Radar echo extrapolation, a critical spatiotemporal sequence forecasting task, requires precise modeling of motion trajectories and intensity evolution from sequential radar reflectivity inputs. Contemporary deep learning implementations face two operational limitations: progressive attenuation of predicted echo intensities during autoregressive inference and spectral leakage-induced diffusion at high-intensity echo boundaries. This study presents RaDiT, a hybrid architecture combining differential transformer with adversarial training for radar echo extrapolation. The framework employs a U-Net backbone augmented with vision transformer blocks, utilizing differential attention mechanisms to govern spatiotemporal interactions. Our differential attention mechanism enhances noise suppression under high-threshold conditions, effectively minimizing spurious feature generation while improving metric reliability. A conditional GAN discriminator is integrated to maintain microphysical consistency in generated sequences, simultaneously addressing spectral blurring and intensity dissipation. Comprehensive evaluations demonstrate RaDiT’s superior performance in preserving spatiotemporal coherence and intensity across 0–90 min forecasting horizons. The proposed architecture achieves CSI improvements of 10.23% and 2.88% at 4 × 4 and 16 × 16 spatial pooling scales, respectively, for ≥30 dBZ thresholds on the CMARC dataset compared to PreDiff. To our knowledge, this represents the first successful implementation of differential transformers for radar echo extrapolation.

Weather radar, as a crucial component of remote sensing data, plays a vital role in convective weather forecasting through radar echo extrapolation techniques. To address the limitations of existing deep learning methods in radar echo extrapolation, this paper proposes a radar echo extrapolation model based on video vision transformer and spatiotemporal sparse attention (ViViT-Prob). The model takes historical sequences as input and initially maps them into a fixed-dimensional vector space through 3D convolutional patch encoding. Subsequently, a multi-head spatiotemporal fusion module with sparse attention encodes these vectors, effectively capturing spatiotemporal relationships between different regions in the sequences. The sparse constraint enables better utilization of data structural information, enhanced focus on critical regions, and reduced computational complexity. Finally, a parallel output decoder generates all time step predictions simultaneously, then maps back to the prediction space through a deconvolution module to reconstruct high-resolution images. Our experimental results on the Moving MNIST and real radar echo dataset demonstrate that the proposed model achieves superior performance in spatiotemporal sequence prediction and improves the prediction accuracy while maintaining structural consistency in radar echo extrapolation tasks, providing an effective solution for short-term precipitation forecasting.

Satellite cloud images exhibit complex multidimensional characteristics, including spectral, textural, and spatiotemporal dynamics. The temporal evolution of cloud systems plays a crucial role in accurate classification, particularly under the coexistence of multiple weather systems. However, most existing models—such as those based on convolutional neural networks (CNNs), Transformer architectures, and their variants like Swin Transformer—primarily focus on spatial modeling of static images and do not explicitly incorporate temporal information, thereby limiting their ability to effectively integrate spatiotemporal features. To address this limitation, we propose SIG-ShapeFormer, a novel classification model specifically designed for satellite cloud images with temporal continuity. To the best of our knowledge, this work is the first to transform satellite cloud data into multivariate time series and introduce a unified framework for multi-scale and multimodal feature fusion. SIG-Shapeformer consists of three core components: (1) a Shapelet-based module that captures discriminative and interpretable local temporal patterns; (2) a multi-scale Inception module combining 1D convolutions and Transformer encoders to extract temporal features across different scales; and (3) a differentially enhanced Gramian Angular Summation Field (GASF) module that converts time series into 2D texture representations, significantly improving the recognition of cloud internal structures. Experimental results demonstrate that SIG-ShapeFormer achieves a classification accuracy of 99.36% on the LSCIDMR-S dataset, outperforming the original ShapeFormer by 2.2% and outperforming other CNN- or Transformer-based models. Moreover, the model exhibits strong generalization performance on the UCM remote sensing dataset and several benchmark tasks from the UEA time-series archive. SIG-Shapeformer is particularly suitable for remote sensing applications involving continuous temporal sequences, such as extreme weather warnings and dynamic cloud system monitoring. However, it relies on temporally coherent input data and may perform suboptimally when applied to datasets with limited or irregular temporal resolution.

Precipitation nowcasting aims to predict short-term precipitation evolution over forecast lead times of 1–6 h and can support hydrological-risk and disaster-prevention applications when near-real-time observations are available. However, precipitation forecasting remains challenging because of rapid spatiotemporal evolution, spatial displacement, and the difficulty of representing localized high-intensity precipitation. To address these issues, this study proposes STAMP-GAN, a spatiotemporal attention-modulated generative adversarial network for regional precipitation sequence prediction. STAMP-GAN combines an AM-ConvLSTM temporal evolution module with spatial attention, efficient channel attention, large-receptive-field context modeling, and temporal-index-conditioned feature modulation. A spatially aligned two-dimensional digital elevation model (DEM) field is retained as static auxiliary geographical information. The STAMP-Net generator uses hierarchical multi-scale feature extraction to reconstruct precipitation structures at different spatial scales while a dual-branch temporal PatchGAN provides adversarial supervision for both the complete forecast sequence and the final three forecast frames. A hybrid objective combines regression, event-based, structural, temporal, and adversarial constraints. Experiments on the ERA5 and CMA-S datasets show that, compared with the best-performing baseline for each metric, STAMP-GAN achieves relative CSI improvements of approximately 6.5% and 9.5%, respectively. The proposed framework provides a data-driven approach for retrospective hourly regional precipitation sequence prediction under gridded meteorological-data conditions, rather than a fully validated operational real-time nowcasting system.