Coalbed methane (CBM), a type of self-sourced unconventional clean energy, occurs in coal seams and their surrounding rocks. Conventional exploration methods for CBM enrichment areas are laborious, while remote sensing provides a new approach to the rapid exploration of such areas. The basic principle behind the remote sensing-based exploration of CBM enrichment areas is as follows: ① The extraction and comprehensive analysis of multi-source data are conducted based on the comparison between the spectral features of typical surface features and those of surface feature anomalies, including rock and mineral alterations, vegetation anomalies, and thermal anomalies, caused by hydrocarbon micro-seepage in CBM enrichment areas, along with data obtained using geophysical prospecting methods like geological, seismic, and magnetotelluric methods; ② The distribution range and gas-bearing properties of CBM enrichment areas are gradually delineated. This paper reviews the hydrocarbon seepage in the CBM enrichment areas and the response mechanisms of spectral anomalies of surface rocks, minerals, and vegetation. It covers the applications of various methods based on the inversion of spectral parameters of surface rocks, minerals, and vegetation, together with the inversion of the spectral anomalies of surface features, in the exploration of potential CBM enrichment areas. Additionally, this paper elucidates the different explanations for surface thermal anomalies caused by CBM-bearing strata, as well as major methods to improve the accuracy of surface temperature inversion and their applications. In the future, the main approach to achieving low-cost, rapid exploration of CBM enrichment areas will be the analysis and information extraction of three-dimensional, multi-source information based on the combination of remote sensing technology with the geological data, seismic exploration, and geomagnetic prospecting of coalfields.

Snow proves to be both an important factor in characterizing the surface cryosphere and a critical parameter for weather and hydrological phenomena. Employing remote sensing to conduct long-term and large-scale monitoring of snow morphologies and their changes plays a vital role in research into global climate change, investigations into hydrology and water resources, and geological disaster prevention. After decades of development, significant progress has been made in the field of remote sensing-based snow monitoring technology both in China and abroad. Accordingly, the products for remote sensing-based snow monitoring have become increasingly abundant, and the snow-orientated inversion algorithms have been continuously improved. This paper provides a summary of the existing, widely applied products after categorizing them into three types: snow-cover extent (SEC), snow coverage, and snow depth/snow water equivalent (SWE) products. Furthermore, this study organizes the commercialized remote sensing inversion algorithms used in existing, typical SEC and SWE products. The review of advances in the relevant scientific research reveals that, with the constant presence of sensors with high temporal and spatial resolutions in China and abroad and the support of both novel optical and microwave data sources and new technologies, researchers have gradually improved the accuracy of snow-orientated inversion algorithms by optimizing these algorithms based on regional characteristics. This will provide more support for continuously improving remote sensing-based snow monitoring products in the future.

In the context of achieving peak carbon dioxide emissions and carbon neutrality, conducting a remote sensing-based ecological assessment and monitoring analysis is greatly significant for ascertaining the ecological condition in time and formulating scientific and reasonable ecological protection policies. The early remote sensing-based ecological assessment indices, simple and involving complex processes, are difficult to find wide applications. In contrast, the remote sensing ecological index (RSEI), contributing to elevated assessment efficiency, has been extensively used. To gain a deeper understanding of RSEI, this study describes its background, calculation method, and research status and provides a summary of the current issues and regional adjustments. Furthermore, it analyzes the main application directions of RSEI, namely the in-depth analyses of regional ecological assessment and change monitoring. Finally, the study proposes that despite a broad space for RSEI development, it is necessary to conduct research into the spatiotemporal scales of images, storage and batch processing capabilities, model adaption, and intelligentization.

