Hyperspectral (HS) images are highly accurate for mineral discrimination, but available areas are limited. For this reason, several methods have been proposed to extend the mineral map of the overlap region between HS and multispectral (MS) images to the surrounding area with no HS image. One such method, proposed by Hirai and Tonooka, discriminates minerals using MS images by obtaining the endmembers of MS images from the positions of the endmember pixels of HS images in the overlap region. While this method (referred to as HT method) has the advantage of being less susceptible to the spectral distortions of HS and MS images, it also has the problem of reduced accuracy due to misalignment between HS and MS images. We proposed an improved HT method that reduces the effects of the above problems by incorporating a process that improves the robustness against misalignment by searching for the best MS endmember pixel around the position of the HS endmember pixel and a process that determines more optimum threshold value of each mineral in the spectral angle mapper method used in the HT method. As a result of evaluation using an AVIRIS image as an HS image and a World View-3 image as an MS image at Cuprite, Nevada, the improved method improved the overall accuracy by 2.6% compared with the original HT method, and in the case that the HS and the MS images were misaligned, the overall accuracy of the original method decreased by 7.0%, while the improved method decreased by only 1.5%. These results indicate that the improved method can perform as expected.

A range-angle–dependent beam pattern can be produced by frequency diversity array (FDA) due to the small frequency offsets between the array elements. The beam pattern can be used to automatically scan an area in its entirety and estimate the distance and angle to a target. However, for constant tracking of the target after recognition, energy is wasted in the scanning mode because of periodic scanning of the main lobe of the beam. To eliminate this energy waste, we propose multiple repeated subpulses of FDA. This scheme achieves stable tracking without range–angle coupling. This special transmission method considerably improves the transmission energy and signal-to-noise ratio, while ensuring accurate range detection and high resolution. The results of a feasibility analysis and simulation experiments verify the superiority of the proposed method.

I make a case to abandon the use of the term “ground truth.” After a brief history of the term that indicates its military origins, two main arguments are presented against the use of the term. The first is from measurement theory. The second is by way of presenting three examples: the first is in crop mapping, in which “ground truth” may have some validity. This is presented as the exception to prove the rule. The second example looks at albedo, in which the measurand is the same for the local and remote measurements, and evidence is provided to support this paper’s argument. The third example looks at the scattering phase center in radar. In this case, the measurand is different for the local and remote measurements, and an argument is given for why this does not warrant the use of the term. Finally, a heuristic checklist that seeks to guide readers to reflect on whether or not their particular field measurements should be referred to as “ground truth” is provided.

The guest editors introduce the Special Section on Multitemporal Remote Sensing Data Processing and Applications.

The counting of crop plants is an important basis for the estimation of pitaya yield. Traditional crop monitoring methods are time-consuming and laborious. The pitaya tree, one of the characteristic economic crops in the complex mountain environment, was selected as the research object. Considering the comprehensive factors such as different seasons, cloud shadow shading, crop interplanting, steep terrain, different breeds, and ages of pitaya trees, the quad-rotor unmanned aerial vehicles (UAVs) were used to collect the visible light images of the planting base of pitaya trees in the test area in karst plateau canyon region and the verification area in a complex environment. First, the characteristics of the RGB values of the eight main ground objects in the test areas are analyzed according to the color index, geometric size, and texture characteristics, which show that pitaya plants, weeds, and shrubs have certain interference with each other. Then the excess green index (ExG) of the images is calculated to enhance the vegetation characteristics and separate vegetation and non-vegetation. Gaussian high-pass filter (GHPF) is used to retain the high-frequency information of pitaya plants on the ExG images. After GHPF processing, the DN values of shrub and weeds are reduced, the edges of the area target are enhanced, and the influences of weeds and cloud shadows on plant identification are eliminated. Finally, through field measurement of pitaya plant data and OTSU, grayscale segmentation was performed on the images processed by GHPF and the pitaya plant information was extracted. Combined with the projection area of single pitaya tree, the number of pitaya trees was obtained by area screening method. The target detection percentage of test area A, test area B, and the verification area in complex environment is, respectively, 96.99%, 94.66%, and 94.30%; the quality percentage of them is, respectively, 92.46%, 90.41% and 91.50%; the branching factor of them is, respectively, 0.05, 0.05, and 0.03, proving that the RGB images collected by UAV can be used in the recognition and counting of pitaya trees in complex mountain environments. Our study can provide a reference for the application of low-cost, high-efficiency UAV visible light remote sensing to precision agriculture in complex mountain environment.

In recent years, ecological and environmental degradations have occurred widely around the world, especially in urban areas with rapid economic development, intensive human activities, and massive urbanization. We aim at proposing an effective and comprehensive urban ecological environment evaluation method based on remote sensing. Harmonic analysis of time series method and spatial–temporal information fusion based on nonlocal means filter was used to improve the quality of long-term remote sensing data from 2003 to 2019. The greenness, dryness, heat, and atmospheric turbidity of the ecological environment were represented by normalized difference vegetation index, normalized difference built-up index, land surface temperature, and aerosol optical depth, respectively. An urban ecological comfort index (UECI) was established by the weighted combination of the above four indicators using entropy method. UECI was calculated at seasonal and annual scales from 2003 to 2019 in Hefei city, China, as a case study. Furthermore, Mann–Kendall trend test, spatial autocorrelation analysis, and spatial clustering analysis were performed on the UECI results. The experimental results show that UECI in Hefei had an increasing trend from 0.5446 to 0.5737 between 2003 and 2019, indicating that its ecological environment is actively improving. UECI is roughly equal in spring, summer, and autumn, and the worst is in winter. The responses of UECI to urban expansion, environmental protection policies, and extreme weather were also explored. Its sensitively analysis results and the fact that the spatial–temporal trends of UECI were consistent with the influencing factors reflected the stability and reliability of UECI, indicating that the proposed UECI performed well in Hefei and could be applied to measure urban ecological comfort level in other study areas.

The Shengjin Lake wetland is a national wetland with rare wintering waterfowl as its main protection object, which is also a significant wetland in the world. It is of great significance to study the land use change and its driving forces for the protection of rare waterfowl and the restoration of ecological environment. The study of land use changes and driving forces in Shengjin Lake wetland can provide reference for land use pattern optimization, the protection of rare waterfowl, and the restoration of ecological environment of Shengjin Lake wetland. Taking Shengjin Lake wetland as an example, based on nine remote sensing images of Shengjin Lake wetland from 1986 to 2019, the land use change process of Shengjin Lake wetland was analyzed by using geographic information technology. Meanwhile, driving factors of land use change were analyzed using mathematical statistics method. The following results were obtained: The areas of grassland, paddy fields, tidal land, and other land types increased from 1986 to 2019, whereas areas of woodland, dry land, water, and reed beach dropped. Land use composite index fluctuated slightly but was at a moderate level overall, indicating land use was in the period of development. The main driving forces of land use change in the study area included terrain topography, climate, population change, social-economic development, and policy factors. Each land use type had its own dominant driving factor. And changes of any type of land use were attributed to multiple driving factors instead of a single one. To summarize, the natural factor provides the intrinsic motivation for land use change, but population is the fundamental factor.

The monitoring of rivers based on remote sensing data provides an opportunity for the observation of the natural dynamics of water and flood conditions. Taking advantage of the availability of radar data, we describe a methodology used to detect the river’s water surface flooding patterns and the size of floods using observations from Sentinel-1 imagery. With the proposed methodology, characteristics of the river’s normal average riverbed surface are determined, as well as the flood footprint. As a case study, it shows the annual and interannual water variability at the lower Coatzacoalcos Basin in the state of Veracruz, Mexico. This site was chosen after considering the evidence of seasonal flooding and the changes that happened during the 2015 to 2017 period. Our paper details the results on surface changes during this time period. The results on water surface cover have been validated through different sources of information to confirm the ability to differentiate flood areas and know the maximum flood extensions. An agreement above 92% is estimated to be reached with the use of the evaluated maps of binary water/no-water. Recurrent flood areas were detected where water remained during one or several weeks during the month of October, both in 2015 and 2017. Finally, a change in the extent of water maps in the framework of the Sustainable Development Goals is achieved with the integration of historical data over time and space of the natural behavior of the rivers in the case study.

The monitoring of threatened habitats is a key objective of European environmental policies. Due to the high cost of current field-based habitat mapping techniques, there is keen interest in proposing solutions that can reduce cost through increased levels of automation. Our study aims to propose a habitat mapping solution that benefits both from the merits of convolutional neural networks (CNNs) for image classification tasks, as well as from the high spatial, spectral, and multitemporal unmanned aerial vehicle image data, which shows great potential for accurate vegetation classification. The proposed CNN-based method uses multitemporal multispectral aerial imagery for the classification of threatened coastal habitats in the Maharees (Ireland) and shows a high level of classification accuracy.

