A CatBoost-based intelligent tropical cyclone (TC) intensity-detecting model was built to quantify the intensity of TCs over the Western North Pacific (WNP) with the cloud-top brightness temperature (CTBT) data of Fengyun-2F (FY-2F) and Fengyun-2G (FY-2G) and the best-track data of the China Meteorological Administration (CMA-BST) in recent years (2015–2018). The CatBoost-based model was featured with the greedy strategy of combination, the ordering principle in optimizing the possible gradient bias and prediction shift problems, and the oblivious tree in fast scoring. Compared with the previous studies based on the pure convolutional neural network (CNN) models, the CatBoost-based model exhibited better skills in detecting the TC intensity with the root mean square error (RMSE) of 3.74 m s⁻¹. In addition to the three mentioned model features, there are also two reasons for the model’s design. On one hand, the CatBoost-based model used the method of introducing prior physical factors (e.g., the structure and shape of the cloud, deep convections, and background fields) into its training process. On the other hand, the CatBoost-based model expanded the dataset size from 2342 to 13,471 samples through hourly interpolations of the original dataset. Furthermore, this paper investigated the errors of this model in detecting the different categories of TC intensity. The results showed that the deep learning-based TC intensity-detecting model proposed in this paper has systematic biases, namely, the overestimation (underestimation) of intensities in TCs which were weaker (stronger) than at the typhoon level, and the errors of the model in detecting weaker (stronger) TCs were smaller (larger). This implies that more factors than the CTBT should be included to further reduce the errors in detecting strong TCs.

Tropical cyclones (TCs) seriously threaten the safety of human life and property especially when approaching a coast or making landfall. Robust, long-lead predictions are valuable for managing policy responses. However, despite decades of efforts, seasonal prediction of TCs remains a challenge. Here, we introduce a deep-learning prediction model to make skillful seasonal prediction of TC track density in the Western North Pacific (WNP) during the typhoon season, with a lead time of up to four months. To overcome the limited availability of observational data, we use TC tracks from CMIP5 and CMIP6 climate models as the training data, followed by a transfer-learning method to train a fully convolutional neural network named SeaUnet. Through the deep-learning process (i.e., heat map analysis), SeaUnet identifies physically based precursors. We show that SeaUnet has a good performance for typhoon distribution, outperforming state-of-the-art dynamic systems. The success of SeaUnet indicates its potential for operational use.

Arctic sea ice has a significant effect on global climate change, ship navigation, Arctic ecosystems, and human activities. Therefore, it is essential to produce high-resolution sea ice maps that accurately represent the geographical distribution of various sea ice types. Based on deep learning technology, many automatic sea ice classification algorithms have been developed using synthetic aperture radar (SAR) imagery over the last decade. However, sea ice classification faces two vital challenges: (1) it is difficult to distinguish sea ice types with close developmental stages solely from SAR images and (2) an imbalanced sea ice dataset has a significantly negative effect on ice classification model performance. In this article, a spatio-temporal deep learning model—the Dynamic Multi-Layer Perceptron (MLP)—is utilized to classify 10 sea ice types automatically. It consists of a SAR image branch and a spatio-temporal branch, which extracts SAR image features and spatio-temporal distribution characteristics of sea ice, respectively. By projecting similar image features to different positions in the spatio-temporal feature space dynamically, the Dynamic MLP model effectively distinguishes between similar sea ice types. Furthermore, to reduce the impact of data imbalance on model performance, the dynamic curriculum learning (DCL) method is used to train the Dynamic MLP model. Experimental results demonstrate that our proposed method outperforms the long short-term memory (LSTM) network approach in distinguishing between sea ice types with similar developmental stages. Moreover, the DCL training method can also effectively improve model performance in identifying minority ice types.

Despite numerous studies, the impact of global warming on the tropical cyclone (TC) track and intensity by reasons of data inhomogeneity in remote sensing and large natural variability over a relatively short period of observation is still controversial. Three carbon-emission sensitivity experiments are conducted to investigate how TC track and intensity respond to changes in the oceanic and atmospheric environment under global warming. The results show a high sensitivity of the simulated TC track and intensity to global warming. On one hand, with increase in carbon emissions, the western Pacific subtropical high expands notably, increasing the poleward steering flow and eventually leading to a poleward shift of TC. On the other hand, the underlying sea-surface temperature and surface-entropy flux increase and, thus, favor the convections near the eyewall. Moreover, the TC structure becomes more upright, which is closely related to the larger pressure gradient near the eyewall. As a result, TC intensity increases with carbon emissions. However, this increase is notably smaller than the maximum potential intensity theory as the TC intensity can reach a threshold if carbon emission still increases in the future. The involved mechanisms on the changes of TC track and intensity are also revealed.

The integration of high-resolution aerosol measurements into an atmospheric chemistry model can improve air quality forecasting. However, traditional data assimilation methods are challenged in effectively incorporating such detailed aerosol information. This study utilized the WRF-Chem model to conduct data assimilation and prediction experiments using the Himawari-8 satellite’s aerosol optical depth (AOD) product and ground-level particulate matter concentration (PM) measurements during a record-breaking dust event in the Beijing–Tianjin–Hebei region from 14 to 18 March 2021. Three experiments were conducted, comprising a control experiment without assimilation (CTL), a traditional three-dimensional variational (3DVAR) experiment, and a multi-scale three-dimensional variational (MS-3DVAR) experiment. The results indicated that the CTL method significantly underestimated the intensity and extent of the severe dust event, while the analysis fields and forecasting fields of PM concentration and AOD can be significantly improved in both 3DVAR and MS-3DVAR assimilation. Particularly, the MS-3DVAR assimilation approach yielded better-fitting extreme values than the 3DVAR method, mostly likely due to the multi-scale information from the observations used in the MS-3DVAR method. Compared to the CTL method, the correlation coefficient of MS-3DVAR assimilation between the assimilated PM₁₀ analysis fields and observations increased from 0.24 to 0.93, and the positive assimilation effect persisted longer than 36 h. These findings suggest the effectiveness and prolonged influence of integrating high-resolution aerosol observations through MS-3DVAR assimilation in improving aerosol forecasting capabilities.

Due to limited in-situ ocean observations, reanalysis data are often considered as an important source for studying tropical cyclone (TC)-induced ocean waves. Here, we introduced a method to quantitatively evaluate the errors of TC-induced ocean waves in reanalysis datasets. The TC data are from the IBTrACS dataset. We compared TC-induced ocean waves in two reanalysis datasets (i.e., ERA5 and WAVERYS) with those in buoy data when TCs are near the buoy stations. We showed that the errors of TC-induced ocean waves in WAVERYS and ERA5 are similar, because the surface winds in these two datasets are the same. We noticed that the errors in the North Pacific are remarkably smaller than those in the North Atlantic due to more realistic probability density of TC-induced ocean waves in the North Pacific compared with those in the North Atlantic. Moreover, the errors are strongly related to significant wave height (SWH) and the distance from TC center. The larger the SWH and the shorter the distance, the larger the error. Furthermore, the errors in TC-induced ocean waves decreased significantly over the past decades.