Issue Ebookcover storyIssue CoverIssue Contents
Latest Issue
Online FirstList of IssuesDocument Types
2026
Issue 10
Issue 9
Issue 8
Issue 7
Issue 6
Issue 5
Issue 4
Issue 3
Issue 2
Issue 1
2025
2024
2023
2022
2021
2020
2019
2018
2017
Vegetation effects on land-surface energy partitioning： Research progress， remote-sensing applications， and prospects AI Introduction
Abstract：Accelerating global warming and intensifying human activities have substantially altered vegetation cover worldwide, with profound effects on regional and global climates. A key mechanism is the influence of vegetation change on land-surface energy partitioning, namely the partitioning of all-wave net radiation to latent, sensible, and soil heat fluxes. This partitioning directly regulates near-surface temperature, atmospheric stability, and boundary-layer dynamics. Clarifying how vegetation affects energy partitioning is therefore essential for understanding the climatic effects of land-cover change and improving climate models and land-management practices.However, isolating and quantifying vegetation-induced responses from background climate variability remain challenging. Nonlinear interactions, pronounced spatiotemporal heterogeneity, and limited long-term observations complicate the separation of vegetation signals from climatic noise. Moreover, the magnitude and even the direction of vegetation-induced shifts in energy partitioning vary with latitude, season, aridity, and ecosystem type, precluding simple generalization.This review synthesizes three decades of research on vegetation-energy partitioning, with an emphasis on remote-sensing applications. Vegetation modifies land-surface energy partitioning through three interconnected pathways: (1) physiological traits—including leaf area, stomatal conductance, and root depth—that regulate surface–atmosphere exchanges of water and heat; (2) changes in albedo and emissivity that alter the surface radiation budget; and (3) hydrological processes, including interception, transpiration, and soil-moisture extraction, that modify latent and sensible heat fluxes. These processes are influenced by human activities, such as deforestation, irrigation, ecological restoration, and urbanization, as well as by natural disturbances, including wildfires and climatic extremes. Interactions among energy fluxes may reinforce or offset one another. For example, greater latent heat flux generally suppresses sensible heat flux and cools the surface during the growing season, whereas reduced evapotranspiration can intensify heatwaves in dry regions. This complexity produces distinct climatic responses across regions and scales and therefore requires integrated assessment.Numerical models, ranging from land-surface schemes to Earth system models, remain the dominant methodological tools. However, observational evidence is becoming increasingly important because model parameterizations and forcings remain uncertain. Satellite remote sensing helps overcome the spatial sparsity of in situ networks by providing broad coverage, long time series, and observations at multiple resolutions. It supplies key variables—including leaf area index, albedo, land-surface temperature, and evapotranspiration—as model inputs and constraints. Remote sensing has thus extended vegetation-energy research to the global scale, enabling attribution of the effects of vegetation greening on the surface energy balance and providing quantitative evidence to inform policy.Future research should prioritize four directions. First, improvements in the spatiotemporal resolution and accuracy of remote-sensing products, including the use of remote-sensing foundation models that integrate deep learning with physical constraints, should enable better representation of fine-scale heterogeneity and rapid land-cover change. Second, greater attention should be given to vegetation transition zones and complex terrain, including arid and semi-arid regions, urban areas, and mountains, where energy-partitioning responses are particularly sensitive but remain poorly understood. Third, integrating coupled biogeochemical processes—especially links among carbon uptake, nutrient cycling, and energy exchange—will support more comprehensive assessments of vegetation-climate interactions. Fourth, multi-platform and multi-sensor data fusion, combined with robust causal-attribution methods, is essential for reducing uncertainty and informing actionable strategies. Collectively, these advances will deepen mechanistic understanding of how vegetation change affects regional climate through energy partitioning and will support sustainable ecosystem management and climate adaptation.Keywords：vegetation;surface energy partitioning;surface radiation budget;remote sensing;latent heat flux;sensible heat flux;land-atmosphere interactions;climate response343|15844|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Research progress in urban thermal environment studies toward public health assessment AI Introduction
Abstract：Against the backdrop of global climate change and rapid urbanization, the widespread replacement of natural land cover with impervious urban surfaces has substantially altered surface energy exchanges and intensified urban thermal stress. Urban thermal environments have consequently become a major challenge to sustainable urban development, with far-reaching implications for thermal comfort, heat exposure, and public health. Quantifying the spatiotemporal dynamics of urban thermal environments, identifying their health consequences, and developing effective mitigation and adaptation strategies are essential to advancing the United Nations Sustainable Development Goals, particularly those concerning sustainable cities and communities, good health and well-being, and reduced inequalities. In recent years, urban thermal environment research has undergone a notable shift from the physical monitoring of thermal conditions toward human-centered assessments of thermal perception, exposure inequality, adaptive capacity, and heat-related health risks. To clarify this evolving research paradigm, this review synthesizes major advances over the past two decades through bibliometric analysis and thematic review. First, it summarizes progress in monitoring urban thermal environments using in situ observations, satellite remote sensing, and numerical simulations. These approaches have improved the characterization of air temperature, land surface temperature, and urban microclimatic processes across multiple spatial and temporal scales. Nevertheless, current monitoring frameworks remain constrained by limited spatial coverage, discontinuous temporal records, inadequate all-weather capability, and uncertainty in cross-scale integration. Second, the review examines the growing emphasis on outdoor thermal comfort, urban heat exposure, and heat-related health risks. Research has increasingly shifted from the physical intensity of urban heat islands to how residents perceive, experience, and respond to thermal stress. However, fine-scale, dynamic assessments of heat exposure remain limited by the lack of globally consistent diurnal temperature data and short-term population mobility information, constraining the evaluation of exposure inequality across places, periods, and