Nitrogen is one of the essential nutrients required for crop growth, and hence should be applied efficiently for attaining optimum yield. To fulfil nitrogen need, absorbed nitrogen in the plant is required to be estimated. Various methods are available to estimate crop nitrogen such as tissue analysis using the methods of Kjeldahl and Dumas, which are accurate, but time-consuming and destructive. Satellite imagery provides a more extensive field view. However, they are limited to their spatial and temporal resolution. Unmanned aerial vehicle (UAV) is emerging as a promising tool that can provide the status of crop nitrogen rapidly with high spatial and temporal resolution. The objective of the study was to evaluate UAV-based imageries to show nitrogen status and predict rice yield at different growth stages. The experiments were conducted using two rice cultivars, six nitrogen applications, two water management practices, and with three replications. Soil plant analysis development (SPAD) meter readings were collected at various growth stages. First, aerial imageries of experimental site were collected using an octocopter UAV equipped with a multispectral sensor that provides reflectance values in four different bands (red, green, red edge, and near-infrared) along with SPAD values for respective flight. Second, aerial images were processed in pix4D software, to identify the most appropriate vegetation index that shows nitrogen status variation in the field and to predict yield using different vegetation indices. Nine vegetation indices were considered: ratio vegetation index, normalized difference vegetation index, normalized green red difference index, red edge difference vegetation index, green ratio vegetation index, green normalized difference vegetation index (GNDVI), wide dynamic range vegetation index, transformed normalized difference vegetation index (TNDVI), and normalized difference red edge. After that, a linear regression model was developed between the representative index and SPAD values. Finally, linear regression models developed by using VI and SPAD values were evaluated and results revealed that GNDVI-based model simulates SPAD values with R² of 0.49, 0.49, and 0.74 at panicle, milky, and booting stages, respectively. It is also found that TNDVI-based linear regression model predicts yield with R² of 0.71 at milky stage.

The coconut tree is an important cash crop in coastal tropical regions. These trees undergo substantial damage during annual cyclones. Early and effective assessment of damage is important to farmers for appropriate monetary compensation. However, the identification of coconut trees presents multiple challenges: (i) the trees are with other vegetation and (ii) cloud cover during monsoons and cyclones precludes using simple imaging techniques. Although automated approaches based on classical machine learning techniques have been attempted for other crops on remote sensed images, these are not adequate for identifying coconut trees from different land cover. We present a hybrid approach termed CocoNet that combines (i) unsupervised K-means for pixel-based coarse classification of land cover and (ii) a patch-based convolutional neural network for fine classification of vegetation to identify coconut farms. The challenge presented by cloud cover is addressed using synthetic aperture radar (SAR) images. Change detection is then performed on bitemporal Sentinel-1 SAR images taken before and after a cyclone to obtain the change map depicting the damage caused. Experimental results show that the proposed two-phase CocoNet model gives a classification accuracy of 92.5% and a change detection accuracy of 89.1%.

Since multitemporal hyperspectral imaging has an excellent ability to observe the Earth’s surface over time, it has been used for various remote sensing applications. On the other hand, multitemporal hyperspectral images (HSIs) contain HSI sequences acquired multiple times over the same scene, resulting in large amounts of data. Conventional HSI compression methods cannot benefit from temporal correlation, which can be very high, depending on the acquisition cycle. We propose a prediction and compression framework that directly considers temporal correlation for the compression of HSIs. The main objective of the proposed method is to predict each spectral signature in the target HSI from the corresponding spectral signature of the reference HSI using a long short-term memory network model that supports clustering. Then, the residual image between the predicted HSI and the target HSI is quantized and entropy encoded for the compression purpose. The experiments are conducted on a ground-based multitemporal dataset named Noguiero, which contains nine HSIs, in terms of prediction and compression performances. Experiments show that the proposed method not only provides the best quality metrics from the perspective of prediction but also has convincing compression performances compared to the other methods.

Based on the data provided from the new satellite constellation, the crop phonology can be mapped accurately with a high spatial and temporal resolution. The potential of Sentinel-2 (high spatial and temporal resolution, 10 m and 5 days) is investigated in order to monitor and forecast the crop growth over different olive groves located in Sfax, Tunisia, from the beginning of 2016 to the end of 2020. The normalized difference vegetation index (NDVI) is used as an indicator of vegetation health due to its high correlation with crop growth health and productivity, and it is particularly forecasted to predict the trends. The prediction is done using two different statistical methods, autoregressive (AR) and Markov chain (MC), based on historical data derived from Sentinel-2 data. The prediction is applied for three different areas having various vegetation types and density. Moreover, for an accurate and precise prediction, our study areas are divided into different homogeneous clusters using the well-known Gaussian mixture model. Furthermore, the performances of our approach are assessed by means of the root mean squared error (RMSE) between the predicted results and the actual data. Globally, the obtained results for all the clusters of each area show that the forecast, using both methods, is accurate where the error is less than 5%. Nevertheless, the MC model displays the highest performance where the predicted NDVI curves of the different study areas are the closest to the actual observation. MC (resp. AR) precision is of 97% with RMSE  =  0.025 (resp. 95% and RMSE  =  0.041) for all the clusters of Jbeniana and Limaya, and 99% with RMSE  =  0.0073 (resp. 98% and RMSE  =  0.0127) for the four clusters of Chaâl.

Combining multiple satellite remote sensing sources provides a far richer, more frequent view of the earth than that of any single source; the challenge is in distilling these petabytes of heterogeneous sensor imagery into meaningful characterizations of the imaged areas. Meeting this challenge requires effective algorithms for combining multi-modal imagery over time to identify subtle but real changes among the intrinsic data variation. Here, we implement a joint-distribution framework for multi-sensor anomalous change detection (MSACD) that can effectively account for these differences in modality, and does not require any signal resampling of the pixel measurements. This flexibility enables the use of satellite imagery from different sensor platforms and modalities. We use multi-year construction of the SoFi Stadium in California as our testbed, and exploit synthetic aperture radar imagery from Sentinel-1 and multispectral imagery from both Sentinel-2 and Landsat 8. We show results for MSACD using real imagery with implanted, measurable changes, as well as real imagery with real, observable changes, including scaling our analysis over multiple years.

Forest fire is a major hazard around the world that seriously affects the terrestrial, aquatic, and atmospheric systems. Remote sensing methods are found to be efficient in mapping forest fires and analyzing the postfire effects through change detection. The land surface temperature (LST) also gets affected with the change of vegetation cover over the region due to the forest fires. This work pertains to analysis of postevent effects of forest fires and regrowth of vegetation using LST and normalized difference vegetation index (NDVI) of 2014, in which many forest fires occurred, and of 2018, to assess the vegetation regrowth in the forest range in and around Tirupati region (Andhra Pradesh, India). The LST was estimated using monowindow algorithm and NDVI by band ratioing method for the Landsat 8 imagery. The Landsat 8 datasets for February and March of 2014 and March of 2018 were used in the study. Analysis using LST and NDVI showed that 112-sq km area was affected by forest fires and detected forest regrowth. A change of 35.8% and 14.69% in mean NDVI and mean temperatures, respectively, was observed for the study area.

High-accuracy forest type mapping is crucial for forest resource inventory to support forest management, conservation biology, and ecological restoration. The Chinese GF-1 satellite can provide high-spatial resolution and low-cost data for forestry monitoring. However, it is difficult to obtain time-series GF-1 images with high spatial resolution and high temporal frequency because of technical constraints and cloud contamination, which affects the accuracy of forest type extraction. We proposed a method for mapping forest type using Chinese GF-1/widefield view (WFV) multi-spectral data, the fused GF-1/WFV normalized difference vegetation index (NDVI) time series, and the phenological parameters. We used the enhanced spatial and temporal adaptive reflectance fusion model algorithm to fuse Chinese GF-1/WFV and Moderate Resolution Imaging Spectroradiometer NDVI to generate high-spatiotemporal resolution NDVI time-series image data and extract phenological parameters in the northeastern part of Hubei Province, China. We then designed four different combinations of GF-1/WFV spectral features, the fused NDVI time-series features, and the phenological parameters. Finally, a random forest algorithm is used to classify forest types in the study area. The results showed that the proposed method can obtain satisfactory results. The overall accuracy based on the combination of GF-1/WFV spectral features, high-resolution time-series NDVI, and phenological features was 89.68%. Compared with the classification using only GF-1/WFV spectral data, the overall accuracy was improved by 9.60% points. These results demonstrate that the proposed method can support forest type survey and mapping at the regional or global scale.