Inland surface water bodies, including rivers, lakes, and reservoirs, are significant freshwater resources for human beings and ecology, and their monitoring and control are greatly significant. Optical remote sensing provides great convenience for the monitoring of surface water resources, proving to be an important means for the information extraction and dynamic monitoring of inland surface water bodies. This study reviews the basic principles, remote sensing data sources, methods, existing issues, and prospects of the information extraction of water bodies. Owing to the unique characteristics of the remote sensing images of inland surface water bodies, their information can be extracted in an accurate, scientific, and effective manner using remote sensing. Multiple remote sensing data resources can be applied to the information extraction, and the optical remote sensing-based extraction methods include the threshold value method, classifier method, object orientation method, and deep learning method. Given that different methods have unique advantages, disadvantages, and applicable conditions, selecting appropriate multi-source data and varying methods based on the conditions of study areas tend to improve the information extraction accuracy. Nevertheless, there still exist some issues in the optical remote sensing-based water body information extraction, such as the balance of spatiotemporal resolution of remote sensing data, the information mining of water body characteristics, the generalization ability of water body models, and the uniformity of criteria for accuracy evaluation.

Timely and accurate detection and statistical analysis of the spatial distributions and time-series variations of water bodies like rivers and lakes holds critical significance and application value. It has become a significant interest in current remote sensing surface observation research. Conventional water body extraction methods rely on empirically designed index models for threshold-based segmentation or classification of water bodies. They are susceptible to shadows of surface features like vegetation and buildings, and physicochemical characteristics like sediment content and saline-alkali concentration in water bodies, thus failing to maintain robustness under different spatio-temporal scales. With the rapid acquisition of massive multi-source and multi-resolution remote sensing images, deep learning algorithms have gradually exhibited prominent advantages in water body extraction, garnering considerable attention both domestically and internationally. Thanks to the powerful learning abilities and flexible convolutional structure design schemes of deep neural network models, researchers have successively proposed various models and learning strategies to enhance the robustness and accuracy of water body extraction. However, there lacks a comprehensive review and problem analysis of research advances in this regard. Therefore, this study summarized the relevant research results published domestically and internationally in recent years, especially the advantages, limitations, and existing problems of different algorithms in the water body extraction from remote sensing images. Moreover, this study proposed suggestions and prospects for the advancement of deep learning-based methods for extracting water bodies from remote sensing images.

Land surface temperature (LST) proves to be an important parameter in surface processes on regional and global scales, and its spatiotemporal information can be obtained through thermal infrared remote sensing. However, the constraints of thermal infrared sensors (TIRSs) themselves and the inability of thermal infrared electromagnetic waves to penetrate clouds render it impossible to obtain LST with a high spatiotemporal resolution currently. This study presents a method for reconstructing hourly LST at 100-m resolution in all weathers. This method consists of three main steps: ① cloudy LST at four moments is reconstructed using a moderate resolution imaging spectroradiometer (MODIS) based on the conventional annual temperature cycle (ATC) model; ② the daily variation curve of LST is estimated based on the daily trend in the skin temperature (SKT); ③ with spectral indices as regressors, spatial downscaling is conducted for the hourly LST using Extreme Gradient Boosting (XGBoost). The results show that the proposed reconstruction method can obtain spatiotemporally continuous LST products, improve the spatial resolution of LST, and provide more details. The validation of the hourly 100-m-resolution LST using data from the surface radiation budget network (SURFRAD) developed by the U.S. indicates that the reconstructed hourly LST exhibits roughly the same trend as the measured values of the SURFRAD. The method for reconstructing all-weather hourly LST boasts high accuracy, with R² of 0.95, a root mean squared error (RMSE) of 3.75 K, and a bias of 0.75 K.

Crown volumes serve as a crucial factor for surface ecological monitoring. Laser point clouds can characterize the fine-scale spatial morphologies of individual trees, providing a data basis for crown volume extraction. However, existing laser point cloud-based methods for extracting crown volumes of individual trees are sensitive to parameters and exhibit low degrees of automation. Based on the analysis of the three-dimensional morphological structures of individual trees, this study proposed a spherical coordinate integration method for extracting crown volumes of individual trees based on the terrestrial laser scanning (TLS) point clouds. First, the crown points were obtained through visual elevation threshold-based segmentation according to the elevation distributions of TLS point clouds. Then, the TLS point clouds were projected onto the spherical coordinate space for infinitesimal segmentation into triangular pyramids. Finally, the crown volumes were determined through the three-dimensional spherical coordinate integration. Six types of TLS point cloud data for individual trees were selected for tests. As indicated by the test results, the proposed method effectively considers factors like crown morphology and point cloud density, achieving a maximum absolute error of 2.33 m³ and a maximum relative error of 3.40% in the crown volume extraction of individual trees. It manifests higher extraction accuracy and stability compared to the existing methods. Therefore, this study holds significant reference value for extracting tree parameters based on TLS point clouds.