social groups. Finally, four future research directions are proposed. (1) Integrated retrieval and reconstruction of multiple temperature parameters should be strengthened by combining ground observations, multisource satellite remote sensing, microwave data, data fusion, deep learning, and physically constrained modeling to generate long-term, high-resolution, all-weather temperature datasets for urban heat studies. (2) Future research should further characterize the interactions of urban heat exposure with urban morphology, ecological structure, infrastructure provision, population dynamics, socioeconomic conditions, and patterns of social development, with particular attention to exposure inequality. (3) Public-health-oriented adaptation should be advanced through nature-based solutions, urban greening, blue–green infrastructure, vertical vegetation, landscape optimization, and climate-sensitive urban design, while fully considering exposure, vulnerability, and access to cooling and medical resources. (4) Under continued warming and increasingly frequent extreme-heat events, interdisciplinary studies should extend heat–health risk assessment across spatial and temporal scales, investigate the effects of persistent heat on vulnerable populations, and support medical-resource allocation, emergency-response planning, and urban heat-risk management.Keywords：Urban heat environment;remote sensing monitoring;urban heatwaves;urban heat island;heat exposure;health risks of high temperatures523|15867|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
Research Progress
Remote sensing of urban six-dimensional space environments： Current status and future challenges AI Introduction
Abstract：As China’s new urbanization process advances, accurate monitoring and assessment of the multidimensional urban environment have become important research frontiers in remote sensing. Using the theoretical framework of the “Urban Six-Dimensional Space,” this paper systematically reviews progress in the remote sensing of six dimensions (urban green space, heat space, gray space, humidity, brightness, and livability). It covers key aspects such as information extraction, pattern evolution, and impact assessment, as well as recent developments. The paper also examines common challenges in current research, including data fusion, model generalization, and cross-scale modeling. On this basis, it outlines future directions for urban environmental remote sensing, including intelligent interpretation, multisource integration, dynamic prediction, and integration of human perceptions. The aim is to provide theoretical guidance and practical pathways for developing a smarter, more refined, and more sustainable remote sensing system for urban environmental monitoring.Keywords：urban environmental remote sensing;urban green space;urban heat space;urban gray space;urban humidity;urban brightness;urban livability24|15779|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Weakly supervised semantic change detection in street view imagery and its application in urban renewal dynamics mapping AI Introduction
Abstract：Street view imagery (SVI) has emerged as an important source of geospatial big data for understanding the urban built environment. Accurate detection of facade-level changes and identification of their semantic categories are crucial for monitoring urban renewal dynamics. However, existing change detection approaches often struggle to separate the temporal ownership of changed objects (change decomposition) and provide semantic information, leading to complex workflows and high data preparation costs. This study aims to develop a weakly supervised semantic change detection framework that jointly integrates change decomposition and semantic labeling and applies it to the dynamic mapping of urban renewal in Guangzhou, China.This study proposes Cross-C2PO, a novel dual-branch architecture designed to achieve end-to-end weakly supervised change decomposition. Different from traditional single-branch models, Cross-C2PO introduces a cross-comparison mechanism to explicitly model asymmetric temporal differences, ensuring completeness and consistency regardless of input order. The proposed model integrates a differentiable union operator to maintain consistency constraints during weak supervision and employs the Cross-MTF feature fusion function to break commutativity, enabling accurate temporal differentiation. Based on the outputs of Cross-C2PO, a semantic change detection workflow that leverages state-of-the-art segmentation models (e.g., DeepLabV3+) is designed without relying on synthetic datasets. Finally, an urban renewal dynamic index is introduced to quantify facade-level changes and visualize renewal patterns across panoramic and directional views (front, back, left, right) in the central districts of Guangzhou from 2013 to 2019, using 11,431 pairs of Baidu street view panoramas.Traditional change detection methods often fail to achieve end-to-end change decomposition and typically rely on complex multi-stage pipelines, limiting their scalability and flexibility. In contrast, the proposed Cross-C2PO framework enables one-stage weakly supervised change decomposition and can be seamlessly integrated with mainstream architectures, providing improved detection accuracy while allowing temporal ownership splitting without the need for additional labels. Experiments on multiple benchmark datasets demonstrate that the proposed method consistently achieves state-of-the-art performance, outperforming existing approaches in binary change detection and decomposition tasks. Ablation studies further validate the effectiveness of the cross-branch structure, Cross-MTF fusion, and the differentiable union operator. When applied to Guangzhou street view imagery, the proposed workflow successfully produces urban renewal dynamic maps, revealing that high-intensity updates are clustered in Liwan and Baiyun industrial areas, whereas moderate changes are mainly observed in residential zones. Furthermore, directional view analysis highlights local disparities and micro-scale renewal patterns.The proposed Cross-C2PO framework offers a simple but effective solution for weakly supervised semantic change detection, allowing for accurate change decomposition without additional synthetic labels. Combined with an interpretable urban renewal dynamic index, the framework provides a scalable and cost-effective approach for urban facade change analysis. This study establishes a connection between street view imagery and AI-based computer vision for urban analytics, offering new insights into spatiotemporal urban renewal dynamics. Future studies will focus on optimizing computational efficiency and extending the proposed method to multi-source data integration for large-scale applications.Keywords：urban renewal;street view imagery;semantic change detection;scene change detection;weak supervision;dynamic index1046|1126|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