Building change detection is an important task in urban applications. In the last decades, airborne LiDAR data have been explored, aiming at automatic or semi-automatic detection of different entities or objects. We show a comparative analysis considering three methods: two previously developed by the authors (called M1 and M2) and another based on the results derived from an open-source software (M3). The first developed method (M1) explores the height entropy concept and is based on threshold empirically determined to separate building and vegetation changes. The second method (M2) considers the planarity attribute and the Otsu algorithm to automatically separate the classes. The main purpose is to highlight differences among the methods, as well as discuss advantages and disadvantages considering a real scene, in which buildings with at least 20  m² were considered. To perform the comparative analysis, qualitative and quantitative evaluation were conducted considering a study area located in the city of Presidente Prudente, Brazil. In the experiments, two airborne LiDAR datasets were used, acquired in 2012 and 2014. The results indicate the potential of methods M1 and M2, presenting F_score around 80% and 83%, respectively. In contrast, the method M3 presented a F_score around 50%.

Guest editors Kun Tan, Xiuping Jia, and Antonio J. Plaza summarize the Special Section on Satellite Hyperspectral Remote Sensing: Algorithms and Applications.

Sparse unmixing has been proven to be an effective hyperspectral unmixing method. The row-sparsity model (using l_2,0 norm to control the sparsity) has outperformed single-sparsity unmixing methods in many scenarios. However, to avoid the NP-hard problem, most algorithms adopt a convex relaxation strategy to solve the l_2,0 norm at the expense of unmixing accuracy and sparsity. In addition, the row-sparsity model might cause aliasing artifacts on the boundaries. To solve these problems, a novel algorithm called two-step iterative row-sparsity hyperspectral unmixing via a low-rank constraint (TRSUnLR) is proposed. TRSUnLR introduces a row-hard-threshold function to solve the l_2,0 norm directly. The low-rank constraint, which can make full use of the global structure of data, is imposed to alleviate the aliasing artifacts. The reweighting strategy is used to further enhance the sparsity. Then we adopt the two-step iterative method under the alternating direction method of multipliers framework to solve the proposed algorithm. Specifically, the current solution is computed by a linear combination of the solutions of two previous iterations. Simulated and real data experiments have proven that the proposed algorithm can obtain better unmixing results.

Spaceborne hyperspectral images are expected to have abundant applications in various fields, particularly in the quantitative observation of the Earth. However, the problem of low spatial resolution has limited their effectiveness to some extent. There are urgent needs for high-spatial-resolution hyperspectral images. We explore fusing the hyperspectral image with panchromatic and multispectral images captured by sensors onboard ZY-1 (02D) to obtain high-spatial-resolution and high-spectral-resolution images. Four approaches to obtain the final high-spatial-resolution hyperspectral image data are proposed, and six fusion methods are used. The fusion results were evaluated using quantitative indicators and classification performance, and image quality greatly improved after fusion. The fusion approach that used multispectral image data as an intermediate layer twice in the fusion process obtained the best fusion effect, and this was verified by both quantitative and application evaluation results. This provides an effective approach to obtain a high-quality, high-spatial-resolution hyperspectral image using the combination of panchromatic, multispectral, and hyperspectral images acquired by the sensors onboard ZY-1(02D).

Self-training is commonly used for hyperspectral imagery semi-supervised learning, which selects confident unlabeled samples and puts them into the labeled set iteratively. However, self-training suffers from the influence of noises and outliers that are wrongly selected as confident samples and guide the accuracy of the classifier going down in iteration. To solve this problem, a self-training algorithm for hyperspectral imagery classification based on a mixed measurement k-nearest neighbor (k-NN) and support vector machine (SVM) framework is proposed. Different from the traditional self-training based on k-NN and SVM, the proposed method utilizes a mixed measurement k-NN to select confident unlabeled samples, and the SVM is applied to help the k-NN method to label the unlabeled samples. The mixed measurement is combined with the spectral distance, spatial distance, and local outlier factor (LOF) distance to explore the feature differences of spectra. Meanwhile, the factors of LOF and k-NN are obtained adaptively, which effectively reduces the computational complexity of the proposed method to select optimized parameters. Experimental results on five hyperspectral images show that the proposed method outperforms the competitors.

Surface water is an essential carrier for the balance of ecosystems and human existence. As applied hyperspectral remote sensing has improved, the research into water extraction from the hyperspectral image gradually attracted more attention. However, delineating water from the hyperspectral image is challenging due to shadows of buildings and trees, and other dark pixels. Furthermore, water indexes that come from multispectral imagery are directly applied to hyperspectral imagery leads to some defects, such as underutilized abundant information and higher misclassifications. Here, we proposed a decision tree water index (DTWI) method to extract water bodies from hyperspectral imagery. Combining the threshold of the near-infrared reflectance and the magnitude of the difference between reflectance of 700 and 730 nm, the DTWI method could effectively extract water from hyperspectral imagery. Three hyperspectral images derived from the study areas within the city, a mixture of city and suburban zone, and suburban zone only, were applied to test the feasibility and flexibility of this method in various sensors and environments. Four alternative methods were used to verify the robustness of the proposed method, including hyperspectral difference water index, normalized difference water index, shaded building index, and support vector machine (SVM). Our results indicated that the proposed method was better suited to extracting water bodies from hyperspectral imagery when compared with other methods and could effectively capture water on different hyperspectral sensors, various heterogeneous surroundings, and multi-spatial scales. The proposed DTWI method has high stability in threshold and band selection, which is comparable to the SVM method in the mean overall accuracy and kappa coefficient (DTWI: 0.98 and 0.81; SVM: 0.98 and 0.82). For the view of elapsed time, calculation, and applicability, the proposed method has the potential of better utilization and rapid response for water extraction in hyperspectral imagery.

Natural ecosystems are highly dynamic and exhibit a highly nonlinear nature. We attempt to detect and analyze changes in a coastal mangrove ecosystem to predict the dynamics of the system and its biodiversity. Multitemporal hyperspectral data have been used to analyze the competition among mangrove and saline blank endmembers and their dominance with time. We aim to predict the ecodynamics of the area through subpixel analysis of multitemporal hyperspectral imagery. The biodiverse coastal zone of Sunderban Biosphere Reserve, West Bengal (a world heritage site), is considered as a case study for predicting the mangrove ecodynamics of the area through Markov chain analysis. The mangrove species of Sunderban vary in their abundance with time due to dynamic weather conditions. The model is applied to hyperspectral data from 2011 and 2014, collected over Henry Island in the Sunderban to predict species dynamics in 2017 and 2020. An endmember transition matrix is framed to determine the endmember dynamics in the area in terms of degradation and regeneration of mangroves and saline blank cover in the area. Based on the transition probability matrix, the abundance values have been predicted for 2017 and 2020. The predicted abundance values have been validated by ground truth values extracted during field visits in 2017. It is observed that, in certain locations, the increase in saline blanks has led to an overall decrease in the proportion of Phoenix paludosa and Ceriops decandra (a salt-intolerant mangrove species) over the years. However, there is an increase in Avicennia marina and Avicennia officinalis, which are salt tolerant and can sustain in extreme saline conditions.

The content of soil heavy metal elements such as copper (Cu) is a crucial indicator for soil environmental monitoring and evaluation. However, traditional field sampling and laboratory chemical analysis methods are time-consuming and labor-intensive. The newly launched ZY1-02D satellite has a large imaging width, nanometer-level spectral resolution, and almost continuous spectral curve, providing a promising means for large-scale rapid estimation of soil heavy metal content. For the first time, our study explores the application potential of ZY1-02D satellite hyperspectral data in estimating soil Cu content in mining areas. A comprehensive approach combining correlation analysis, feature selection, and a regression model is proposed to predict the soil Cu content. First, the correlations between sampling soil Cu content and spectral features (including dozens of spectral indices and different transformed spectra calculated based on the pixel spectrum) were analyzed. Then the random frog algorithm was chosen to reduce the spectral features’ dimensionality and select the proper predictors. Several regression models of soil Cu content, including Gaussian process regression (GPR), boosted trees, bagged trees, and support vector machine, were established using these predictors. The model comparison shows that the GPR model has the best agreement with the sampling data. Furthermore, the accuracy of the GPR model with different numbers of predictors was also analyzed. The results show that the proposed estimation approach based on correlation analysis, random frog algorithm, and GPR model has a good performance compared to sampling data (R²  =  0.83, adjusted R²  =  0.81, RMSE  =  364.61  mg  /  kg, and RPD  =  1.83). And the results further indicate the great potential of AHSI/ZY1-02D hyperspectral data in predicting soil Cu content in mining areas.