The monitoring and assessment of urban ecological quality holds critical significance for sustainable urban development. To assess the ecological quality of developed coastal cities in China in recent years, this study investigated Nanjing City based on the Sentinel-2A remote sensing images obtained in 2021. It constructed a novel remote sensing green index (RSGI) model involving green spaces, blue spaces, buildings, and impervious surfaces for assessing the ecological quality of Nanjing. First, neural network supervised classification was applied to the Sentinel-2A remote sensing images, constructing the RSGI to assess the ecological quality of various districts in Nanjing. Then, the correlations between the RSGI and urban ecological factors were analyzed using the Pearson correlation coefficient. Finally, the ecological similarity between the districts was analyzed using the agglomerative hierarchical clustering method. The results of this study are as follows: (1) The ecological quality of Nanjing presented a pattern of low RSGI values in the central portion and high RSGI values in the surrounding areas, with the highest and lowest RSGI values (0.86 and 0.38) observed in Luhe and Qinhuai districts, respectively, differing by 0.48; (2) The RSGI exhibited a positive correlation with the density of green spaces and negative correlations with the densities of population, buildings, and impervious surfaces, all at the 0.01 level; (3) With the ecological similarity of 70% as the threshold, 11 districts in Nanjing were categorized into four clusters: Qinhuai, Gulou, and Jianye districts in the first cluster, Yuhuatai and Qixia districts in the second cluster, Xuanwu and Gaochun districts in the third cluster, and the rest four districts in the fourth cluster. The results of this study can provide a scientific basis for subsequent urban planning and sustainable development of Nanjing.

Rapid identification and detection of landslides can both meet the requirement of timely responses to disasters and hold great significance for loss assessment and rescue post-disaster. This study proposed a deep learning-based automatic information extraction method for landslides to improve their detection accuracy. Specifically, the model input of this method includes the remote sensing images of the target areas, data from digital elevation models, and variation characteristics extracted using robust change vector analysis (RCVA). Furthermore, a U-Net model integrating dense upsampling and asymmetric convolution is designed to improve the identification accuracy. Taking Wenchuan, Sichuan Province as the study area, this study designed experiments to test the pixel-level image segmentation accuracy of landslides using different data combinations and methods. The results indicate that the improved U-Net model proposed in the study can produce the optimal image segmentation results of landslides.

Building heights are necessary for urban informatics, providing a significant basis for the planning and early warning of risks for urban construction. The shadow method, which can measure the heights of urban buildings on a large scale at a low cost, faces challenges such as low efficiency, accuracy, and robustness in building height inversion in complex shadow scenes. This study proposed a measurement method for these scenes. First, the shadows were measured and delineated using the fishing net method combined with multiple constraints. Second, the shadow lengths of all the zones divided were obtained, and the optimal values were determined using the quartile method and the bidirectional approximation strategy. Third, the shadow lengths were determined through a comprehensive assessment of the optimal values of all zones. The results show that 90.6% of building heights calculated using the new method exhibited absolute errors ranging from 0 to 5 m. Therefore, this method features elevated accuracy of building height inversion for various complex shadow scenes, laying a basis for research into the inversion and expansion of urban building heights.

Drought is identified as one type of the most serious natural disasters affecting agricultural production, and developing a comprehensive drought index holds practical significance for the assessment of drought in various districts and counties in Chongqing. By coupling the soil moisture and precipitation Z-index data downscaled using the XGBoost algorithm and using the Copula function, this study developed a gridded comprehensive drought index CSDIM-A on a kilometer scale with the boundaries of various districts and counties of Chongqing as the spatial division criteria and decades of a month as time units. Using this index, this study assessed the spatiotemporal characteristics of drought and performed an experimental demonstration of the index using Chongqing as the study area. The results indicate that the downscaling enhanced the spatial continuity of remote sensing-based products, thus providing support for the subsequent construction of a comprehensive drought index on a kilometer scale. The generalized extreme value distribution and the t location-scale distribution applied to the fitting of the data distributions of soil moisture and precipitation in most districts and counties of Chongqing, respectively, while the Frank-copula function suited for the fitting of the joint distribution of binary variables on a scale of a month decade. As validated based on soil moisture content, CSDI_M-A can more effectively reflect drought than the precipitation Z-index, with its spatial distribution in various districts and counties consistent with the actual drought data. This indicates that the CSDI_M-A can be used as a reference for drought assessment.