Abstract：Quantitative analysis of the coupling coordination relationship between the ecological environment and socioeconomic development in China’s poverty-alleviated counties, as well as the identification of influencing factors, can contribute to summarizing poverty alleviation experiences and consolidating the achievements of poverty eradication efforts.Based on long-term and high-resolution remote sensing datasets, this study constructed a comprehensive evaluation index system for the ecological environment and socioeconomic development across 832 poverty-alleviated counties in China. The development levels and coupling coordination status of the two dimensions in 2010, 2015, and 2020 were quantitatively assessed. Furthermore, a Boosted Regression Tree model was employed to identify the contribution rates of various indicators to the regional coupling coordination development.Results indicated that the ecological environment in poverty-alleviated counties generally exhibited a higher level of comprehensive development compared with socioeconomic factors. However, socioeconomic development progressed at a faster pace than ecological improvements. Moreover, the average annual growth rates of both dimensions during 2015-2020 were higher than those during 2010-2015. Spatially, the coupling coordination degree was highest in northeastern counties and lowest in northwestern counties, revealing a distribution pattern of “high in the east and low in the west” and a progressive improvement trend from coastal to inland areas. Most counties were categorized as “economically lagging.” Among the influencing indicators, population size, gross domestic product, and nighttime light intensity exhibit particularly substantial contributions to the coupling coordination degree.Based on the poverty alleviation paths and practical experiences of typical regions, this study concludes that implementing context-specific industrial poverty alleviation strategies is crucial for accelerating socioeconomic development in poverty-alleviated counties. The establishment of diversified, locally distinctive industries is identified as a key approach for consolidating poverty alleviation achievements and preventing the recurrence of poverty.Keywords：poverty-alleviated counties;ecological environment;socioeconomic development;Remote sensing dataset;coupling coordination;boosted regression tree model;contribution rate;spatial distribution pattern;industrial poverty alleviation818|525|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Estimation and spatiotemporal evolution analysis of annual 500 m anthropogenic heat flux in the Guanzhong Plain urban agglomeration AI Introduction
Abstract：Anthropogenic Heat Flux (AHF) refers to the total amount of human-generated heat emissions per unit area within a unit of time. As a key contributor to the formation of the urban heat island effect, anthropogenic heat emissions directly influence urban thermal environments by releasing waste heat into the atmosphere. Therefore, studying the spatiotemporal distribution of AHF is essential for understanding the formation and evolution of urban thermal environments and provides important theoretical and practical implications for mitigating and regulating urban ecological issues.Aiming to characterize the spatial distribution of anthropogenic heat emissions under limited sample data at a regional scale and analyze the spatiotemporal evolution patterns of different AHF types in the Guanzhong Plain urban agglomeration in China, this study first employed a modified emission inventory method to estimate AHF values for prefecture-level cities. The modification primarily addresses the overestimation of residential building heat emissions through the exclusion of private vehicle energy consumption, which is already accounted for in residential energy use. Additionally, the calculation of transportation AHF was refined by incorporating heat emissions from public transportation. Subsequently, a multivariate linear regression model was constructed using multi-source spatial datasets, including point of interest (POI) data, nighttime light data, building height data from the global human settlement layer, and population distribution data from WorldPop, to estimate different types of anthropogenic heat emissions. This approach enabled the acquisition of annual anthropogenic heat emission datasets from 2016 to 2021 for the Guanzhong Plain urban agglomeration at a 500-m spatial resolution, followed by a spatiotemporal analysis of their emission characteristics.The following results are obtained(1) Multivariate linear regression demonstrates high feasibility for AHF gridding, with all fitted models achieving high accuracy and R2 values exceeding 0.9. Among the models, the building AHF model achieves the highest accuracy, with an R2 of 0.98. (2) POI data plays a crucial role in the gridded allocation of anthropogenic heat emissions by effectively reflecting the spatial heterogeneity of different types of anthropogenic heat emissions. Therefore, POI data is an important source for estimating the spatial distribution of AHF from various heat sources. (3) The spatial distribution of AHF in the Guanzhong Plain urban agglomeration is uneven, with high-value areas mainly concentrated in economically developed, flat, and highly urbanized central regions, particularly the northern central urban area of Xi’an. Temporally, AHF exhibits an overall upward trend.The spatiotemporal distribution data of anthropogenic heat emissions generated in this study can provide valuable data references for urban thermal environment assessments and thermal comfort simulations. Furthermore, this information can support the development of targeted heat mitigation strategies, contributing to more sustainable urban planning and improved urban living environments.Keywords：Anthropogenic heat;point of interest (POI) data;nighttime light data;multi-source spatial data;human activities;multiple linear regression;inventory method;Guanzhong plain urban agglomeration;spatiotemporal change1037|1329|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Long-term water-color evolution mechanisms in large shallow lakes under anthropogenic disturbance： A case study of Lake Chaohu AI Introduction
Abstract：Large shallow lakes are critical to regional water supply and ecological security. Understanding their long-term water-quality dynamics is therefore of substantial scientific and practical importance for watershed management and ecological protection. However, these lakes are ecologically vulnerable and highly sensitive to external disturbances. The extent to which their water quality exhibits phased and nonlinear responses to intensive human activities and sustained environmental management remains poorly understood. In this study, Lake Chaohu, the fifth-largest freshwater lake in China, was selected as a representative large shallow lake. The Forel-Ule index (FUI), which provides an integrated measure of water optical properties and water-quality status, was used to characterize long-term changes in water quality. Using long-term MODIS remote sensing data from 2000 to 2024, an improved, high-accuracy water-color inversion model was applied to generate a long-term water-color series. Validation yielded an overall Mean Relative Error (MRE) of 1.38% and a Root Mean Square Error (RMSE) of 3.32°. The spatiotemporal patterns of water color and their drivers under long-term anthropogenic disturbance were then systematically examined. The results showed that the multiyear mean FUI of Lake Chaohu ranged from 12 to 15, with