Dimensionality reduction (DR) plays an important role in hyperspectral image processing. With enough efficient labeled samples, linear discriminant analysis (LDA) can find the optimal projections by maximizing the between-class scatter variance meanwhile minimizing the within-class scatter variance. However, LDA is not suitable for dimensional reduction of non-Gaussian data and will cause overfitting with insufficient samples. A semisupervised graph embedding and spatial neighbor-based discriminant analysis (SEGSA) method is proposed for DR. First, SEGSA constructs a similarity connected graph among samples without any parameter then uses the labeled samples to filter out valid unlabeled samples and assigns them weak labels. Second, SEGSA utilizes the label and weak label information of the samples to construct the interclass penalty weight graph and intraclass similarity weight graph, which uses different strategies to reduce the effect of weak labels’ inaccuracy. Finally, by mining the spatial neighbor information of samples, SEGSA constructs a spatial similarity graph and maintains the spatial neighbor structure while maximizing the interclass distance and minimizing the intraclass distance. Experimental results on hyperspectral images validate the advantage and effectiveness of the proposed method compared with other DR methods.

Hyperspectral remote sensing is considered an effective tool for monitoring inland water quality. Non-optically active water quality parameters are of great significance to the aquatic environment, although they are rarely used in practical remote sensing applications. This study aims to improve the performance of non-optically active water quality parameter retrieval models by optimizing the wavelength selection and apply to the newly hyperspectral imagery from the Advanced HyperSpectral Imager (AHSI) and Orbita HyperSpectral (OHS) sensors. Focusing on dissolved oxygen, chemical oxygen demand (COD), ammonia nitrogen, and total phosphorus (TP), we propose a hyperspectral dimension reduction method based on the variable importance projection (VIP) and segmented principal component analysis (SPCA) method to determine the sensitive bands of different water quality parameters. A total of 81 in-situ samples of water quality parameters and water spectral reflectance were collected in Shanghai between 2018 and 2019. These were analyzed and used to establish quantitative retrieval models. Furthermore, the principal component regression, partial least squares regression, and back-propagation (BP) network models were compared and partly applied to satellite hyperspectral images. The final results show that models based on VIP-SPCA performed better in the validation set, and the best model was COD estimated by BP (VIP-SPCA) with a coefficient of determination (R²) raised from 0.56 to 0.74. The mean absolute percentage error ranged from 14.23% (COD) to 24.11% (TP). Overall, the AHSI and OHS concentration maps had consistent spatial distributions with monthly monitoring data and reasonable concentration levels. Therefore, the results validate the great potential of hyperspectral remote sensing for inland water quality parameter retrieval using VIP-SPCA.

A hyperspectral image (HSI) contains hundreds of spectral bands, which provide detailed spectral information, thus offering an inherent advantage in classification. The successful launch of the Gaofen-5 and ZY-1 02D hyperspectral satellites has promoted the need for large-scale geological applications, such as mineral and lithological mapping (LM). In recent years, following the success of computer vision, deep learning methods have shown their advantage in solving the problem of hyperspectral classification. However, the combination of deep learning and HSI to solve the problem of geological mapping is insufficient. We propose a new 3D convolutional autoencoder for LM. A pixel-based and cube-based 3D convolutional neural network architecture is designed to extract spatial–spectral features. Traditional and machine learning methods are employed as competing methods, trained on two real hyperspectral datasets, and evaluated according to the overall accuracy, F1 score, and other metrics. Results indicate that the proposed method can provide convincing results for LM applications on the basis of the hyperspectral data provided by the ZY-1 02D satellite. Compared with traditional methods, the combination of deep learning and hyperspectral can provide more efficient and highly accurate results. The proposed method has better robustness than supervised learning methods and shows great promise under small sample conditions. As far as we know, this work is the first attempt to apply unsupervised spatial–spectral feature learning technology in LM applications, which is of great significance for large-scale applications.

To address the problem of “pseudochange” caused by illumination, phasing, and shadows in multiperiod remote sensing images, a bitemporal, hyperspectral remote sensing image change detection method is proposed. The method uses a pyramid self-attention mechanism, and the attention module is introduced to simulate the attention mechanism of human eyes, where more attention is given to a small number of important objects. The model architecture uses a general encoder–decoder paradigm, in which shared encoders extract common features, whereas individual decoders learn task-specific representations. The classifier takes fused data to locate where the changes occurred. The method is tested on the ZY1-CD dataset.

Convolutional neural networks (CNNs) have shown tremendous success for hyperspectral image classification in recent years. CNNs are capable of capturing multi-scale spectral–spatial characteristics of hyperspectral image pixels leading to good classification results. Despite the good accuracy, most of the classifiers misclassify some pixels and generate noisy classification maps. A deep CNN and Markov random field (MRF)-based two-stage classification framework is developed for hyperspectral images. The input image is first classified with the help of a deep CNN classifier. The results provided by CNN are further refined by applying stochastic relaxation labeling using MRF on the first-stage classification map to produce a refined classification map with better accuracy. This two-stage classification approach is particularly helpful if smaller misclassified regions are generated during the first-stage classification. Experiments are performed on one satellite-borne and three airborne hyperspectral images: Dioni, Indian Pines, Pavia University, and Salinas. The results show that the proposed method yields good classification accuracy and smoothed classification maps. The refinement by MRF relaxation improved the overall classification accuracy of the first-stage classifier by more than 2% for all the images. The overall classification accuracy in terms of κ coefficient is obtained as 0.9844, 0.9678, 0.9843, and 0.9841 for Dioni, Indian Pines, Pavia University, and Salinas images, respectively, which is comparable or better than several existing methods.

Soil is one of the essential natural resources that is at risk from heavy metal pollution. The traditional sampling method for soil heavy metal monitoring and assessment cannot meet the requirements for large-scale areas. The purpose of this study was to estimate the soil heavy metal concentrations based on Gaofen 5 (GF5) satellite hyperspectral imagery for the assessment of the heavy metal pollution in the study area and to analyze the scale effect under different resolutions. A total of 96 topsoil samples were collected in this work, and these samples were analyzed for the arsenic (As), cadmium (Cd), chromium (Cr), copper (Cu), nickel (Ni), lead (Pb), and zinc (Zn) contents. To solve the problem of the insignificant features caused by the complex imaging conditions of spaceborne hyperspectral satellite imagery, the binary weight symbiotic organisms search algorithm (BWSOS) was developed. After feature selection based on the BWSOS method, the heavy metal contents are inverted by the use of support vector machine regression. The experimental results show that the BWSOS feature selection method shows a good performance, with the Rp2 values for As, Cd, Cr, Cu, Ni, Pb, and Zn being 0.67, 0.68, 0.73, 0.71, 0.66, 0.65, and 0.71, respectively. Based on the estimated heavy metal concentration maps, the geoaccumulation index (I_geo), the pollution index, and the potential ecological risk index were calculated to assess the heavy metal pollution status in the study area. The results showed that only As contamination is present at a significant level, but with a low level of potential risk for the whole study area. A comparison with the results obtained using HyMap airborne hyperspectral imagery showed that the GF5 satellite hyperspectral imagery can obtain consistent results for heavy metal pollution assessment. The airborne hyperspectral imagery can provide more fine details, whereas the spaceborne hyperspectral imagery is more suitable for large-scale pollution assessment at a low cost.

Grain sorghum (Sorghum bicolor [L]. Moench) is a crucial crop to the world’s semiarid regions, as it can produce grain and biomass yields in precipitation-limited environments. Many genotypes have a characterized form of drought resistance known as the stay-green (SG) trait, enabling sorghum plants to resist postflowering drought stress that can severely reduce yields. Breeding for SG sorghum lines is considered vital for sorghum breeders around the world, but selecting for SG traits currently relies on methods that are labor-intensive and time-consuming. Using unmanned aerial systems capable of capturing high-resolution imagery offers a solution for reducing the time and energy required to select for these traits. A field study was conducted in Manhattan, Kansas, where 20 Pioneer^® sorghum hybrids were planted in a randomized complete block design with three replications per hybrid. Imagery was collected with a DJI^® Matrice 200™ equipped with a MicaSense^® RedEdge-MX™ multispectral camera. Flight altitude was 30 m, and flights were collected under clear, sunny skies within ±2.5  h of solar noon. Ground-measured data included visual senescence ratings, fresh and dry plant biomass, leaf area index, and final grain yield. After correlation and regression analysis, results indicated significant relationships with the near-infrared spectral band with fresh and dry plant biomass samples, the green normalized difference vegetation index scores at flowering were the most related to final grain yield, and the visible atmospherically resistant index was the most related to visual senescence scores. Significant spectral band/vegetative indices were clustered into groups, and significant differences were found between various traits. We have developed a methodology for SG sorghum growers to collect, process, and extract data for more efficient identification of traits of interest.