In the change detection for high-resolution remote sensing images, non-subsampled contourlet transform (NSCT) and change vector analysis (CVA) cannot ensure high detection accuracies under single thresholds due to significantly different changes in surface features. Hence, under the framework of change vector analysis in posterior probability space (CVAPS), this study proposed a NSCT-based change detection method combining fuzzy C-means (FCM) clustering and a simple Bayesian network (SBN): the FCM-SBN-CVAPS-NSCT method. First, the proposed method coupled FCM with an SBN to generate a change intensity map in posterior probability space. Then, the change intensity map was decomposed into submaps of different scales and directions through NSCT. The reconstructed change intensity map was optimized by preserving the details and eliminating noise in the high-frequency submaps. Finally, the multi-scale and multi-directional change detection in posterior probability space was achieved, enhancing the change detection accuracy. As indicated by the experimental results, the Kappa values obtained by the proposed method for three study areas were 0.100 9, 0.056 6, and 0.067 4 higher than those derived from the FCM-SBN-CVAPS method, demonstrating certain superiority.

Hyperspectral technology has become the major means of coastal wetland monitoring. However, traditional hyperspectral classification methods usually face challenges such as insufficient feature extraction, the same surface features corresponding to different spectra, and fragmented scenes. To solve these problems, this study proposed a new classification method by applying Transformer to hyperspectral classification. This vision Transformer (ViT)-based method expanded the receptive field by learning global spatial features using non-local technology, thus overcoming the insufficient extraction of discriminant features. Meanwhile, this method enhanced the cross-layer information interchange through cross-layer adaptive residual connection, thus eliminating information loss. This study, taking NC16 and NC13 wetland datasets of the Yellow River Delta as experimental data, compared the classification method proposed in this study to support vector machine (SVM), one-dimensional convolution neural network (1DCNN), contextual deep convolution neural network (CDCNN), spectral-spatial residual network (SSRN), hybrid spectral network (HybridSN), and ViT. The comparison results show that the new method yielded significantly elevated overall accuracy (OA) of up to 96.24% and 73.84%, average accuracy (AA) reaching 83.42% and 74.87%, and Kappa coefficients of up to 94.80% and 68.94%, respectively for the two datasets.

The pansharpening of remote sensing images is a process that extracts the spectral information of multispectral images and the structural information of panchromatic images and then fuse the information to form high-resolution multispectral remote sensing images, which, however, suffer a lack of spectral or structural information. To mitigate this problem, this study proposed a pansharpening algorithm for remote sensing images based on a multi-depth neural network. This algorithm includes a structural protection module and a spectral protection module. The structural protection module extracts the high-frequency information of panchromatic and multispectral images through filtering and then extracts the multi-scale information of the images using a multi-depth neural network. The purpose is to improve the spatial information extraction ability of the model and to reduce the risk of overfitting. The spectral protection module connects the upsampled multispectral images with the structural protection module through a skip connection. To validate the effectiveness of the new algorithm, this study, under consistent experimental conditions, compared the new algorithm with multiple pansharpening algorithms for remote sensing images and assessed these algorithms from the angles of subjective visual effects and objective evaluation. The results indicate that the algorithm proposed in this study can effectively protect the spectral information of multispectral images and the structural information of panchromatic images in solving the lack of structural information in current algorithms for image fusion.