a mean hue angle (α) of 211.67°±2.58°, indicating an overall green-to-yellow water color. Spatially, a clear decreasing gradient from west to east was observed, with mean α values of 213.13°±2.22°, 211.67°±2.58°, and 209.79°±2.16° in the western, central, and eastern lake regions, respectively. This spatial pattern was broadly consistent with the distribution of the major inflowing rivers and differences in their pollution loads. At the seasonal scale, water color exhibited a distinct bimodal pattern. The highest α value occurred in autumn (214.23°±3.12°), indicating the most turbid conditions, whereas the lowest occurred in summer (208.74°±3.03°), indicating the clearest conditions. Seasonal variations in natural factors, particularly wind speed, precipitation, the normalized difference vegetation index (NDVI), and temperature, regulated these intra-annual changes. Over the long term, the annual mean α increased at a rate of 0.15° yr-1 from 2000 to 2024, with the largest increase occurring in the eastern lake region (0.26°/a). The Mann-Kendall and Pettitt tests both identified approximately 2014 as a key transition point. The decreasing trend during 2000—2014 (-0.13°/a) shifted to a significant increasing trend during 2014—2024 (0.44°/a), indicating a clear transition from earlier improvement to subsequent deterioration in the water environment. SHAP-based attribution further showed that anthropogenic factors, including gross domestic product (GDP), impervious surface area, nighttime light, and population, contributed substantially more to long-term water-color changes than natural climatic factors and were the dominant drivers. Early watershed pollution-control and ecological-restoration measures effectively reduced environmental pressure and improved water color. However, as urbanization and socioeconomic activities continued to intensify, these management gains were gradually eroded. This resulted in an imbalance in the lake’s pressure-response relationship and renewed deterioration in water color. Meanwhile, the Changjiang-to-Huaihe River Diversion Project and extreme climate events further intensified hydrodynamic disturbances and external pollutant inputs, increasing interannual variability and the risk of short-term water-color deterioration. Overall, the results revealed distinct phased, nonlinear, and spatially heterogeneous patterns in the long-term evolution of water color in large shallow lakes under sustained human disturbance. These findings provide a scientific basis for long-term water-environment management and adaptive regulation in Lake Chaohu and other large shallow lakes.Keywords：shallow lake;Chaohu Lake;water color index;long-term remote sensing;anthropogenic disturbance320|104|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
Socioeconomic Remote Sensing
Remote sensing analysis of spatiotemporal changes in global cropland water-use efficiency toward the sustainable development goals AI Introduction
Abstract：Cropland Water-Use Efficiency (WUE) reflects the coupling between cropland productivity and water consumption and is an important indicator for evaluating agricultural water use and coordinated progress toward the Sustainable Development Goals (SDGs). This study aims to characterize the spatial patterns and temporal changes in global cropland WUE from 2001 to 2024 and to quantify the relative contributions of net primary productivity (NPP) and evapotranspiration (ET) to changes in WUE, and explore their implications for progress toward the SDGs.Cropland WUE was defined as the ratio of NPP to ET. Using a global remote-sensing-derived cropland WUE dataset for 2001-2024, the spatial differentiation, temporal trends, and change patterns of cropland WUE were systematically analyzed. Irrigated and rainfed croplands were also compared. In addition, the relative contributions of NPP and ET to changes in WUE were quantitatively decomposed using a contribution decomposition method based on the total differential. Combined trend types of WUE, NPP, and ET were further analyzed to examine concurrent changes in productivity and water consumption and their implications for coordinated progress toward the SDGs.(1) The mean global cropland WUE during 2001-2024 was approximately 1.08 g(C)/kg(H2O), with pronounced spatial heterogeneity. Among climate zones, WUE was highest in arid regions and lowest in temperate regions; among continents, it was highest in Europe and lowest in Africa. (2) Global cropland WUE increased significantly at a rate of approximately 0.003 g(C)/(kg(H2O)·a), corresponding to an overall increase of about 6.4%. The increase in WUE was faster in irrigated than in rainfed croplands. Approximately 27% of global cropland areas showed significant increases in WUE, mainly in the major agricultural regions of India and China, whereas only about 7% showed significant decreases, mainly in east-central Europe, southern Russia, Central Asia, and southern South America. Among major grain-producing countries, China and India showed the fastest increases; Brazil and Canada exhibited steady improvements; the United States remained broadly stable; and Ukraine, Russia, and Argentina showed declining trends. (3) NPP was the dominant contributor to global cropland WUE trends. Within cropland areas exhibiting significant WUE trends, the relative contribution of NPP exceeded that of ET across approximately 96.5% of the area. (4) Within these areas, approximately 73.6% exhibited concurrent increases in WUE, NPP, and ET, whereas only 5.5% exhibited increasing WUE and NPP alongside decreasing ET. A further 19.8% exhibited concurrent declines in NPP and WUE.Global cropland WUE generally increased from 2001 to 2024, but this increase was driven predominantly by increases in cropland productivity rather than reductions in water consumption. To advance SDG 2 (Zero Hunger) and SDG 6 (Clean Water and Sanitation) in a coordinated manner, the sustainability of WUE improvements should be evaluated in the context of concurrent changes in cropland productivity and water consumption. Region-specific management should also account for local water-resource constraints. These findings provide a scientific basis for assessing agricultural sustainability and developing regionally adapted water-resource management strategies.Keywords：cropland water-use efficiency;net primary productivity;evapotranspiration;spatio-temporal change;sustainable development goals344|60|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Winter wheat yield prediction using the CBA-WOFOST model with remote-sensing-based biomass assimilation AI Introduction