Ship detection in synthetic aperture radar (SAR) images plays an important role in remote sensing, but it is still full of challenges in the deep learning area. The primary problem is that ships in SAR images have different sizes and orientations. The off-the-shelf detectors are not able to adapt to the situation. Recurrent feature pyramid networks are presented to detect ships with different sizes especially the small ones. Rotatable region proposal network is used for locating ships with a tighter rectangle. Rotatable anchors with sizes, aspect ratios, and angles are designed according to the distribution of ships in dataset. Multiratio region-of-interest pooling is used for projecting arbitrary-oriented proposals to fixed length vectors. Angle-related intersection-of-unit (ArIoU) is used for evaluating the intersection of rotatable proposals. ArIoU can be an indicator for nonmaximum suppression (NMS) and also is used for preparing negative and positive proposals. A loss function is proposed to compute loss between bounding boxes. The sinusoidal function is used for solving the problem of unstable angle. We also use a dataset called SSDD+ (SAR ship detection dataset plus) to evaluate different methods. Experiments based on SSDD+ show that our method achieves state-of-the-art performance. The dataset and the code will be public at https://zhuanlan.zhihu.com/p/143794468.

Due to the rapid developments of smart cities, 3D modeling of urban infrastructures has become essential for making decisions in urban management and planning applications. 3D modeling of the bridge structure, as one of the most important civil structures in transportations, includes many challenges due to geometrical and structural complexities. A knowledge-based approach based on a 2D projection method is proposed for automatic 3D reconstruction of the main bridge elements including the railing, body, piers, and abutment using dense point clouds of unmanned aerial vehicle images. The knowledge is the geometric relations between the bridge elements to model different types of bridge structures. In the first step, the input point cloud is segmented into the major elements using a fuzzy c-means clustering algorithm. Next, the 2D model of each element is generated using a 2D projection-based reconstruction technique. Then 3D models of elements are created by extruding 2D models into the 3D space. Finally, an integrated 3D model of the bridge structure is formed by merging individual 3D models in a CAD format. In order to evaluate the performance of the proposed modeling framework, the length differences of the elements are compared with a reference model. In addition, the geometric distances between the final models and the input point cloud are calculated. Accordingly, a median error of about 3 cm between the length of the elements in the reference and reconstructed models and 3.8 cm between the reconstructed models and corresponding point clouds highlight the effectiveness of the proposed method to generate 3D models of different bridges.

The Brazilian Savanna, also known as Cerrado, is considered a global hotspot for biodiversity conservation. The detailed mapping of vegetation types, called physiognomies, is still a challenge due to their high spectral similarity and spatial variability. There are three major ecosystem groups (forest, savanna, and grassland), which can be hierarchically subdivided into 25 detailed physiognomies, according to a well-known classification system. We used an adapted U-net architecture to process a WorldView-2 image with 2-m spatial resolution to hierarchically classify the physiognomies of a Cerrado protected area based on deep learning techniques. Several spectral channels were tested as input datasets to classify the three major ecosystem groups (first level of classification). The dataset composed of RGB bands plus 2-band enhanced vegetation index (EVI2) achieved the best performance and was used to perform the hierarchical classification. In the first level of classification, the overall accuracy was 92.8%. On the other hand, for the savanna and grassland detailed physiognomies (second level of classification), 86.1% and 85.0% were reached, respectively. As the first work that intended to classify Cerrado physiognomies in this level of detail using deep learning, our accuracy rates outperformed others that applied traditional machine learning algorithms for this task.

The research aimed to evaluate the spectral behavior of remotely piloted aircraft system (RPAS) images taken of agricultural crops during the day to answer the following hypotheses: (1) there are no significant differences between spectral reflectance information obtained through RPAS images and data from spectroradiometry; (2) the survey time can be defined as a limiting factor in obtaining accurate reflectance data; and (3) it is possible to characterize homogeneous observation windows throughout the day. With regular intervals of 1 h, 12 surveys were made in rice and soybean cultivars, using a Sequoia^® camera, embedded in RPAS, and the FieldSpec^®3 spectroradiometer, to validate the image spectra. Normalized Difference Vegetation Index (NDVI), Optimized Soil Adjusted Vegetation Index (OSAVI), and Transformed Chlorophyll Absorption Reflectance Index (TCARI) derived from aerial images were subjected to Kruskal–Wallis and hierarchical cluster tests to define homogeneous observation times. The Sequoia^® system made it possible to obtain reflectance data that met the validation criteria, even operating under varying interaction factors during the day, so the flight schedule did not result in loss of information about the vegetation. The hierarchical cluster analysis allowed the detection of temporal patterns of NDVI, OSAVI, and TCARI throughout the day. For rice and soybeans, three more homogeneous observation windows were characterized. In terms of temporal pattern of NDVI, OSAVI, and TCARI, the Kruskal–Wallis analysis was quite sensitive to the inherent variation of the crops throughout the day, so a hypothesis of homogeneous monitoring windows input throughout the day could not be proven.

Hyperspectral image (HSI) classification is major and necessary task related to HSI analysis in the field of remote sensing. The fundamental steps in this task are band selection (BS), spatial feature extraction, and classification. HSIs generally equipped with rich spectral and spatial information and having the properties such as non-stationary and non-Gaussian. To process such rich information, wavelet transform (WT) is the perfect candidate. Also, the multiscale system in wavelets will be used in complete information extraction. So, classification task is implemented by the application of WT in BS and spatial feature extraction. Here, BS is achieved by the combination of clustering based on key band identification and ranking using wavelet entropy (WE). Discrete wavelet transform is applied along three dimensions to extract spatial features. The extracted spectral and spatial features are used in final classification by convolution neural networks classifier. From the obtained results it is observed that, with the advantage of using WT, the proposed method has successfully addressed overfitting and huge data dimensionality problems. Comparison and detailed analysis of the class-wise accuracies also clearly shows the impact of using WT in HSI classification. Evaluation results on three publicly used datasets namely Indian Pines, University of Pavia, and Salinas shows the significant performance over state-of-the-art methods. The proposed method has attained overall accuracy of 93.85%, 99.05%, and 97.13% for the three datasets, respectively.

Multi-temporal high-resolution land cover (LC) information is of great significance to landscape monitoring, environmental assessment, and local climate change. Given the LC map in the former phase, an automatic LC updating approach based on change detection of high-resolution remote sensing images is proposed. First, object-based change detection is implemented combining spectral bands, normalized difference vegetation index, and normalized difference water index. Second, the changed objects are classified using training samples generated from the unchanged area, and the LC labels of the training samples were transferred from the LC map in the former phase. Finally, as the updated objects with abnormal area (AREA) or perimeter-area ratio (PARA) are recognized as slivers or spurious stretches, and removed using specifically designed rules, an AREA-PARA-based updating method is proposed to update the LC map. Two pairs of GaoFen-1 panchromatic and multispectral sensor images acquired in 2013 and 2015 of two areas in Jiangsu, China, were used to validate the effectiveness of the proposed method. Results and the comparisons with two other updating methods manifested the superiority of the APU method in reducing abnormal LC fragmentation and shape complexity, and maintaining LC consistency between two phases.

Recently, deglaciated landscapes are ideal natural arenas to investigate ecological succession processes. However, ground data acquisition remains complicated as glacier forefields are often difficult to access and fieldwork possibilities remain limited. Remote sensing offers an opportunity to bypass this issue and increase spatial and temporal coverage of ecological parameters. The Landsat satellites (5 to 8) provide reflectance data for the past 40 years, which align with recent phenomena of glacier retreat and related ecological and geomorphological dynamics in glacier forefields. Difficulties remain as information retrieved from 30-m Landsat pixels are the result of a mixture of objects influencing reflectance signals. Here, we used a submeter multispectral unmanned aerial vehicle (UAV) image of the Glacier noir foreland, France, to assess the sensitivity of Landsat normalized difference vegetation index (NDVI) to subpixel vegetation and topographic components. We found a twofold linear relationship (a  =  0.456) and high sensitivity between fractional vegetation cover (FVC) and Landsat NDVI with detection of low vegetation changes (FVC  >  5  %  ) at low NDVI values (<0.1) (F-score  =  0.75). We also showed that vegetation height and subpixel topographic heterogeneity leads to misestimation of vegetation cover as quantified by Landsat NDVI. Overall, our comparative analysis using very-high resolution UAV imagery provides support for the use of widely available Landsat imagery for investigating vegetation dynamics in glacier forefields.