Deep learning-based scene analysis of remote sensing images serves as a critical means for post-earthquake damage assessment. Given scarce images of damaged buildings, constructing high-quality sample sets of remote sensing scene images holds crucial significance for improving the accuracy of scene recognition and classification. The proportion of contextual information in scene images, as a significant reference for remote sensing analysis, is a key factor affecting the construction effects of sample sets. Currently, the appropriate proportion of contextual information remains under-studied in the sample set construction method. Aiming to construct high-quality sample sets, this study designed a method for adjusting the proportion of contextual information in scene images. It investigated the impacts of different proportions of contextual information on the construction of scene sample sets, exploring the optimal proportion range of contextual information. This study constructed six sample sets of scene images under different proportions of contextual information for training and testing in five classic convolutional neural network (CNN) models. It analyzed the classification results of all the CNN models under different proportions of contextual information. The results indicate that with the proportion of contextual information being 80%, the classification accuracy of the CNN reached an optimal value of 92.22%, which decreased to 89.03% with the proportion of contextual information at 95%. Among all the CNN models, GoogLeNet exhibited superior classification performance with an average accuracy of 93.13%. This study enables the setting of proper proportion ranges of contextual information in scene sample sets, thus effectively improving the classification accuracy of remote sensing scene images, and guiding the construction of sample sets of remote sensing scene images for damaged buildings.

The concentrations of negative oxygen ions and particulate matter 2.5 (PM_2.5) serve as important indicators in the assessment of the degrees of air freshness and cleanliness. Based on 2018-2022 data from 50 negative oxygen ion observation stations affiliated with the Fujian meteorological departments, along with the ecological parameters such as aerosol, vegetation index, and surface brightness temperature obtained by satellite-based remote sensing inversion, this study built estimation models for the concentrations of negative oxygen ions and PM_2.5 using the Cubist machine learning method. Accordingly, it developed an air freshness index (AFI), and the fine-scale mesh-based monitoring of regional air freshness was achieved. The results show that the estimation model for the negative oxygen ion concentration yielded goodness of fit of 0.838 and 0.526 for the training and test sets, respectively. In comparison, the estimation model for the PM_2.5 concentration exhibited goodness of fit of 0.968 and 0.867 for the training and test sets, respectively. Then, this study developed the AFI by comprehensively considering negative oxygen ions and PM_2.5. Then, this study graded the AFI using the frequency quartiles of the statistical data series combined with the spatiotemporal changes in negative oxygen ions. The results indicate that the AFI monitoring results based on meteorology, remote sensing, and machine learning algorithms are consistent with the actual conditions.

High-resolution remote sensing technology can greatly enhance the efficiency of investigations into active faults due to its high ability to identify the fine structures of microlandforms. This study presents a systematic summary of the symbols of remote sensing images for active faults. By comprehensively utilizing data from the Landsat8 and GF-2 satellites, as well as previous results and field geological surveys, this study analyzed and examined the active faults in the Maoyaba area of western Sichuan Province through the interpretation of remote sensing images of both macro- and microlandforms. The results show that, besides the Yidun-Litang fault zone, several nearly-W-E-trending normal active faults occur in the study area. Based on this finding, as well as the analysis of the regional geological setting, it can be concluded that crustal materials along the southeastern margin of the Qinghai-Tibet Plateau were continuously squeezed out laterally under the background of the intense collision and compression between the Indian and Eurasian plates, leading to the formation of two conjugate faults: the dextral Batang strike-slip fault and the sinistral Litang strike-slip fault. The joint control of both faults resulted in the local extension of the study area and the formation of nearly-W-E-trending fault structures, which govern the development and evolution of the Damaoyaba Basin, the Xiaomaoyaba Basin, and the Cuopu Basin in the north.