Abstract：With growing global food-security challenges and the increasingly adverse impacts of extreme environmental changes on agricultural production, accurate regional-scale crop-yield prediction has become a key issue in agricultural remote sensing and crop modeling. In this context, the assimilation of remote sensing data into crop models has been widely applied. By integrating remote sensing observations with crop-growth processes, this approach improves the dynamic characterization of crop conditions. However, existing studies have mainly focused on Leaf Area Index (LAI) or Soil Moisture (SM) as assimilation variables, whereas relatively limited attention has been given to the feasibility and effectiveness of Above-Ground Biomass (AGB) assimilation.In this study, the Dung Beetle Optimizer (DBO) and non-dominated sorting genetic algorithm II (NSGA-II) were used for model calibration. In combination with the ensemble Kalman filter (EnKF), the above-ground biomass estimated by the crop biomass algorithm for wheat (CBA-Wheat) model was assimilated into the WOrld FOod STudies (WOFOST) model, thereby establishing a CBA-WOFOST above-ground biomass assimilation framework integrating remote sensing estimation with crop growth modeling. The framework enabled dynamic biomass updating and simulation of component-wise dry-matter partitioning, and its performance in simulating biomass and yield was evaluated at the regional scale.The results indicated that the CBA-Wheat model based on relative growing degree-days (RGS) and Enhanced Vegetation Index 2 (EVI2) achieved superior performance in biomass estimation. The coefficient of determination (R2) and Root Mean Square Error (RMSE) for the validation set were 0.84 and 2303 kg/hm2, respectively. At the regional scale, the biomass-assimilated WOFOST model achieved satisfactory accuracy, with coefficients of determination (R2) of 0.86 and 0.60 for biomass and yield simulations, respectively, and corresponding RMSE values of 2411 kg/hm2 and 1261 kg/hm2.The CBA-WOFOST framework developed in this study achieved relatively high simulation accuracy and provides a reliable approach for assimilating multisource remote sensing observations into crop models, thereby improving the accuracy of winter wheat growth monitoring and yield assessment.Keywords：WOFOST;CBA-Wheat;EnKF;Sentinel-2;winter wheat;AGB;EVI2;DBO17|23|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27InSAR coherence time-series analysis and burned-area extraction under forest fire disturbance AI Introduction
Abstract：Existing remote sensing parameters for forest-fire monitoring, such as thermal anomalies, the normalized burn ratio, and the radar burn ratio, primarily capture changes in surface radiative characteristics. However, the mechanisms underlying structural changes in forested areas before and after burning remain insufficiently understood. Given the high sensitivity of interferometric synthetic aperture radar (InSAR) coherence to changes in surface materials and structural properties, this study proposes a method for identifying forest damage and extracting burned areas through time-series analysis of InSAR coherence.First, a sequence of InSAR coherence maps for adjacent SAR acquisition dates is constructed from time-series SAR imagery. Histogram statistics of coherence values within the study area are then used to identify the time window associated with abrupt regional coherence changes caused by forest damage. Subsequently, normalized differences between pre- and post-damage coherence are calculated to construct a time series of InSAR coherence indices, and an iterative self-organizing clustering algorithm is applied for unsupervised burned-area extraction. To evaluate the feasibility of the proposed method, two forest fires that occurred in Xichang and Muli in 2020 were selected as test cases. Burned areas derived from Sentinel-2 optical imagery were used as cross-validation data to evaluate the accuracy of InSAR-based burned-area extraction.The experimental results indicate that forest damage can be identified with relatively high accuracy through time-series statistical analysis of InSAR coherence. The extracted burned areas agree closely with those derived from Sentinel-2 imagery. The extraction accuracies for the Xichang and Muli fire events were 82.71% and 66.14%, respectively, with commission errors below 3.6% and omission errors below 6.5% in both cases.Time-series analysis of InSAR coherence can effectively identify forest damage and extract burned areas. By focusing on the decoherence caused by forest destruction before and after fires, the proposed method provides a useful approach to remote-sensing-based forest-damage identification and burned-area extraction under persistent cloud cover or smoke-obscured conditions.Keywords：forest fire;forest damage identification;burned area extraction;interferometric synthetic aperture radar;coherence coefficient;time-series analysis;unsupervised classification127|59|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
Forestry and Agriculture
Spatio-temporal prediction and cause mechanism analysis of InSAR long-term land subsidence based on bidirectional PhConvLSTM AI Introduction
Abstract：Local land subsidence can occur during airport construction and operation. Its long-term accumulation may pose safety hazards to aircraft operations and increase airport maintenance costs; therefore, high-precision prediction of the spatiotemporal distribution of land subsidence is of substantial practical value. Conventional spatiotemporal prediction models based on convolutional long short-term memory (ConvLSTM) have several limitations, including a limited capacity to capture long-term temporal features, inadequate use of temporal information, and incomplete representation of local extreme-value features. To address these limitations, this study proposes a bidirectional peephole-connected ConvLSTM (PhConvLSTM) that incorporates peephole connections and considers both forward and backward time sequences for spatiotemporal land-subsidence prediction.Method First, persistent scatterer interferometric synthetic aperture radar (PS-InSAR) was used to obtain time-series land-subsidence observations. The time-series results were then interpolated and smoothed to generate complete spatiotemporal images. Second, 3D convolution, max pooling, and min pooling were combined to extract key information from these images, and a bidirectional PhConvLSTM model was constructed for spatiotemporal prediction. A temporal attention mechanism was introduced to increase the model’s focus on critical time steps, and uncertainty in the land-subsidence predictions was estimated. Furthermore, the effects of climate change, foundation treatment, and transportation infrastructure construction on airport land subsidence were analyzed. The causes of subsidence at Beijing Daxing International Airport and the triggering mechanisms of regional land subsidence were examined to provide a reference for geological-disaster early warning.Result Beijing Daxing International Airport was selected as the study area. The bidirectional PhConvLSTM model predicted the spatiotemporal distribution of airport land subsidence, with the maximum cumulative subsidence reaching 135.7 mm by February 2025. On the test set, the maximum prediction error was less than 16 mm, and the maximum goodness of fit R2 reached 0.99. Compared with the ConvLSTM model, the average MSE and MAE were reduced by 61.31% and 45.87%, respectively.Conclusion The experimental results demonstrate that the proposed bidirectional PhConvLSTM model enables high-precision land-subsidence prediction and has strong practical applicability. Uneven land subsidence at Beijing Daxing International Airport is jointly driven by meteorological variability, differences in foundation treatment, and long-term dynamic traffic loads.Keywords：land subsidence;PS-InSAR;spatio-temporal prediction;ConvLSTM;cause mechanism analysis22|35|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Investigating the seismic damage characteristics of the MS 6.8 Dingri， Xizang earthquake using multi-source remote sensing data AI Introduction
