence and Spatiotemporal Hybrid Intensity Estimation Network), a hybrid framework for Tropical Cyclone (TC) intensity estimation. SOPHIENet integrates a convolutional neural network, the Convolutional Block Attention Module, and a physical‐persistence adjustment mechanism. By extracting spatiotemporal features from multi‐temporal satellite infrared (IR) imagery, SOPHIENet captures TC cloud‐structure evolution. The embedded attention module refines channel‐wise and spatial representations, improving the feature‐learning capability of the deep learning (DL) component. A key innovation is the physical‐persistence adjustment mechanism inspired by the Dvorak technique, which converts historical TC intensity‐evolution characteristics over the Western North Pacific (WNP) into multi‐timescale constraints on intensity change. These constraints suppress physically unreasonable short‐term fluctuations in the initial DL estimates, improving the temporal consistency, physical plausibility, and reliability. In an independent operational evaluation using FY‐4B IR observations for 43 TCs (20,144 samples) over the WNP during 2023–2024, SOPHIENet achieved a mean absolute error (MAE) of 2.8 m/s. A temporally collocated comparison of 1,841 samples from 26 TCs in 2024 yielded an MAE of 2.7 m/s for SOPHIENet against China Meteorological Administration best‐track 2‐min maximum sustained wind (MSW), compared with 3.4, 3.6, and 4.0 m/s for D‐PRINT, AiDT, and Advanced Dvorak Technique v9.0, respectively, against Joint Typhoon Warning Center best‐track 1‐min MSW. The comparison is not strictly homogeneous because of differences in wind‐averaging conventions, best‐track references and TC center inputs. More broadly, SOPHIENet also showed competitive performance relative to existing DL approaches and stable performance across TC intensity categories, with slight underestimation for super typhoons and greater uncertainty during rapid intensity evolution.

This paper introduces a comprehensive Tropical Cyclone (TC) dataset based on China’s FengYun (FY) series geostationary satellite observations in the Western North Pacific (WNP), designated as FYSATTY. This innovative dataset integrates high-resolution geostationary meteorological satellite observations with best-track data, providing a unique resource for TC monitoring and analysis in the WNP region. FYSATTY is derived from infrared, water vapor, and visible channel imagery captured by the FY-2 and FY-4 satellites, which is processed using an equal-interval latitude-longitude grid system within a 1500-km radius centered on each TC. The best-track data, sourced from the China Meteorological Administration (CMA), is interpolated into 3-h intervals to ensure temporal consistency with the satellite observations. This dataset consists of 99714 observations for 614 TCs from 2005 to 2024 and will be updated annually. It represents the most comprehensive collection of TC satellite imagery derived from the FY series of geostationary meteorological satellites of the National Satellite Meteorological Center (NSMC). It encompasses a synthesis of infrared, water vapor, and visible channel data from the FY-2 and FY-4 satellite series. The dataset has been meticulously cropped and georeferenced using the CMA best-track data, with fundamental geographic spatial calibration completed. Comparative analysis with analogous elements within other TC geostationary meteorological satellite datasets reveals a high degree of consistency. It is suitable for reanalyzing TC best-track data, assessing climatic characteristics of TC satellite observations, and enhancing forecasting capabilities. This dataset is particularly useful for TC research, forecasting, and machine learning, and is publicly available.

Ocean surface wind vector (OWV) is a key variable for ocean remote sensing and tropical cyclone (TC) monitoring. This study presents the first comprehensive intercomparison of Ku-band OWV products from FY-3E/WindRAD and HY-2B/SCA scatterometers using full-year data from 2022 (583,805 spatiotemporal collocations), with both sensors sampling the morning–evening local-time sector in sun-synchronous orbits. Results indicate strong agreement in wind speed (R = 0.95; mean bias −0.47 m/s; RMSE 1.30 m/s) and wind direction (mean bias 0.22°; std 28.13°) for wind speeds ≥ 3.4 m/s (Beaufort scale B3 and above), with the highest consistency across Beaufort scale 3–8 (B3–B8); however, at wind speeds greater than 20.8 m/s (B9) the bias increases. A fusion leveraging FY-3E’s fine resolution and HY-2B’s wide coverage is implemented and applied to Super Typhoon Hinnamnor (2022), enhancing the spatial coverage and structural detail of TC winds. Quadrant 34 kt wind radii (R34) are estimated from the fused wind fields and evaluated against the best-track data from the Joint Typhoon Warning Center (JTWC), showing close agreement during compact, symmetric TC stages but larger differences during structural reorganization. Overall, the findings confirm inter-satellite consistency for the two Chinese scatterometers and demonstrate the practical value of a multi-source fusion approach that benefits TC monitoring, wind radii estimation, and marine weather services.