Assessing regional habitat quality holds great significance for maintaining regional biodiversity, enhancing human well-being, and achieving regional sustainable development. Based on the land use data of 2000, 2010, and 2020, this study analyzed the spatiotemporal characteristics of the habitat quality in the ecological green integrated demonstration area in the Yangtze River Delta using the InVEST model and the habitat quality index method. Furthermore, this study explored the relationship between regional habitat quality and land use. Key findings are as follows: ① From 2000 to 2020, the study area exhibited moderate habitat quality, with the habitat quality index trending downward and the habitat degradation gradually mitigating. Regarding the districts and counties in this area, Qingpu District of Shanghai, Wujiang District of Suzhou, Jiangsu, and Jiashan County of Jiaxing, Zhejiang (the two districts and one county) showed a downward trend in the habitat quality from 2000 to 2010. In contrast, from 2010 to 2020, Qingpu District and Jiashan County exhibited improved habitat quality, while Wujiang District still maintained a downward trend in the habitat quality; ② From 2000 to 2020, the study area primarily featured moderate habitat, with major land types including cultivated land and grassland. During this period, the water and wetland areas in the north and center, respectively exhibited the highest habitat quality, while the construction land in the study area displayed poor and inferior habitat; ③ There is a strong correlation between the habitat quality and land use structure in the study area. Specifically, areas with more intense changes in land use feature more significant variations in habitat quality. The results of this study will provide a reference for biodiversity conservation and land use management in the study area.

Landslides may cause the loss of lives and property, and an accurate and complete map showing the spatial distribution of landslides and the determination of landslide susceptibility areas assist in guiding the optimization of the production, living, and ecological spaces. However, landslide investigations are complicated by dense vegetation. LiDAR technology enables the presentation of actual terrain features, thereby achieving landslide identification in densely vegetated areas. This study obtained the LiDAR point cloud data of the study area through ground-imitating flight and then built a digital elevation model (DEM) through data processing. Then, based on mountain shadow analysis, color-enhanced presentation, and 3D scene simulation, the locations and scales of existing landslides in the study area were identified. The field verification revealed an interpretation accuracy of landslides of up to 86.4%. For the assessment of landslide susceptibility areas, this study, with existing landslides as samples, delineated landslide susceptibility areas through remote sensing classification for the first time. Specifically, images were synthesized using the landslide-related elevations, slopes, and surface undulations, and then landslide susceptibility areas were determined using the support vector machine (SVM) classification method. The analysis of the inspection samples reveals a landslide identification accuracy of 81.91%. The results show that the image identification based on high-accuracy LiDAR data and visually enhanced images allows for the delineation of small landslides and that the SVM classification method enables the accurate location of landslide susceptibility areas. This study provides a basis for the future planning and optimization of the production, living, and ecological spaces.

This study constructed an assessment index system for ecological vulnerability based on the vulnerability scoping diagram (VSD) model. It dynamically assessed the three-phase ecological vulnerability of Kangding City from 2011 to 2019 using the analytic hierarchy process - principal component analysis - entropy weight method (AHP-PCA-EWM). Through the analysis of spatio-temporal variations, spatial correlations, and driving factors, it revealed the spatio-temporal differentiation characteristics and driving mechanism of Kangding’s ecological vulnerability, aiming to provide suggestions for the ecological restoration, conservation, and sustainable development of Kangding. The results of this study are as follows: ① Throughout the study period, Kangding exhibited an overall moderate ecological vulnerability, with increased potentially- and slightly-vulnerable areas and a decreased severely vulnerable area, suggesting a promising ecological evolutionary trend. Moreover, the ecological vulnerability of Kangding manifested spatial distributions characterized by high-high clusters in the western and southeastern portions and low-low clusters in the northeastern and middle portion; ② The spatial distributions of ecological vulnerability in Kangding were subjected to various internal and external factors, with natural driving factors like vegetation cover, biological abundance, soil and water conservation, and meteorology being the dominant ones.

