Search

Abstract Due to the powerful ability in capturing the global information, transformer has become an alternative architecture of CNNs for hyperspectral image classification. However, general transformer mainly considers the global spectral information while ignores the multiscale spatial information of the hyperspectral image. In this paper, we propose a multiscale spectral–spatial convolutional transformer (MultiFormer) for hyperspectral image classification. First, the developed method utilizes multiscale spatial patches as tokens to formulate the spatial transformer and generates multiscale spatial representation of each band in each pixel. Second, the spatial representation of all the bands in a given pixel are utilized as tokens to formulate the spectral transformer and generate the multiscale spectral–spatial representation of each pixel. Besides, a modified spectral–spatial CAF module is constructed in the MultiFormer to fuse cross‐layer spectral and spatial information. Therefore, the proposed MultiFormer can capture the multiscale spectral–spatial information and provide better performance than most of other architectures for hyperspectral image classification. Experiments are conducted over commonly used real‐world datasets and the comparison results show the superiority of the proposed method.

Abstract North China experienced an extreme precipitation event from July 29 to August 1, 2023 (i.e., the “23.7” event) causing severe floods, significant infrastructure damage and multiple fatalities. To enhance comprehension of the mechanism behind the extreme precipitation of the “23.7” event, water vapor transport paths and sources were determined, and water vapor contribution of each source was quantitatively evaluated based on Lagrangian methods. Results showed that the extreme precipitation of the “23.7” event was closely related to large‐scale water vapor transport and convergence from low‐latitude oceans. There were five main water vapor sources which corresponded to five transport pathways. Path 1 was derived from tropical West Pacific, containing the most trajectories (195), carrying the most water vapor (69.3%) and contributing the most to the extreme precipitation of the “23.7” event (45.7%). Path 2 was guided by the cross‐equatorial flow through South China Sea, contributing to 10.1% of the precipitation. Path 3 originating from eastern tropical Indian Ocean and Path 4 from the west source near the Caspian Sea contributed less to the precipitation. Last but not the least, water vapor evaporation from eastern China contributed more than 30% to the extreme precipitation, making this region another important water vapor source.

Abstract Unlike prior researches focusing on interannual or interdecadal changes of the Northwest Pacific subtropical high (NWPSH), this paper emphasizes the interdecadal changes in the relationship between May‐June and July‐August NWPSH. The correlation coefficients between them are 0.66 (passing the 99.9% confidence level) and 0.12 during 1979–2002 and 2003–2023. Therefore, the May‐June NWPSH has a strong connection with July‐August East Asian precipitation during 1979–2002, but not during 2003–2023. The interdecadal variations in the relationship of NWPSH are due to interdecadal changes of ENSO. During 1979–2002, the eastern‐type ENSO in prior winter, having longer durations, can influence May‐June and July‐August NWPSH and the North Indian Ocean. However, the central‐type ENSO in prior winter only persists it's influence until May‐June, and changes into a Pacific Dipole pattern in July‐August during 2003–2023. Despite the North Indian Ocean continually warms during 2003–2023, the tropical Pacific Dipole offsets its impacts on the NWPSH.

Abstract Deep Neural Networks have been successfully applied in hyperspectral image classification. However, most of prior works adopt general deep architectures while ignore the intrinsic structure of the hyperspectral image, such as the physical noise generation. This would make these deep models unable to generate discriminative features and provide impressive classification performance. To leverage such intrinsic information, this work develops a novel deep learning framework with the noise inclined module and denoise framework for hyperspectral image classification. First, the spectral signature of hyperspectral image is modeled with the physical noise model to describe the high intra‐class variance of each class and great overlapping between different classes in the image. Then, a noise inclined module is developed to capture the physical noise within each object and a denoise framework is then followed to remove such noise from the object. Finally, the CNN with noise inclined module and the denoise framework is developed to obtain discriminative features and provides good classification performance of hyperspectral image. Experiments are conducted over two commonly used real‐world datasets and the experimental results show the effectiveness of the proposed method. The implementation of the proposed method and other compared methods could be accessed at https://github.com/shendu‐sw/noise‐physical‐framework.

Abstract Compared to interannual and interdecadal variations, inadequate attention has been given to intraseasonal changes in the summer northwest Pacific subtropical high (NWPSH). During 1981–2005, the north Indian Ocean (NIO) and NWPSH exhibit significant relationships from June to August, while the influences of the Boreal Summer Intraseasonal Oscillation 1 (BSISO1) in June and August differ from its impacts in July, resulting in a significant correlation between the NWPSH in June and August. The models capture NIO's roles but underestimate BSISO1's influences, leading to an underestimation (overestimation) of the relationship between NWPSH in June and August (adjacent two months). A dynamical‐statistical approach based on the close relationship between June and August NWPSH is proposed to improve models' prediction capabilities. Improved results show the mean correlation coefficients (root mean square errors) between prediction and observation for August precipitation and geopotential height increase by 57% and 21% (decrease by 0.27 mm/day and 1.61 m).