Accurate identification of the center of a tropical cyclone (TC) is crucial for TC forecasting and early warning. For weak TCs [i.e., maximum sustained wind (MSW) ≤ 17 m s ⁻¹ ], the determination of their near-surface centers is more challenging due to weaker central convection and less pronounced circulation features. Existing objective TC positioning methods often exhibit limited performance for weak TCs, and few are specifically designed for weak TCs. The Automated Rotational Center Hurricane Eye Retrieval (ARCHER) algorithm, which locates TC centers by jointly finding the spiral cloud band center and assumed eye center in satellite images, is a relatively accurate and objective approach. However, it was designed primarily for use with TCs of at least tropical storm intensity (MSW ≥ 17.2 m s ⁻¹ ), and it is less effective for weaker TCs. This study analyzed 8000 TC samples in the western North Pacific (WNP) during 2000–12 to develop M_ARCHER, a modified version of the ARCHER algorithm with adjusted parameters to locate the centers of weak TCs in the WNP. M_ARCHER employs two sets of new optimized parameters tailored for microwave and infrared satellite observations. These parameters include the inclination angle of the spiral cloud band pattern, range and increment for determining the assumed eyewall inner edge radius, and the weights assigned to spiral and eye centering, among others. These modifications improve the algorithm’s capability to detect the asymmetric and less-defined structures of weak TCs. Independent tests of the M_ARCHER algorithm show median location errors for weak TCs of 44 km (microwave) and 57 km (infrared) relative to the China Meteorological Administration best track data. For readily available infrared observations, M_ARCHER shows a 20% improvement in location accuracy relative to the original algorithm. Notably, strong wind shear can disrupt the structure of a TC, making the spiral-eye configuration within the M_ARCHER unrealistic and reducing the location accuracy. Significance Statement: Accurately locating the centers of weak tropical cyclones (TCs) remains a challenge in meteorology, and existing methods commonly generate substantial errors. This study aims to improve the accuracy associated with locating the center of a weak TC. We modified the Automated Rotational Center Hurricane Eye Retrieval (ARCHER) algorithm by adjusting its parameters, thereby optimizing its ability to locate weak TCs. Test results indicate that this modified algorithm (M_ARCHER) reduces the location error by 20% when using infrared observations. This approach increases the location accuracy of weak TCs and will greatly improve meteorological monitoring and early warning systems. Future studies should explore the performance of M_ARCHER with respect to various meteorological conditions and extreme weather events.

Current deep learning methods for predicting tropical cyclone intensity are restricted by inadequate utilization of climate factors, insufficient feature fusion, and the inability to automatically adapt input data to changes in the scale of cyclone structures. To address these challenges, this article proposes a multimodal data fusion model guided by sequential infrared satellite images (IRGMF). The IRGMF model integrates a broader range of climate factors for a more comprehensive, enriched feature representation of tropical cyclone intensity. In addition, we designed a temporal attention module that allows the model to focus on critical moments in the sequential data for a more precise capture of the dynamic features of climate factors as they evolve. To enhance prediction accuracy, we also developed an attention-cropping module that adaptively crops irrelevant information outside the scale of a tropical cyclone from reanalysis data. To validate the effectiveness of the proposed model, we conducted experiments on multiple datasets, including CMA-BST, ERA5, and GridSat-B1. Experimental results show that, compared to existing methods, the proposed IRGMF model demonstrates significant advantages in predicting tropical cyclone intensity and provides superior accurate information for meteorological forecasting and decision support.