A conventional possibilistic c-mean (PCM)-based algorithm uses means or variance–covariance statistical variables computed from training data sets for classification. Two objectives were covered; in the first objective, the PCM classifier using the training samples as “individual sample as mean” (PCM_IM) in place of “mean” was experimented with handling heterogeneity within the class. The second objective, the PCM-based local convolutional method, was tested with different parameters to handle the noisy pixel. Local convolution algorithms reduce the effect of noise, and PCM_IM handles heterogeneity to enhance the classification performance within the class. The optimized algorithm was first calculated in this research by applying the local convolutional methods for different distance measures and fuzziness factors. This optimized algorithm was obtained by classifying the dense forest, eucalyptus, riverine sand, water, wheat, and grassland classes using the multispectral images of Landsat-8 and Formosat-2 of the Haridwar area. It was observed that the Bray–Curtis distance measures at m  =  2.9 (fuzziness factor) for possibilistic local information c-means using the base PCM classifier yields the highest overall accuracy (OA) (86.37%), irrespective of the classifier. Then, PCM_IM provides a classwise and OA that handles heterogeneity within the class. The classwise accuracy of dense forest, eucalyptus, grassland, sand, water, and wheat is 88.12%, 95.29%, 78.79%, 89.17%, 89.72%, and 92.43%, respectively, and the OA is 85.72%, whereas OA of the PCM classifier using the training sample as mean is 75.18%.

Landslides are a special kind of geological hazard found throughout the entire world. The methods to predict the timing of landsliding and to study its kinematics or consequences are timely topics in research. Terrestrial laser scanning technology has been used for predicting and researching geological hazards for two decades. This research focuses on using terrestrial light detection and ranging (LiDAR) to measure the slow-moving displacement of both the surface and the subsurface soil of landslides and investigating LiDAR system accuracy. Physical modeling was done to simulate the performance of special targets mounted on rods driven into an active landslide, measured using terrestrial LiDAR to improve the scanning accuracy. The primary conclusions from this research are as follows: (1) several tests were done to prove that the terrestrial three-dimensional LiDAR scanner used in this research can precisely obtain the three-dimensional position, displacement, and rotational angle of scanning targets, which can present shallow surface movement of slow-moving landslides. (2) Two kinds of special scanning targets were used for the laser scanner to measure the displacement of surface and subsurface soil in sandbox tests. These consist of spherical Styrofoam targets mounted on rigid metal rods driven into the soil to help the LiDAR scanner improve the accuracy of scanning results and obtain the shallow surface soil movement in the test model. (3) A large-scale field test was designed to prove the feasibility of extending the function of the LiDAR system to obtain the shallow subsurface displacement of landslides and improve the accuracy of the scanning results.

Forest inventory and monitoring is normally carried out based on field measurements of biophysical attributes such as diameter, height, and the number of trees, which is a labor, time, and money consuming activity. Satellite data associated with artificial intelligence tools are alternative approaches to estimate forest parameters at large scales. We assessed correlation of forest variables measured in the field with different vegetation indices (VIs) (normalized difference vegetation index, soil adjusted vegetation index, modified soil adjusted vegetation index, enhanced vegetation index, and enhanced vegetation index adjusted 2.2), retrieved from Sentinel-2 imagery to predict the volume of commercial trees (VCC) showing a minimum commercial diameter   (  MCD  )    ≥  50  cm in a sustainable forest management plan in the Brazilian Amazon region. A total of 150 artificial neural networks (ANNs) of the multilayer perceptron type were trained and supervised. Subsequently, the five best-performing networks were retained based on the fit and accuracy statistics. The ANN-1 showed the best statistical results [root-mean-square error <10  %   and correlation coefficient   (  r  )    >  0.98] to predict the VCC using as input variables the number of trees per hectare showing MCD  ≥  50  cm and all tested VIs. Our study shows promising results that may contribute to improving forest management planning at large scales in remote areas in tropical regions.

Hydrothermal alteration mapping is of great significance for porphyry copper deposit (PCD) exploration. Due to different metallogenetic environments, hydrothermal processes are heterogeneous especially in large-scale ore districts, leading to diverse physicochemical properties and spectral variations of wall-rock alteration. Conventional mapping methods mainly rely on shallow spectral features to identify hydrothermal alteration and cannot fully characterize the properties of the materials. By contrast, deep learning (DL) frameworks with multiple nonlinear layers can extract deep spectral and spatial features from hyperspectral imagery which is more robust and discriminative. Deep models were employed to the GaoFen-5 (GF-5) hyperspectral dataset in the Duolong district, to investigate the role of deep features in hydrothermal alteration mapping. For this purpose, we assess stacked autoencoder and several convolution neural networks which are computationally affordable for achieving large-area mineral exploration. More importantly, the mixed convolutions and covariance pooling algorithm achieved the best mapping results with the overall accuracy of 96.77%. The feature fusion method is also recommended, because of its lightweight structures. These two spectral–spatial classifiers produced appealing classification performance while reducing the misclassification among spectrally similar minerals and alleviating noise, demonstrating their powerful learning capabilities. Finally, spatial patterns of the hydrothermal alteration in the Duolong district can be taken as an important indicator of erosion degree of the porphyry-epithermal system. It is proved that the combination of the DL methods and the GF-5 hyperspectral data is effective in large-area mineral exploration and can be applied to other PCDs in the neighboring Bangong Co-Nujiang metallogenic belt.

Water depth estimation models based on optical satellite images often require ground-truth data for supervised training procedures. However, in the South China Sea (SCS), the ground truth data are limited or outdated. Therefore, it is challenging to derive reasonable water depths around islands or coral reefs without prior knowledge. ICESat-2 is a space-borne LiDAR satellite launched in September 2018 that provides geolocated height at the laser footprint on a global scale and opens an opportunity for bathymetric mapping in regions normally inaccessible. The combination of ICESat-2 data with Sentinel-2 optical images is developed to extend the application of satellite-derived bathymetry (SDB). Three SDB algorithms, including ratio transform (RT), multiple linear regression, and classification-based (CB) algorithms, are applied to investigate the water-depth retrieval capabilities. Five islands located in different parts of the SCS are selected, analyzed, and evaluated with the support of Google Earth Engine. Comparing the ICESat-2 water depth profiles against airborne LiDAR data, the statistical indexes of R² and RMSE reached 0.99 and 0.31 m, respectively. This demonstrates the suitability of using ICESat-2 data as a reliable data source in shallow water. On the other hand, the CB model is used to address the issue of heterogeneity by dividing the target islands into groups based on spectral characteristics. The results show that an integration of ICESat-2 and Sentinel-2 imageries can achieve R² at 0.75 to 0.95 and RMSE at 0.66 to 1.87 m with the deepest pixels at 19 to 32 m across these five islands. We conclude that the ICESat-2/Sentinel-2 synergy scheme is capable of overcoming current limitations in various regions and thus can fill the gaps in bathymetry charts.

Domain adaptation is a proven hyperspectral image (HSI) classification approach aimed at transferring knowledge from a label-rich domain to a label-scarce domain. Existing literature assumes a closed-set scenario in which both the source and target domains share exactly the same label space (“known classes”). However, this assumption may be too ideal in practice. Often, the target domain contains private classes unknown to the source (“unknown classes”). It requires domain adaptation methods to classify the known classes accurately while simultaneously rejecting unknown classes. Focusing on the open-set setting, this paper creatively proposes a hyperspectral open set domain adaptation model based on adversarial learning with a three-dimensional convolutional neural network as the feature extractor, which can sufficiently explore joint spatial-spectral information of HSI and improve classification performance significantly. In addition, this model introduces a dynamic weighting scheme based on multiple auxiliary classifiers for inhibiting negative transfers during adversarial training. Experiment results on three benchmark hyperspectral datasets verify the superiority of the proposed approach for the hyperspectral open set classification. Compared with state-of-the-art techniques with and without using target samples during training, the proposed method improves the mean AUC values by at least 0.157, 0.028, and 0.163 on the Pavia University, Pavia Centre, and Indian Pines datasets, respectively.