Abstract：On January 7, 2025, an Ms 6.8 earthquake struck Dingri, Xizang, causing heavy casualties. The affected area has an average altitude of over 4000 meters, posing substantial challenges for field investigations and seismic research. This study aims to comprehensively characterize the earthquake’s seismogenic mechanism, building-damage characteristics, and spatial distribution of secondary hazards. The potential of domestically developed high-resolution satellite data for earthquake emergency monitoring is also evaluated.Using multitemporal and multisource domestic remote sensing data, this study applied deep-learning methods to rapidly identify and assess damaged buildings. Coseismic deformation fields were derived from land-exploration satellite data to analyze seismogenic structures and source mechanisms. Gaofen-series satellite images were used to map the surface rupture zone and earthquake-induced secondary geological hazards. Furthermore, deformation-field data were integrated to analyze the spatial distribution of secondary landslide hazard chains.Earthquake-induced secondary geological hazards were concentrated mainly in seismic-intensity zones VI and VII, and their spatial density increased with seismic intensity. The earthquake resulted from normal faulting; maximum coseismic surface deformation exceeded 2 meters, and maximum fault displacement reached 6 meters. A discontinuous, nearly north-south-trending surface rupture zone extended along the eastern bank of Dengmocuo fault, with an interpreted length of approximately 10 km and a maximum width of 300 m. Changsuo Township sustained the most severe building damage, with a damage rate exceeding 75%. Damaged buildings were concentrated along the northwestern hanging wall of the Demacuo fault.This study systematically characterizes the mechanisms and spatial distribution of damage caused by the Dingri earthquake. The results provide baseline data for in-depth analyses of its seismogenic mechanism and disaster characteristics, as well as for subsequent risk assessment. They also demonstrate the utility of high-resolution data from multiple domestic satellites for rapid emergency monitoring of earthquake disasters.Keywords：Earthquake;earthquake emergency response;multi-source remote sensing;identification of earthquake damage information;surface rupture;InSAR;Dingri earthquake14|22|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
GeologyandDisasters
Seasonal dynamics-aware quantification of traffic sign visibility and simulation of street tree pruning AI Introduction
Abstract：Occlusion of traffic signs by street trees threatens road safety. Existing studies, however, have largely focused on static occlusion detection and lack an integrated technical pathway linking occlusion quantification to pruning decisions.This study developed a decision framework linking dynamic quantification of seasonal occlusion to the simulation and evaluation of different pruning intensities. An 8.0 km road section in Zhangjiang Hi-Tech Park, Shanghai, was selected as the study area, encompassing 74 cantilever-mounted traffic signs and 1524 street trees. First, high-precision street-tree models were reconstructed from point clouds (mean error, 2.0 cm), and virtual foliage was added to simulate seasonal variation. Second, view-frustum spatial analysis was used to quantify the pavement-level visibility of traffic signs in winter and summer. Finally, pruning was simulated using two branch-diameter thresholds, 5 cm and 10 cm, and the outcomes were evaluated in terms of pruning cost, ecological loss, and landscape impact.Mean pavement-level traffic-sign visibility was 91.15% in winter and 85.26% in summer, a decrease of 5.89 percentage points. Sign No. 7 showed the greatest seasonal decrease (41.45%), indicating the most pronounced seasonal effect. Traffic signs with shorter cantilevers were more susceptible to occlusion. Under the 5 cm pruning scheme, branch-volume loss was 6.20% and leaf loss was 20.70%. Under the 10 cm scheme, these losses increased to 19.20% and 39.21%, respectively, whereas the number of pruning points decreased from 558 to 219. Although the higher-intensity scheme reduced the number of pruning points and associated costs, it caused substantial structural damage to the trees. For example, a tree near sign No. 43 lost 57.69% of its branch volume, and some trees were reduced almost to bare trunks.The proposed framework maps quantitative occlusion estimates to pruning-threshold selection criteria and reveals an asymmetric trade-off between fewer pruning points and greater structural damage to trees. It therefore provides a scientific basis for coordinated street-tree management that balances traffic safety, ecological conservation, and landscape quality.Keywords：mobile laser scanning point cloud;street tree;traffic sign;occlusion analysis;pruning simulation;traffic safety227|74|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27A systematic illumination and color normalization method for airborne multi-array imagery accounting for radiative transfer effects AI Introduction
Abstract：Airborne multi-array stitching cameras are widely used in remote sensing and mapping because of their high resolution, wide swath coverage, and efficient data-acquisition capacity. However, variations in hardware and installation angles introduce systematic differences in radiative-transfer processes during imaging, adversely affecting the radiometric and color consistency of stitched images.This study proposes a systematic illumination and color-normalization method for airborne multi-array imagery that accounts for radiative-transfer effects and specifically addresses the structural and optical characteristics of multi-array camera systems. The method decomposes the influence of the radiative imaging process into three components: lens-induced nonuniform optical attenuation, differences in atmospheric transmission paths, and inconsistent reflectance responses of surface features in overlapping areas. Corresponding corrections—irradiance adjustment, dehazing enhancement, and brightness and color matching—are applied sequentially to achieve consistent illumination and color despite differences in imaging paths.Aerial experiments using the ALC-2000 dual-array camera were conducted in Inner Mongolia. Statistical analysis of image brightness showed that the proposed method effectively suppressed overbright regions in the stitched virtual images, reduced overall luminance discrepancies, and eliminated color mismatches.The proposed method provides a robust solution for radiometric harmonization in multi-array aerial imaging.Keywords：multi-array images;illumination and color correction;irradiance inconsistency;dark channel dehazing;radiation imaging model13|22|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Photovoltaic hot spot detection method with infrared-visible complementary feature enhancement and adaptive fusion AI Introduction