Typhoons are extreme weather events that pose significant threats to human life and property. Sequential satellite imagery of typhoon clouds, which is rich in spatio-temporal information, plays a critical role in understanding their formation, development, and evolutionary dynamics. Recently, the rapid advancement of deep learning that combines physical insights with data-driven has opened new avenues for research in Earth science. In this study, for high-quality prediction of sequential typhoon cloud images, we propose a physics-enhanced deep learning model termed C²PhyNet. Specifically, we introduce a novel disentangling spatio-temporal block integrated with a criss-cross physics-enhanced unit. To further improve the fine structural details in the predicted typhoon cloud images, a concurrent spatial and channel squeeze-and-excitation attention mechanism is incorporated into both the encoder and decoder modules. Our quantitative analysis demonstrates the superiority of the proposed approach over existing sequential image prediction models on the publicly available Digital Typhoon dataset. The experimental results show that our method achieves superior performance, with a Structural Similarity Index Measure (SSIM) of 0.8200 and a Peak Signal-to-Noise Ratio (PSNR) of 23.26. C²PhyNet is capable of generating high-quality sequential typhoon cloud images, which can significantly enhance the ability of meteorologists to forecast typhoon-related details with greater accuracy. Furthermore, our research contributes to improved risk mitigation and more effective disaster warning and management strategies.

Tropical cyclone (TC) is a highly catastrophic weather event, and accurate estimation of intensity is of great significance. The current proposed TC intensity estimation model focuses on training using satellite images from a single or two channels, and the model cannot fully capture features related to TC intensity, resulting in low accuracy. To this end, we propose a double-layer encoder-decoder model for estimating the intensity of TC, which is trained using images from three channels: infrared, water vapor, and passive microwave. The model mainly consists of three modules: wavelet transform enhancement module, multi-channel satellite image fusion module, and TC intensity estimation module, which are used to extract high-frequency information from the source image, generate a three channel fused image, and perform TC intensity estimation. To validate the performance of our model, we conducted extensive experiments on the TCIR dataset. The experimental results show that the proposed model has MAE and RMSE of 3.76m/s and 4.62m/s for TC intensity estimation, which are 15.70% and 20.07% lower than advanced Dvorak technology (ADT), respectively. Therefore, the model proposed in this article has great potential in accurately estimating TC intensity.

Tropical cyclone (TC) is a major devastating natural disaster that leads to property destruction worth billions of dollars and threatens millions of lives. However, rapid changes in TC are the main source of the current TC forecast error. This study proposes a model for estimating the rapid weakening of TC intensity based on deep learning using infrared satellite images and sea surface temperature (DEEP_RW_TCIE). This model is in two parts: a many-to-many TC intensity estimation network, composed of spatiotemporal code and decode; and a network with multi-layer perception as the core for constraining TC intensity estimation sequence, based on sea surface temperature (SST) and intensity change rate (ICR). We investigated the effects of different time series lengths, different ranges of SST, and different feature vector composition methods on the effect of the rapid weakening (RW) of the TC intensity estimation model. Moreover, we verified the rationality and feasibility of the proposed method through the analysis of experimental methods. The results show that the TC intensity at several moments before rapid TC weakening is of great significance for estimating a current rapid TC weakening, SST, and ICR. Our method greatly reduces the estimating error of the rapid TC weakening intensity. The mean of absolute errors (MAE) and the root mean square error (RMSE) are 6.68 kt and 8.68 kt, respectively, which decrease by over 10%, compared to the benchmark convolutional neural network CNN-TC model.

Tropical cyclone (TC) size is an important parameter for estimating TC risks, such as precipitation distribution, gale-force wind damage, and storm surge. This paper uses the TC size dataset compiled by the Shanghai Typhoon Institute of China Meteorological Administration (STI/CMA) to investigate the interannual, monthly variation in TC size, and the relationships between TC size and intensity in the WNP basin from 1981 to 2009. The results show that the annual mean TC size oscillated within the range of 175–210 km from 1981 to 2002, then decreased following 2003. For the monthly average TC size, there are two peaks in September and October. The TC size, overall, becomes larger with increasing intensity; the samples with an unusually large size are mainly concentrated near a 40 m s⁻¹ intensity. After the TC intensity exceeds 40 m s⁻¹, the number of unusually large size samples gradually decreases. About 60% of the TCs reach their maximum size after reaching the peak intensity, and the average lag time is 8.3 h.