The scale-invariant feature transform (SIFT) algorithm is widely used in remote sensing optical image matching because of its robustness and universality. However, owing to the large amount of data and complex and nonlinear change spectral information of multisource and multiscale optical remote sensing images, the matching performance of classic SIFT and related improved algorithms is not ideal. To mitigate this, we propose an SIFT matching algorithm for remote sensing optical images based on entropy classification optimization. After using the classic SIFT to extract features based on the entropy classification standard, the features are classified to reduce the number of invalid features in the matching process and high-precision transformation parameters. Then, the projection model is used to detect stable features to obtain correctly matched pairs, which addresses the problem of minimal matched pairs due to complex and nonlinear changes. We compared it with commonly used algorithms, such as the SIFT, uniformly robust SIFT, and scale normalization and size classification algorithms. The proposed algorithm was successfully applied to multisource and multiscale remote sensing images. Experimental results show that the proposed algorithm performs better than the comparison algorithms in terms of precision, distribution quality, and in matching highly differing images.

We present an approach for segmenting an indoor unstructured point cloud into multiple rooms without additional information. Our proposed approach starts by applying a cloth simulation filter (CSF) to the raw dataset to detect point cloud-related ceiling patches without inverting the point cloud. Next, a grid map analysis is conducted for initial room segmentation. It is updated using a morphological erosion process and a neighborhood filter with an adaptive threshold. Finally, boundary recovery is utilized to correct for any incomplete room boundaries obtained from the previous steps. The capabilities and accuracy of our approach were evaluated on different point cloud datasets, and the average recall and precision were 97.13% and 96.60%, respectively. Further validation with the datasets of different levels of noise and registration errors show that this approach achieves an average recall and precision of 99.24% and 99.90%, respectively.

Object tracking is a technique used in computer vision and image processing applications. This technique fails to track objects when using only visible sensors in situations with illumination variations, occlusion, camouflage, and adverse climatic circumstances. Thermal sensors are fairly illumination invariant. However, in situations where the temperature is not changing much between the foreground and the background, thermal sensors might not provide the best results, as everything might be single color coded. Visible sensors are known to perform better in these conditions. Thus, the joint use of the visible and thermal sensors would be most beneficial in making a robust object tracking system under such challenging conditions. The proposed method performs object tracking by fusing bimodal information using particle filter with structural similarity index metric (SSIM) as a cue as well as a correlation metric for modality selection. The key idea of this method is to adaptively choose the most discriminating modality with respect to the object being tracked. This is performed by calculation of SSIM between reference and successive frames of both the modalities. The method is evaluated by testing on a variety of extremely challenging video sequences in both imaging modalities and has proven to perform better compared to single modality.

To realize inverse synthetic aperture radar imaging with limited bandwidth and short coherent processing interval, in view of pattern-coupled sparse structure of imaging targets, a two-dimensional joint sparse imaging algorithm, namely two-dimensional reweighted alternating direction method of multipliers (2D-RADMM), is proposed. The main idea of the proposed algorithm contains two aspects: one is expanding one-dimensional RADMM (1D-RADMM) to 2D-RADMM by matrix processing, so as to reduce the computation and memory usage. The other is designing different pattern-coupled modes to further improve the imaging performance. Extensive experiments based on two sets of measured data demonstrate that compared with traditional unweighted algorithms, the proposed algorithm has excellent imaging quality under noisy and sparse conditions, and its imaging time is <2  s, which can be suitable for the real-time inverse synthetic aperture radar imaging.

Due to the small number of remote sensing image datasets, it is difficult to train deep neural networks, so we first constructed a two-branch network based on exponentially learning multi-labeled remote sensing image features. In addition, most multi-labeled remote sensing image classification networks use ResNet as the backbone network, which ignores inter-channel correlation, so we used SE-ResNet as the two-branch backbone network. Finally, since most traditional methods focus only on the visual elements in an image or only on the dependencies between multi-labels, we combined the two and constructed a multi-label remote sensing image classification network, Dual-branching Channel Attention and Graph Convolution Network (DCA-GCN), based on a two-branch network and graph convolution, using the two-branch channel attention structure to extract richer image features from remote sensing images and the graph convolution network to establish the dependencies between multi-labels. DCA-GCN achieves relatively excellent results on three publicly available multi-label remote sensing datasets.

Maneuvering target will generate nonlinear migration through range cells (MTRC) and high-order phase term, which make inverse synthetic aperture radar image obtained by range-Doppler algorithm severely defocused in azimuth domain. A rotational motion estimation method is proposed to eliminate the high-order phase term and nonlinear MTRC, which greatly improves the computational efficiency. Frequency axis reversal transform is applied to directly compensate the range migration and Radon-nonuniform fractional Fourier transform is proposed to estimate the rough motion parameter. Then, the search algorithm-based minimizing entropy is used to obtained fine estimation. Based on the two steps, the parameter can be estimated quickly and accurately. Then, resampling and Keystone transform are used to eliminate the high-order phase term and nonlinear MTRC. Finally, focused image can be obtained by performing Fourier transform along the slow-time dimension. Experimental results based on simulated and real data show the effectiveness and computational efficiency of the proposed algorithm.

A complex pattern of urban demographic transition has been taking shape since the onset of the COVID-19 pandemic. The long-standing rural-to-urban route of population migration that has propelled waves of massive urbanization over the decades is increasingly being juxtaposed with a reverse movement, as the pandemic drives urban dwellers to suburban communities. The changing dynamics of the flow of residents to and from urban areas underscore the necessity of comprehensive urban land-use mapping for urban planning/management/assessment. These maps are essential for anticipating the rapidly evolving demands of the urban populace and mitigating the environmental and social consequences of uncontrolled urban expansion. The integration of light detection and ranging (LiDAR) and imagery data provides an opportunity for urban planning projects to take advantage of its complementary geometric and radiometric characteristics, respectively, with a potential increase in urban mapping accuracies. We enhance the color-based segmentation algorithm for object-based classification of multispectral LiDAR point clouds fused with very high-resolution imagery data acquired for a residential urban study area. We propose a multilevel classification using multilayer perceptron neural networks through vectors of geometric and spectral features structured in different classification scenarios. After an investigation of all classification scenarios, the proposed method achieves an overall mapping accuracy exceeding 98%, combining the original and calculated feature vectors and their output space projected by principal components analysis. This combination also eliminates some misclassifications among classes. We used splits of training, validation, and testing subsets and the k-fold cross-validation to quantitatively assess the classification scenarios. The proposed work improves the color-based segmentation algorithm to fit object-based classification applications and examines multiple classification scenarios. The presented scenarios prove superiority in developing urban mapping accuracies. The various feature spaces suggest the best urban mapping applications based on the available characteristics of the obtained data.

Hyperspectral remote sensing imagery contains rich spectral image information, which shows a strong ability to distinguish targets on the ground. Anomaly target detection does not require prior information about the target; this characteristic makes it more convenient to detect the target because prior information is hard to acquire. Therefore, anomaly target detection is widely used in hyperspectral imagery applications. Many anomaly target detection algorithms are proposed by researchers. We propose two models to improve the traditional Kernel Reed-Xiaoli (RX) and Orthogonal RX method. First, Gram–Schmidt Orthogonalization and Householder Transformation are respectively used to construct a data-related basis to approximate the kernel function, and a model based on definite orthogonal features is created. Second, parameters for the models are adjusted to evaluate the efficiency of the algorithms. Finally, experiments conducted with both simulated and real hyperspectral data sets are applied to verify whether the proposed algorithms are effective for hyperspectral anomaly detection. Quantitative evaluation shows that the proposed algorithms are superior to other state-of-the-art algorithms.

The wide application of agricultural greenhouses globally has brought economic benefits; however, it has also led to many environmental problems. The timely and accurate acquisition of greenhouse areas and distribution is valuable for authorities seeking to optimize regional agricultural management and mitigate environmental pollution. Automatic extraction of the greenhouses from high spatial resolution remote sensing (RS) imagery based on deep learning can reduce labor costs and improve operational efficiency to have better application prospects. In this paper, we propose a multi-channel fused fully convolutional network (FCN) optimized by the optimal scale object-oriented segmentation results for agricultural greenhouse extraction from high spatial resolution RS imagery. First, to make full use of remote sensing feature images of target objects and to not increase the complexity of the deep learning network, we constructed a decision-level fusion FCN network that can simultaneously input multiple remote sensing images for preliminary extraction of greenhouse. Second, to address a defect in the classical FCN network causing the easy loss of ground object details, we optimized the preliminary extraction results from FCN by the results of object-oriented segmentation. Finally, the optimized greenhouse extraction results were processed by the mathematical morphology, and the final extraction results were obtained. The experimental results demonstrate that: (1) Multi-channel fused FCN model could use the unique spectral characteristics of different ground objects. (2) Optimized initial extraction results from FCN based on the optimal scale object-oriented segmentation results could fully maintain the edge details of the greenhouse. Experimental results show that the proposed method can extract the greenhouse effectively. The precision and F value of our proposed method are 92.68% and 0.94.

