Remote Sensing Application Center

Introduction Karst regions in Southwest China are ecologically fragile and face complex governance challenges during rapid socio-economic transformation. Clarifying the evolutionary characteristics and governance constraints of ecological civilization construction in these regions is critical for promoting sustainable development and precision governance. Methods This study developed a multidimensional evaluation framework comprising 29 indicators to assess ecological civilization construction in typical karst regions. Using Guizhou Province as a case study, construction levels were systematically measured at the provincial, municipal, and county scales for three key temporal nodes (2012, 2018, and 2023). Spatial autocorrelation models were applied to reveal spatio-temporal differentiation patterns and spatial dependence characteristics. Results The results show that from 2012 to 2023, the comprehensive ecological civilization index in Guizhou Province increased significantly, with an average annual growth rate of 10.02%, indicating the effectiveness of the green transition during the modernization process. However, pronounced spatial non-equilibrium persists, forming a county-level pattern characterized by “high-value clustering in the southeast and low-value clustering in the west,” accompanied by a continuously strengthening positive spatial autocorrelation. Discussion Further analysis reveals that the observed spatial disparities are driven by deep-seated mechanisms, including governance cost differentials imposed by karst habitat constraints, path dependency of resource-based industries, and cultural disconnection during rapid urbanization. These findings provide a theoretical basis for differentiated and precision-oriented governance strategies in ecologically fragile regions and offer a valuable “China Solution” for achieving global sustainable development goals in similar karst environments.

Accurate monitoring of total suspended matter (TSM) concentration is essential for aquatic ecosystem protection and water quality assessment. Multispectral remote sensing provides an effective approach for large-scale TSM monitoring. However, robust retrieval models are difficult to develop due to limited in situ data. This study presents a Deep Feature Extraction–Machine Learning fusion framework that integrates a pre-trained back-propagation neural network (BPNN) with support vector regression (SVR) to enhance TSM retrieval. High-level spectral features extracted by BPNN are used as inputs to SVR (termed DFE-SVR) for regional TSM retrieval, using in situ measurements from five inland lakes in Jiangsu and Anhui Provinces, China. The generated TSM maps showed spatial patterns consistent with TSM concentration distributions visually observed in true-color imagery. Validation results demonstrated that DFE-SVR outperformed BPNN and SVR models, achieving R2 of 0.85 and 0.90 and RMSE of 7.95 and 4.76 mg/L for GF-1 and Sentinel-2 imagery, respectively. Compared with SVR models using principal component analysis or band combinations, DFE-SVR reduced RMSE by over 20%. Under reduced training samples, the DFE-SVR model also maintained higher stability and accuracy. These findings showed its potential for multispectral water quality monitoring with limited in situ data.

Abstract The escalating frequency and magnitude of floods pose significant challenges to traditional flood control measures with fixed defense capacities, such as levees, dams, and reservoirs. Flood Detention Areas (FDAs) have been widely implemented in many countries as solutions to extreme flood events, functioning by temporarily holding excess floodwaters and serving as farmland and residential zones during periods of inactivity. However, uncertainties regarding their flood control effectiveness, concerns over post‐activation livelihood recovery, and the lack of management guidelines have limited their broader adoption and effective utilization. Here we proposed an innovative remote sensing and AI based framework to derive the lifespan of 3D flood structure incorporating flood extent and depth. Taking the once‐in‐a‐century flood in China's Hai River Basin (HRB) in 2023 as example, the thrived two‐month‐long catchment‐scale 229,000‐km 2 3D flood structure enabled the first quantitative evaluation to the flood control and drainage capacities of FDAs. The results revealed that eight activated FDAs detained 2.7 billion m 3 of floodwaters, with most of the water drained within two weeks. River conveyance capacity was identified as a pivotal factor in the flood drainage process. The 3D flood structure further highlights disparities in detention processes between upstream and downstream FDAs due to cascade effects, providing valuable guidance for their management. These findings resolve the key uncertainties and debates regarding FDAs, providing insights to enhance FDA management, and supporting their broader implementation in flood‐prone regions with extensive agricultural land use and dense populations, thereby mitigating future flood risks and minimizing associated losses.

Abstract. Place names are important basic geographic information resources for surveying and mapping. They play a significant role in maintaining national security, national defense construction, economic development, and other aspects. The construction and development of global geographic information resources require the continuous, efficient acquisition, and integration and updating of place names data. This paper takes the boundaries of the lowest administrative units in the task area as the matching scope and employs a string edit distance algorithm based on the substitution rules of common name synonyms for the similarity calculation and integration and updating of place names. It also uses a multilingual translation method based on a transliteration rule library for foreign language translation, with the addition of collaborative processing verification. This ultimately improves the matching efficiency of multi-source place names elements and ensures the accuracy of foreign place names translation. The research findings have been tested in some national regions and can provide a reference for the production and updating of place names data worldwide.

