This study quantifies the spatial heterogeneity of nonlinear signals, background noise and vertical velocities in Global Navigation Satellite System (GNSS) vertical time-series across the Tibetan Plateau (TP), using multisource loading corrections to isolate tectonic deformation. We analysed 20 yr of GNSS data (2002-2021) from Crustal Movement Observation Network of China and Nevada Geodetic Laboratory networks, processed via GipsyX and referenced to ITRF2014. Non-tidal atmospheric, oceanic and hydrological loading effects were applied utilizing operational models from Gravity Recovery and Climate Experiment (GRACE) mascon data (Center for Space Research/Jet Propulsion Laboratory/Goddard Space Flight Center), followed by common mode error (CME) filtering. The findings highlight significant spatial heterogeneity: Monsoon-dominated southern TP exhibits 10-20 per cent Root Mean Square (RMS) reduction after non-tidal atmospheric-oceanic (AO) loading corrections, while northern TP shows minimal improvement (<10 per cent), highlighting non-atmospheric noise dominance. Integration of AO and GRACE-modelled hydrological (AOG) loading corrections outperform AO+hydrological (AOH) model, achieving 25-35 per cent RMS reduction in glacier-covered Himalayas by resolving cryospheric mass loss. Spectral and principal component analysis analyses confirm AOG's superiority in suppressing interannual signals (PC1 variance: 62.7 per cent versus AOH's 60.3 per cent), particularly in monsoon-ENSO (El Ni & ntilde;o-Southern Oscillation)-affected regions. Noise modelling demonstrates high-spatiotemporal correlations (63.1 per cent white noise, WN + FN in raw data), with flicker noise (FN > 5.2 mm) linked to seismic activity in southeastern TP and power-law noise (PL > 3.5 mm) to permafrost dynamics in the north. Post-AOG_CME processing simplifies noise structures (WN + Generalized Gauss-Markov dominance: 32.9 per cent), reducing velocity uncertainties by 26.9 per cent and revealing a residual +1.2 mm yr(-1) uplift in the southern inner TP, indicative of mid-crustal flow. Persistent uncertainties (>0.55 mm yr(-1)) along the Himalayan thrust front correlate with deep lithospheric boundaries. Our findings demonstrate the necessity of integrating GRACE-derived corrections with CME filtering to accurately delineate tectonic signals within the intricate suture zones of the TP, offering crucial insights into plateau-wide geodynamic processes.
True Digital Orthophoto Maps (TDOMs) are essential products for digital twins and Geographic Information Systems (GIS). Traditionally, TDOM generation involves a complex photogrammetric pipeline, which may deteriorate due to various challenges, including inaccurate Digital Surface Model (DSM), unreliable occlusion detections, and visual artifacts in weakly textured regions and reflective surfaces, etc. To address these challenges, we introduce Tortho-Gaussian, a novel method inspired by 3D Gaussian Splatting (3DGS) that generates TDOMs through orthogonal splatting of optimized anisotropic Gaussian kernels. More specifically, we first simplify the orthophoto generation by orthogonally splatting the Gaussian kernels onto 2D image planes, formulating a geometrically elegant solution that eliminates the need for an explicit DSM and occlusion detection. Second, to produce TDOM of large-scale area, a divide-and-conquer strategy is adopted to optimize memory usage and time efficiency of both training and rendering for 3DGS. Lastly, we design a fully anisotropic Gaussian kernel that adapts to the varying characteristics of different regions, particularly improving the rendering quality of reflective surfaces and slender structures. Extensive experimental evaluations demonstrate that our method outperforms existing commercial software in several aspects, including the accuracy of building boundaries and visual quality in low-texture regions and on building facades. These results underscore the potential of our approach for large-scale urban scene reconstruction, offering a robust alternative for enhancing TDOM quality and scalability. Project Web: https://github.com/xwangSGG/Tortho_Gaussian.
Pre-seismic anomalies are attracting more and more attention in seismology. However, this issue has not been thoroughly researched. Superconducting gravimeters have high observation accuracy (10-9 similar to 10-10 m/s2) and can be used to study low-frequency pre-seismic gravity anomalies (the contribution of the internal mass redistribution). We extracted the pre-seismic gravity anomalies of 2010 Maule earthquake (Mw 8.8) and 2011 Tohoku-Oki earthquake (Mw 9.0) from the data of Global Geodynamics Project network. Through power spectrum density analysis and band-pass filtering of the gravity data, we found pre-seismic gravity anomalies continued for 126 h before the Maule earthquake, and the maximum reached 6 lgal. For the Tohoku-Oki earthquake, the pre-seismic anomalies continued for 89 h before the main shock, and the maximum anomaly was 26.5 lgal. Furthermore, all the SGs in Europe had detected clear pre-seismic gravity changes more than 3 lgal, but some other stations not. This discrepancy may be due to differences in local geological structure and sensitivity, despite the gravimeters having the same observation accuracy. From the results of great earthquakes, we found the gravimeters in Europe have a high sensitivity in detecting pre-seismic gravity anomalies. (c) 2025 The Author(s). Published by Elsevier B.V. on behalf of COSPAR. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).
Federated learning is a machine learning paradigm through training on locally private data and aggregating local models to generate a federated model. However, due to the heterogeneity problems (data heterogeneity and model heterogeneity), federated learning suffers from convergence difficulties and excessive aggregation overhead on decentralized participants. Additionally, federated learning faces privacy concerns during the aggregation of local models. To this end, in this work, we propose FEDINF, an efficient and secure inference with federated participants. Specifically, FEDINF features the following characteristics. FEDINF overcomes convergence challenges through federated inference instead of federated training, which reduces computation and communication overhead. Moreover, we design secure computation protocols and aggregation mechanisms to measure contributions, and handle both data and model heterogeneity without sacrificing privacy. Results of experimental evaluations on common datasets demonstrate that the proposed FEDINF outperforms the existing federated learning approaches in terms of efficiency and heterogeneity.
Recently, large vision-language models (VLMs) have made marked strides in vision-language capabilities through visual instruction tuning, showing great promise in the field of remote sensing image interpretation. However, existing remote sensing vision-language models (RSVLMs) often fall short in capturing the complex characteristics of remote sensing scenes. While some recent RSVLMs, such as EarthGPT and LHRSBot, have started to incorporate multiscale features, they typically adopt relatively shallow fusion strategies and lack tightly integrated mechanisms for fine-grained vision-language alignment. In this paper, we present Aquila, an advanced vision-language foundation model designed to enable richer visual feature representation and more precise vision-language feature alignment for remote sensing images. Our approach introduces a learnable hierarchical spatial feature integration (SFI) module that supports high-resolution image inputs and aggregates multi-scale visual features, allowing for the detailed representation of complex visual information. Additionally, the SFI module is repeatedly integrated into the layers of the large language model (LLM) to achieve deep vision-language feature alignment, without compromising the model's performance in natural language processing tasks. These innovations significantly improve the model's ability to learn from image-text data. We validate the effectiveness of Aquila through extensive quantitative experiments and qualitative analyses, demonstrating its superior performance.