Abstract：Photovoltaic (PV) systems are being increasingly deployed in complex environments, where PV panels are susceptible to various defects caused by environmental stress and internal degradation. Among these defects, hot spots caused by localized overheating can substantially reduce power generation efficiency and may even lead to thermal damage or fire hazards under severe conditions. Therefore, accurate and efficient detection of PV hot spots is crucial for ensuring the safe operation and maintenance of PV systems. However, traditional manual inspection methods are inefficient, costly, and difficult to adapt to large-scale PV plants. Although unmanned aerial vehicle (UAV)-based detection methods provide efficient and flexible solutions, most existing approaches mainly rely on thermal infrared images, thereby limiting detection performance due to low spatial resolution, insufficient texture information, and susceptibility to complex environmental interference.Aiming to address the limitations of single-modal UAV inspection, this study introduces a visible–thermal infrared UAV image fusion algorithm for PV hot spot defect detection. First, a dual-branch feature extraction network based on the YOLOv5 backbone is developed to independently extract complementary features from visible and infrared images. Aiming to effectively exploit the complementary information across different modalities, a multi-scale feature fusion strategy is designed. Specifically, a complementary feature enhancement module is introduced to strengthen cross-modal feature representations, while an adaptive feature fusion module dynamically adjusts the fusion weights of visible and infrared features to suppress redundant information and retain discriminative features. Furthermore, a Mamba-based feature pyramid network is developed by incorporating dual-branch ConvMamba modules, thereby improving global context modeling capability and enhancing the representation of multi-scale hot spot features. Finally, the fused features are transmitted to the detection head for accurate localization and classification of PV hot spot defects.Experiments conducted on a self-constructed UAV-based visible–thermal infrared PV hot spot dataset demonstrate that the proposed method achieves superior detection performance in mountainous centralized PV plants and urban rooftop PV scenarios. Compared with the single-modal YOLOv5 baseline, the proposed algorithm improves mAP@0.5 by 7.4% and 14.6% in mountainous and rooftop environments, respectively. Moreover, compared with existing multi-modal detection methods, the proposed approach achieves superior performance in terms of mAP@0.5, mAP@0.5:0.95, parameter quantity, computational complexity, and inference efficiency, demonstrating enhanced detection accuracy and robustness under complex conditions.The proposed visible–thermal infrared UAV image fusion algorithm effectively combines thermal anomaly information from infrared images with structural and textural details from visible images, thereby addressing the limitations of single-modal detection methods. Experimental results reveal that multi-modal fusion substantially improves the accuracy and reliability of PV hot spot defect detection in complex environments. This study offers an effective technical approach for UAV-based intelligent PV system inspection and provides theoretical support for the development of automated PV operation and maintenance technologies.Keywords：object detection;image fusion;unmanned aerial vehicle;infrared imagery;visible-light imagery;hotspot detection15|23|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27Sea ice and water classification for SWOT wide-swath observations using combined unsupervised and supervised machine learning AI Introduction
Abstract：Accurate ice-water classification is essential for altimetry-based estimation of sea ice freeboard and, subsequently, sea ice thickness. The Surface Water and Ocean Topography (SWOT) mission provides the first wide-swath altimetry observations over polar oceans covered by sea ice, creating new opportunities for high-resolution two-dimensional mapping of sea ice freeboard and thickness. However, SWOT differs substantially from conventional nadir altimeters, and the backscatter characteristics of sea ice and leads in SWOT observations remain insufficiently characterized. Moreover, few studies have investigated ice-water classification using SWOT data, limiting the application of SWOT observations to sea ice thickness retrieval. Therefore, using Sentinel-2 imagery as reference data, this study develops a classification method that combines unsupervised and supervised machine learning to distinguish among open leads, frozen leads, and sea ice.Sentinel-2 classification results derived using reflectance thresholds were first interpolated onto the SWOT grid to generate initial class labels. To address label inaccuracies arising from intersensor differences and interpolation errors, a Gaussian mixture model was applied to refine the initial labels based on the intrinsic clustering patterns of SWOT observations. Given the limited separability among classes in the original SWOT feature space, multiple derived features were constructed, including local elevation, texture, and multiscale regional statistical features. ERA5 wind speed was also incorporated to account for its influence on SWOT backscatter. Extreme gradient boosting (XGBoost) was then used for supervised classification to generate the final ice-water classification results. Finally, the leads identified by the proposed method were used to construct a local sea surface reference and estimate SWOT sea ice freeboard.Evaluation against an independent, manually labeled validation dataset yielded an overall classification accuracy of 95%. Among the Random Forest (RF), Artificial Neural Network (ANN), and Extreme Gradient Boosting (XGBoost) models, XGBoost achieved the best performance in lead identification. Refining the initial labels using an unsupervised Gaussian Mixture Model (GMM) improved validation accuracy by approximately 5 percentage points. Feature-combination analysis showed that the best performance was achieved by augmenting the original feature set with local elevation, wind speed, texture, and multiscale regional statistical features. Compared with the conventional minimum-point method, the proposed method yielded SWOT freeboard estimates that were more strongly correlated with ICESat-2 ATL10 freeboard, and the standard deviation of the SWOT-ICESat-2 freeboard differences was approximately 2.0 cm lower.The proposed SWOT wide-swath ice-water classification method generates and refines training labels by integrating optical imagery with unsupervised learning, thereby reducing the burden of manual sample annotation. Incorporating wind speed and derived features characterizing local elevation, texture, and multiscale regional statistics enhances the separability of different surface types and improves classification performance. This method provides a practical framework for ice-water classification and sea ice freeboard retrieval from SWOT observations in polar oceans.Keywords：SWOT satellite;sea ice and water classification;sea ice leads;sea ice freeboard;machine learning22|26|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27A multiscale segment anything model with fine-tuning and high-frequency enhancement for semantic segmentation of remote sensing images AI Introduction