Pine wilt disease (PWD) is identified as a major disease endangering the forest resources in China. Investigating the deep semantic segmentation network (DSSN)-based unmanned aerial vehicle (UAV) remote sensing identification can improve the identification accuracy of discolored trees infected with PWD and provide technical support for the enhancement and protection of the forest resource quality. Focusing on the pine forest in Laoshan Mountain in Qingdao, this study obtained images of suspected discolored trees through aerial photography using a fixed-wing UAV. To examine four deep semantic segmentation models, namely fully convolutional network (FCN), U-Net, DeepLabV3+, and object context network (OCNet), this study assessed the segmentation accuracies of the four models using recall, precision, IoU, and F1 score. Based on the 2 688 images acquired, 28 800 training samples were obtained through manual labeling and sample amplification. The results indicate that the four models can effectively identify the discolored trees infected with PWD, with no significant false alarms. Furthermore, these deep learning models efficiently distinguished between surface features with similar colors, such as rocks and yellow bare soils. Generally, DeeplabV3+ outperformed the remaining three models, with an IoU of 0.711 and an F1 score of 0.711. In contrast, the FCN model exhibited the lowest segmentation accuracy, with an IoU of 0.699 and an F1 score of 0.812. DeeplabV3+ proved the least time-consuming time for training, requiring merely 27.2 ms per image. Meanwhile, FCN was the least time-consuming in prediction, with only 7.2 ms needed per image. However, this model exhibited the lowest edge segmentation accuracy of discolored trees. Three DeepLabV3+ models constructed using Resnet50, Resnet101, and Resnet152 as front-end feature extraction networks exhibited IoU of 0.711, 0.702, and 0.702 and F1 scores of 0.829, 0.822, and 0.820, respectively. DeepLabV3+ surpassed DeepLabV3 in the identification accuracy of discolored trees, with the letter showing an IoU of 0.701 and an F1 score of 0.812. The train data revealed that DeepLabV3+ exhibited the highest identification accuracy of the discolored trees, while the ResNet feature extraction network produced minor impacts on the identification accuracy. The encoding and decoding structures introduced by DeepLabV3+ can significantly improve the segmentation accuracy of DeepLabV3, yielding more detailed edges. Therefore, DeepLabV3+ is more favorable for the identification of discolored trees infected with PWD.

To explore the feasibility and accuracy of the method of utilizing hyperspectral data to estimate the contents of elements Cd and As for soil quality elevation of cultivated land, this study delves into the extraction of characteristic bands of the spectra of both elements and the modeling of quantitative hyperspectral inversion. The characteristic bands of spectra were extracted using multiple methods derived from the combination of four spectral transformations and two feature selection methods, with the former comprising first-order /second-order differential (FD/SD), reciprocal logarithm (LR), and continuum removal (CR) and the latter consisting of the competitive adaptive reweighted sampling (CARS) method and the Pearson correlation coefficient (PCC) analysis. Based on this, the element content inversion was conducted using the partial least squares regression (PLSR) and the particle swarm optimization optimized random forest regression (PSO-RFR), followed by the verification of inversion accuracy. The results indicate that the FD-CARS-PLSR inversion model exhibited the best prediction effect for both elements, with maximum determination coefficients R² of 0.863 and 0.959 and relative percent differences (RPDs) of 2.799 and 5.119 for Cd and As, respectively. The FD and SD spectral transformations combined with the CARS method can improve the accuracy of the PLSR inversion model. The results of this study can provide a reference for the rapid estimation of the contents of Cd and As in soil.

Based on the survey and monitoring results obtained using satellite remote sensing technology, this study investigated the nonmetallic mines for building materials under the jurisdiction of municipal administrative units in the Tibet Autonomous Region. It conducted an exploratory analysis by correlating the land damage, the restoration and control of mine environments, and the damage scales of mining areas in these mines with typical socio-economic factors like gross regional product (GRP), population, zoning area, and population density. The results show that higher GRP or population density is associated with a larger mining scale of nonmetallic mineral resources for building materials. This conclusion provides a research basis for predicting the mining scales of nonmetallic mines for building materials in certain regions in China’s western provinces and achieving harmonious development between the geological environments of such mines and socio-economic construction.

This study aims to explore the object-oriented classification based on moderate-resolution remote sensing data. Using the Landsat8 OLI, Landsat5 TM, and GF1 data obtained from the northern mountainous area and the southern plain area in Hebei Province, this study compared the land cover classification effects of four classifiers: support vector machine (SVM), random forest (RF), decision tree (DT), and naive Bayes (NB). Moreover, it analyzed the impacts of critical parameters in SVM, RF, and DT on the classification results. The findings indicate that the classification results of the classifiers vary slightly in the two study areas, with their effects decreased in the order of SVM, NB, RF, and DT. The classification accuracies of SVM and DT fluctuated significantly with parameter changes. With C values not below 10³ and gamma values not exceeding 10^-1, SVM can yield classification accuracies above 90% in all cases. With depth values over 3, DT exhibits relatively high and stable classification accuracies. With parameter changes, RF manifests slightly varying classification accuracies with nonsignificant variation patterns. The results of this study serve as a reference for exploring the object-oriented land cover classification based on moderate-resolution remote sensing data.