Transplantation therapy using induced pluripotent stem cell (iPS) cell-derived dopamine (DA) progenitors for Parkinson's disease (PD) has attracted attention as an innovative treatment to restore DA neurons in PD, which leads to the improvement of motor disturbance. iPS cells are multipotent stem cells with very high proliferation activity, created by reprogramming mature somatic cells through the transduction of four transcription factors. Relative to fetal midbrain DA neurons, iPS cells have advantages in terms of ethical aspects and availability. On the other hand, the most serious concern associated with therapies with ES/iPS cells is the risk of tumor growth that is caused by the proliferation of undifferentiated ES/iPS cells. Human ES cells that differentiate into DA neurons have been shown to form teratomas. Another concern is graft-induced dyskinesia (GID). GID, which is likely caused by several factors including contamination with serotonergic neurons, developed in the recipients of fetal ventral midbrain (VM) in randomized controlled trials. To enrich the DA progenitor cells and eliminate unwanted cells, a protocol for sorting midbrain DA neurons with antibodies against CORIN, a marker for floor plates, was developed. CORIN-sorted dopamine progenitors were transplanted into the bilateral putamina of MPTP-treated Parkinson models of cynomolgus monkeys, resulting in 18F-DOPA PET-positive graft formation and motor improvement without tumor formation two years after the surgeries. Very recently, a phase I/II trial of iPSC-derived, CORIN-sorted dopaminergic cells for Parkinson's disease was completed (The Kyoto Trial) (Takahashi, 2020; Sawamoto et al., 2025) [1,2]. Based on the results of the trial, we would like to discuss the current status and future perspectives of iPS cell therapy for PD.

Although current infrared (IR) and Visible (VIS) image dual-domain fusion (IVIDF) networks (e.g., SFDFusion) have explored the combination of spatial and frequency do mains, there are still three core limitations: 1) Spatial domain feature extraction relies on fixed convolutional kernels, which are difficult to adapt to irregular edges such as road cracks; 2) The independence of frequency domain amplitude and phase spectrum processing leads to texture structure mismatch; 3) The lack of a cross-domain interaction mechanism causes feature conflicts. To overcome these fundamental limitations, we propose DPFFusion: a dual-domain parallel feature fusion network that introduces three key innovations: First, we develop a Dual Modal Refinement Module (DMRM) that dynamically adapts receptive fields through deformable convolutions and integrates cross-modal attention to suppress feature redundancy. Second, we construct an Amplitude-Phase Synergy Fusion Module (APSFM) that explicitly models cross-modal correlations between AMPir and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PHA_vis</i> through channel attention. Finally, we introduce Cross-Domain Dynamic Attention (CDFM) using a hybrid at tention mechanism to resolve inter-domain feature conflicts. Evaluations on benchmark datasets (M3FD, MSRS, RoadScene) show DPFFusion outperforms SFDFusion by 28.9% in Mutual Information (MI) on M3FD, with real-time inference (28 ms/frame) and minimal parameters (0.14M). Ablation studies confirm deformable convolution boosts irregular edge response by 58%, while the amplitude-phase synergy reduces Frequency Energy Ratio (FER) by 15.2% and enhances texture fidelity by 39.2% in Qabf metric. The code is open source: https: //github.com/NUAA-RS/DPFFusion.

Effective management of corn straw and stubble is critical in conservation tillage, as it impacts soil health and productivity. However, accurate classification of different types of straw cover has been hindered by their similar spectral and spatial characteristics and the low spatial resolution of hyperspectral satellite imagery. Moreover, traditional convolution neural network (CNN)-based methods, which rely on first-order statistics for feature extraction, often struggle to extract distinguishable features of highly similar objects effectively, thereby reducing classification accuracy. In this study, a second-order statistical-feature extraction algorithm based on CNN that uses fused multispectral and hyperspectral data was tested for its ability to classify types of straw cover. In the first step, coupled non-negative matrix factorization (CNMF) was used to fuse hyperspectral and multispectral images effectively, thereby enhancing the spatial resolution of the hyperspectral data. In this study, we integrated pointwise convolution (PWC), depthwise convolution (DWC), and a distance covariance matrix (DCM) to form a mixed convolution and DCM (MCDCM) network; we used this to extract and integrate deep spectral–spatial features of the hyperspectral images. Our experimental results show that the MCDCM network significantly improved classification accuracy compared to traditional methods, with accuracy rates for the different straw-cover types exceeding 90% and overall accuracy reaching 98.26%. The fused image also exhibited better preservation of feature edges and contours. The accurate identification of corn-straw-cover types achieved with the proposed MCDCM method is a major step in optimizing conservation-farming practices, improving soil fertility and farm productivity, and supporting sustainable ecological development.