Abstract：As a representative vision foundation model, the Segment Anything Model (SAM) has demonstrated strong performance and generalization across various downstream computer vision tasks. However, directly applying SAM to the semantic segmentation of remote sensing images remains challenging because of the substantial domain gap between natural images and remote sensing scenes. SAM is trained primarily on large-scale natural-image datasets and therefore cannot readily adapt to the complex characteristics of remote sensing images, including pronounced scale variations, diverse spatial distributions, and heterogeneous land-cover patterns. Moreover, SAM’s image encoder is based on the vision Transformer (ViT) architecture, which uses self-attention primarily to represent global context but has limited ability to capture local textures and high-frequency spatial details. Consequently, important boundary information and fine-grained structural features of remote sensing objects may be attenuated during feature extraction, resulting in blurred object boundaries and reduced segmentation accuracy.To address these limitations, this study proposes a multiscale SAM with fine-tuning and high-frequency enhancement (MSHF-SAM) for the semantic segmentation of high-resolution remote sensing images. The framework improves SAM’s adaptability to remote sensing scenarios by enhancing high-frequency feature representation, optimizing model transfer strategies, and employing a more effective segmentation decoder. Specifically, a high-frequency enhancement module is incorporated into the original SAM image encoder to strengthen the extraction and preservation of high-frequency spatial information, thereby improving the model’s representation of fine-grained textures and object boundaries. A parameter-efficient fine-tuning strategy is also adopted to mitigate the mismatch between the feature distributions learned by the pretrained SAM model and those of remote sensing images, enabling effective domain adaptation with only a small number of trainable parameters.In the decoder stage, the original SAM mask decoder is replaced with a MACU-Net-based decoder to achieve more accurate multiscale feature reconstruction. In addition, a lightweight boundary-refinement branch is developed to enhance the representation of object contours and improve segmentation detail while maintaining computational efficiency. Through joint optimization of the encoder and decoder, the proposed method effectively integrates global semantic information with local spatial details, yielding more precise segmentation results for complex remote sensing scenes.Extensive experiments were conducted on two widely used high-resolution remote sensing semantic segmentation datasets, LoveDA and ISPRS Vaihingen. The results demonstrate that MSHF-SAM outperforms several state-of-the-art segmentation methods in terms of mean intersection over union (mIoU), F1-score, and other evaluation metrics. Compared with the classical UNetFormer model, the proposed method improves mIoU by 4.33% on the LoveDA dataset and 2.85% on the Vaihingen dataset. These results confirm the effectiveness of the proposed high-frequency enhancement strategy, parameter-efficient fine-tuning scheme, and multiscale decoding architecture. The framework provides a promising approach for adapting large-scale vision foundation models to semantic segmentation of remote sensing images. It also demonstrates the potential of foundation models for intelligent remote sensing image interpretation.Keywords：deep learning;segment anything model (SAM);semantic segmentation;high-frequency enhancement;parameter-efficient fine-tuning;remote sensing images393|74|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27LSENet： A local subdomain-aware and edge-enhanced dynamic fusion network for semantic segmentation of airborne LiDAR point clouds AI Introduction
Abstract：Airborne LiDAR point-cloud semantic segmentation is fundamental to three-dimensional (3D) scene understanding and supports applications in smart cities, urban modeling, infrastructure extraction, land surveying, and environmental monitoring. Existing methods often inadequately coordinate local geometric representation with contextual modeling. To address this limitation, we propose LSENet (local subdomain-aware and edge-enhanced dynamic fusion network). LSENet extracts fine-grained geometric features through adaptive local-subdomain partitioning, facilitates information exchange through cross-subdomain interactions, and improves feature representation and boundary discrimination through an edge-enhanced dual-attention fusion module. Experiments on the Vaihingen 3D and DFC2019 datasets yielded Overall Accuracies (OAs) of 83.9% and 97.8% and mean F1 scores of 70.5% and 89.1%, respectively, demonstrating the effectiveness of LSENet and its applicability across scenes.LSENet uses a U-Net-style encoder–decoder architecture comprising three core modules. The Local Subdomain Space-aware feature Encoding and Aggregation (LSSEA) module constructs an adaptive local coordinate system using principal component analysis (PCA) and encodes relative position, spatial distribution, orientation, and curvature to obtain fine-grained geometric representations. The Local Subdomain feature Enhancement and Interaction Attention (LSEIA) module enhances features within each subdomain and establishes cross-subdomain interactions to improve local contextual modeling. The Edge-enhanced Grouped Multi-scale Dual-Attention dynamic fusion (EGMDA) module dynamically fuses edge-enhanced grouped channel attention with multi-scale spatial attention, thereby strengthening boundary representation and classification accuracy.Comparative experiments were conducted on two widely used benchmark datasets: ISPRS Vaihingen 3D and DFC2019. On Vaihingen 3D, LSENet achieved an OA of 83.9% and a mean F1 score of 70.5%, yielding the highest OA among the compared methods. Performance improved substantially for challenging classes, including power lines, roofs, and trees, confirming the effectiveness of the proposed modules in preserving complex structures and distinct boundaries. On DFC2019, LSENet achieved an OA of 97.8% and a mean F1 score of 89.1%; it again ranked first in OA and showed strong cross-scene generalizability. Visual comparisons further indicate that LSENet produces more spatially coherent regions and clearer boundaries than the baseline.LSENet integrates local geometric awareness, cross-subdomain interaction, edge-preserving dual attention, and dynamic fusion to address limitations of existing point-cloud segmentation networks. Results on two airborne LiDAR benchmarks confirm its effectiveness, robustness, and cross-scene generalizability. Future work will develop more discriminative geometric descriptors and lightweight dynamic-inference strategies and will evaluate LSENet on newly released, large-scale, diverse airborne LiDAR datasets to improve its applicability in real-world settings.Keywords：airborne LiDAR point clouds;deep learning;semantic segmentation;local subdomain awareness;dynamic fusion;edge-enhance297|86|0<HTML><L-PDF> <Enhanced-PDF> <Meta-XML>Updated：2026-10-27
Models and Methods
数据分布的基本统计量及遥感应用 AI Introduction
植被辐射传输模型概念介绍 AI Introduction
Remote Sensing Encyclopedia