This study aims to solve the challenges faced in the prediction of sugarcane yield in Guangxi, such as varied crops, complex investigations in the sugarcane planting areas, and difficult acquisition of remote-sensing images caused by the changeable weather. To this end, an improved semantic segmentation algorithm based on Sentinel-2 images was proposed to automatically identify sugarcane planting areas, and an extraction method for representative spectral features was developed to build a sugarcane yield prediction model based on multi-temporal Sentinel-2 and Landsat8 images. First, an ECA-BiseNetV2 identification model for sugarcane planting areas was constructed by introducing an efficient channel attention (ECA) module into the BiseNetV2 lightweight unstructured network. As a result, the overall pixel classification accuracy reached up to 91.54%, and the precision for sugarcane pixel identification was up to 95.57%. Then, multiple vegetation indices of different growth periods of the identified sugarcane planting areas were extracted, and the Landsat8 image-derived vegetation indices were converted into Sentinel-2 image-based ones using a linear regression model to reduce the differences of the indices derived using images from the two satellites. Subsequently, after the fitting of time-series data of the extracted vegetation indices using a cubic curve, the maximum indices were obtained as the representative spectral features. Finally, a yield prediction model was built using multiple machine learning algorithms. The results indicate that the test set of the decision tree model built using the fitted maximum values of the vegetation indices yielded R? of up to 0.759, 4.3%, higher than that (0.792) of the model built using the available actual maximum values. Therefore, this method can effectively resolve the difficulty in developing an accurate sugarcane yield prediction model caused by changeable weather-induced lack of remote sensing images of sugarcane of the key growth periods.

The Jinsha River basin is a typical area with a high incidence of geologic hazards in China. To accurately identify the potential landslide hazards in the basin, this study processed the data of Sentinel-1A’s ascending and descending orbits using the SBAS-InSAR technique. From two detection directions, it conducted early landslide identification and deformation monitoring of the Baige landslide on the bank of the Jinsha River and its lower reaches covering approximately 100 km. The results show that: ① The combined detection based on Sentinel-1A’s ascending and descending orbits effectively reduced the interference of geometric distortions, enabling the identification of long-term creep hazard sites; ② The deformation rates along the line-of-sight (LOS) of ascending and descending orbits ranged from -142 to 80 mm/a and -71 to 56 mm/a, respectively. Combined with the visual interpretation of optical remote sensing images, two large landslide clusters consisting of nine landslides were detected; ③ The analysis of surface deformation characteristics was conducted on three typical landslides: the Sela landslide, the Shadong (Xiongba) landslide, and the Nimasi talus slide. The analysis results reveal that the maximum deformation was associated with the peak rainfall and river runoff, which constituted the significant factors influencing landslide deformation. The results of this study serve as a reference for the prediction, early warning, prevention, and control of basin-scale geologic hazards in flood seasons.

In recent years, target detection, as an important branch of computer vision technology, has been widely applied in fields such as medicine, military affairs, and urban rail transit. As satellite and remote sensing technologies advance, images obtained using these technologies contain abundant information. This makes it crucial to conduct automatic target detection and understanding of these images. However, due to the random directions and dense distribution of targets in remote sensing images, conventional methods are prone to lead to missing or incorrect detection. In response, this study proposes a multi-convolution kernel feature combination-based adaptive region proposal network (MFCARPN) algorithm for multi-directional detection. This algorithm introduces multiple convolution kernel features for target extraction. The weight parameters of these convolution kernel features can be determined through adaptive learning according to the differences between the targets, yielding the characteristic patterns that match better with targets. Meanwhile, in combination with the original features of the targets, the parameters of the classification and regression model vary dynamically according to the difference between targets. Thus, the RPN’s adaptive ability can be improved. The experimental results indicate that the mAP of the standard dataset DOTA reached up to 75.52%, which is 0.5 percentages higher than that of the baseline algorithm GV. Therefore, the MFCARPN algorithm proposed in this study proves effective.