Analyzing the impervious surface area (ISA) of China’s three major urban agglomerations and expansion pattern of representative cities is of great significance for formulating context-specific urban planning policies and enhancing the resilience of urban development. This research employs impervious surface expansion indicators to reveal the spatiotemporal dynamics of impervious surface area, and further applies the location centrality index (LCI) to investigate the expansion structures and patterns of five megacities. The results indicate that the ISA of China’s three major urban agglomerations expanded from 22,544.10 km² to 69,348.59 km² from 1985 to 2024; both the expansion rate and intensity have exhibited a distinct decelerating trend. However, significant regional variation exists in the peak expansion phases: the Guangdong–Hong Kong–Macao Greater Bay Area Urban Agglomeration (GBA) peaked during the Initial Exploratory Phase (1985–1995), whereas the Yangtze River Delta Urban Agglomeration (YRD) and Beijing–Tianjin–Hebei Urban Agglomeration (BTH) reached their peaks during the High-speed Development Phase (2005–2015). At the urban scale, LCI-based analysis of urban expansion across five megacities reveals that adjacency expansion type has become the predominant form of growth, exceeding 50% in most cities. These findings enhance our understanding of evolutionary trajectories of ISA within urban agglomerations and the expansion patterns of five megacities, providing a scientific basis for sustainable urban planning and urban resilience.

Assessing wildfire susceptibility requires integrating environmental and anthropogenic factors to quantify the probability and vulnerability of fires in a given area. Many existing machine-learning models offer high predictive power but limited interpretability, restricting their utility for operational decision-making. This study is the first to apply the intrinsically interpretable deep network TabNet to wildfire susceptibility modeling. By fusing multi-source data and leveraging TabNet’s feature-mask matrix, we achieve accurate prediction and built-in explanation without relying on auxiliary tools. On a dataset of 133,811 samples, the proposed model achieves an Area Under the Curve (AUC) of 0.760, recall of 0.883, precision of 0.395, and an F1.5 score of 0.640, outperforming XGBoost (version 1.5.0) and other baseline models. The importance rankings derived from the feature-mask matrix align with the Shapley Additive Explanations (SHAP) results, confirming the reliability of the explanations. This approach combines predictive accuracy with transparency, providing a deployable framework for wildfire early warning, risk management, and ecosystem conservation.

This paper presents the second generation of the China Seismo-Electromagnetic Satellite (CSES) Global Geomagnetic Field Model (CGGM-2), specifically designed to derive candidate models for the fourteenth generation of the International Geomagnetic Reference Field (IGRF-14). To build this CGGM-2 model, we utilize more than six years of geomagnetic field data collected by the high-precision magnetometer (HPM) onboard CSES. The model is parametrized by expanding the internal field to spherical harmonic degree and order 45, now incorporating lithospheric contributions. This was not the case for the first generation CGGM model, which was limited to spherical harmonic degrees up to 15. We now also implement order six B-splines in time with one year knot spacing for degrees 1 to 13 to capture non-linear temporal variations in the core field, superseding the linear approximation used in CGGM. Despite a number of challenges, such as boom deformation on which the HPM is located (away from the star imager that provides attitude information) and high-latitude magnetic field platform disturbances, the CGGM-2 model succeeds at capturing the non-linear temporal variations of the geomagnetic field. In particular, the SV-2025-2030 candidate model derived from CGGM-2 performs particularly well among all 18 candidate models for IGRF-14. Candidate models derived from the CGGM-2 are also the only IGRF-14 candidate models using CSES data only and not directly relying on any ESA’s Swarm satellite data, thus providing useful candidate models to compare against all other IGRF candidate models. Our findings underscore the potential of CSES data for geomagnetic field modeling and the likely benefits of CSES’s revisiting capability for accurately capturing the main field's variations. It is expected that future data from the forthcoming companion CSES-02 mission will help improve the temporal–spatial resolution of the model and mitigate high-latitude magnetic platform disturbances. Graphical Abstract

Computing power network (CPN) is designed to utilize multi-dimensional resources to complete computing tasks. However, in practical applications, the CPN architecture has difficulty in coordinating cross-domain heterogeneous resources, making it impossible to achieve the real-time and high scalability requirements of computationally intensive and time-sensitive tasks such as levee piping hazard inspection via remote sensing in emergency scenarios. Based on this, we propose a communication and computation integrated network architecture, referred to as (Com)²INet, that integrates “sensing”, “transmission”, and “computation” phases. In the sensing phase, thermal infrared imagery is utilized to retrieve land surface temperature fields through radiative transfer mechanisms, providing a reliable foundation for visual segmentation of piping hazards. In the transmission phase, we adopt the designed multi-path transmission mechanism to promote the efficient data flow across heterogeneous networks. In the computation phase, the proposed SACM algorithm, which is functionally decomposed and implemented as service chains within the proposed network architecture, dynamically processes the retrieved temperature fields to achieve precise hazard identification. This integrated framework ensures seamless interaction between sensing, communication, and computation, addressing the challenges of real-time hazard detection in emergency scenarios.

Remote sensing image super-resolution requires local detail and wider structural context. The mixture-of-experts (MoE) designs compared here select or combine expert outputs without explicitly passing behavior statistics to subsequent spatial–frequency restoration. We propose EBSFFormer, which aggregates routing preferences and response-weighted proportions within restoration groups and converts them into channel gains for later spatial and frequency branches. Heterogeneous experts and a content–cumulative-change router supply these statistics. We train on AID and evaluate on AID, DIOR, DOTA, and NWPU-RESISC45 under bicubic $\times 4$ degradation. Across three AID training seeds, EBSFFormer obtains 31.2639 ± 0.0009 dB peak signal-to-noise ratio (PSNR), with mean improvements of 0.0121 dB over MambaIRv2 and 0.1037 dB over SwinIR. Single-checkpoint evaluation on NWPU without fine-tuning gives the highest Y-channel PSNR and structural similarity index (SSIM) among the compared methods and the third-lowest learned perceptual image patch similarity (LPIPS). Component ablations, red–green–blue (RGB) metrics, and ground-truth-referenced visual comparisons characterize the reconstruction gains. EBSFFormer has fewer parameters and lower mean latency than MambaIRv2, but the highest peak allocated memory in the runtime comparison.