As one of the eight wonders in the world, the virtual restoration of Terracotta Warriors is of great significance to archaeology. However, some parts of fragments were corroded for thousands of years, resulting in the existence of several holes in most of the restored cultural heritage artifacts. Based on the structural and textural information, we present a framework for filling the hole. First, a method based on the Poisson equation was employed to fill the hole on the triangular mesh model. Then, to complete the surface color and texture information of the three-dimension (3D) model, make the hole patch, and the original model surface texture natural transition, the 3D problem is converted into two-dimension (2D) image inpaint problem, and a refined network is added into EdgeConnect to generate a higher resolution result. A set of experiments is performed to evaluate the performance of our proposed framework. We hope the proposed framework can provide a useful tool to guide the virtual restoration of other cultural heritage artifacts.

Forward-looking scanning radar is capable of obtaining the real beam image of terrain in front of the flight platform, and can be used in military and civilian fields, such as sea surface surveillance, search and rescue, etc. However, it is difficult to extract multiple targets from the real beam image due to its poor azimuth resolution. A multiple-target extraction scheme based on low-rank and sparse matrix decomposition is proposed for forward-looking scanning radar. In the scheme, an image with good azimuth resolution is obtained first by the deconvolution preprocessing, and then it is used to map a patch-image. Second, using the low-rank property of the patch-image and the sparse property of the targets of interest, the target extraction is converted into an optimization problem of low-rank and sparse matrix decomposition. Third, the extraction results are obtained by solving this optimization problem using the alternating direction method of multipliers. Finally, simulation and experiment results are given to verify the effectiveness of the proposed scheme.

Serious range-azimuth coupling and spatial variability of imaging parameters limit the focusing depth of conventional synthetic aperture radar (SAR) imaging methods for maneuvering high-squint frequency modulated continuous wave -SAR. To expand the effective imaging area, an imaging method based on time-frequency-domain perturbation is proposed. First, the slant range model based on data recording parameters is optimized to remove Doppler frequency shift and secondary coupling caused by in-pulse movement. Then, the spatial variability of range cell migration (RCM) is corrected by constructing the time-domain perturbation function to achieve unified RCM correction. In azimuth processing, the linear spatial variability of the first- to third-order Doppler parameters and the second-order space variation of azimuth modulation frequency are removed by the time-frequency perturbation filtering functions. The focus quality of the nonscene center is effectively improved. Finally, the effectiveness of the algorithm is verified by both the simulation data and the measured data.

Research on sea clutter modeling is meaningful for the sea clutter jamming rejection in radar detection. The previous methods cannot fit the characteristics such as the peak value and amplitude width of the sea clutter amplitude distribution curves. Thus, an estimation method of sea clutter based on the multi-feature-point model validation is proposed. First, the amplitude distribution characteristics and temporal correlation of sea clutter are analyzed, and the spherically invariant random process is used to simulate the K-distribution sea clutter model. Then, six feature points are constructed, (i.e., the maximum probability density, the amplitude value of maximum probability density, the 3 dB amplitude width, the amplitude widths corresponding to 1/3, and 2/3 of the maximum probability density, and the amplitude critical value corresponding to probability density lower than 0.01). Based on the multigroup feature points of the amplitude distribution curves, the radial basis function (RBF) neural network is trained to derive the relationship between the multiple features and the shape parameter and predict the shape parameter of the measured data. Finally, the simulation results show that the proposed method can fit the characteristics of the actual sea clutter amplitude distribution more accurately than the previous estimation methods.

Recently, the deep learning (DL)-based solution for single image super-resolution reconstruction has been extensively used. Though these methods have shown exceptional results, the computational requirement is very high and requires high-end graphic processors. Motivated to establish an efficient method without such requirement, we propose a modified Markov random field (MRF) model and two-dimensional (2D) phase congruency-based single-image super-resolution reconstruction method for remote sensing images. The 2D phase congruency-based feature extraction method is used to compute features such as edge and texture maps to achieve higher accuracy in finding similar example patches in feature space. To address an essential aspect of successful texture reconstruction, we have incorporated a texture prior for the computation of joint probability in our work. Image Euclidean distance (Ieuc) is integrated to achieve higher accuracy in finding the similarity between image patches in feature space and modeling compatibility functions. The experimental results demonstrate that the results of the proposed method are at par with the DL-based methods and outperform other state-of-the-art methods in perceptual quality as well as peak signal-to-noise ratio and structural similarity index parameters. Moreover, it shows significant improvement in the texture regions while reconstructing sharper edges for ×4 upscaling.

In this research, a new image fusion technique was developed to merge the spatial details in the panchromatic (PAN) image and the spectral details in the multispectral (MS) image to produce fused images with the highest spatial and spectral qualities. This fusion technique is superior to the other existing fusion techniques in that it can produce sharp images with no color distortion for all types of land cover. This is because of using weighting parameters for all MS bands in the calculation of the I component according to the intersecting areas between the MS bands and the PAN band, and the I component will be an average of the four modified MS bands, which minimizes the gray level differences between PAN and I images. Moreover, using the variable vegetation coefficient β for the NIR band makes this technique produce the highest spectral and spatial qualities for each land cover class. To evaluate the performance of the ELSHORA fusion technique, it was compared to the traditional intensity-hue-saturation (IHS) fusion technique, the fast IHS (FIHS) fusion technique, and four other standard image fusion techniques; principal component analysis, hyper-spherical color space, multiplicative resolution merge and Brovey transform. These fusion techniques were applied to three sets of GeoEye-1 PAN and MS images covering different land covers in Tanta city, Egypt. The fused images are compared statistically and visually to the original PAN and MS images. The results showed that the ELSHORA fusion technique was chosen as the best method for all datasets. This is because it has produced the highest spectral quality in which no color distortion can be detected and has provided significant spatial quality, visually and analytically. To test the spatial and temporal transferability of ELSHORA fusion technique, it was compared to the original FIHS fusion technique to fuse three different datasets covering different classes of land cover (urban, agriculture, and mixed) located in different places and acquired at different times. The results indicate that ELSHORA fusion technique is spectrally and spatially in the lead for all datasets. This study proved the efficiency of ELSHORA fusion technique to produce images of high spectral and spatial resolution at different times and places.

To verify the necessity of the time synchronization mechanism in a multibaseline synthetic aperture radar (SAR) system and to acquire the allowable magnitude of the residual time synchronization error after compensation, the influence of the time synchronization error on the altimetry accuracy of multibaseline SAR is studied. Based on a time synchronization error model, the influence of the time synchronization error on the interferometric phase and the influence of the interferometric phase error on the altimetry accuracy of the multibaseline SAR are analyzed, and accordingly, an error transfer model from the time synchronization error to the altimetry accuracy of the multibaseline SAR is established. According to the established error transfer model, the necessity of the time synchronization mechanism in the multibaseline SAR system is verified, and a formula of the allowable magnitude of the residual time synchronization error is derived. The formulas of the single-baseline SAR and the multibaseline SAR to evaluate the altimetry accuracy are compared, and the condition that the altimetry accuracy of the multibaseline SAR is higher than that of the single-baseline SAR is clarified. The accuracy of the derived formula is verified by the performed simulation.

Nowadays, the low-rank representation (LRR) and deep learning-based methods have received much attention in anomaly detection for hyperspectral images (HSIs). However, most of these methods mainly focus on the powerful reconstruction capability of the neural networks while ignoring the potential probability distribution of both anomalies and background pixels. To solve the problem, we propose a sparse component extraction-based probability distribution representation detector (SC-PDRD) framework, which integrates the characteristic of the sparse component obtained by the LRR model with the powerful probability representation ability of the variational autoencoder (VAE) network. The LRR model effectively separates the anomaly component from the background, which also serves as the prior anomaly distribution for each pixel. Moreover, the VAE architecture tries to recover the potential anomaly distribution using the sparse detection map in the feature space. In addition, we employ the Chebyshev neighborhood to leverage spatial information. The modified Wasserstein distance measures the distance between the test pixel and its neighborhood. The final detection map is attained by combining the prior anomalous degree of the anomalies with the output of the VAE network. Experimental results on three real HSIs demonstrate the effectiveness and superiority of SC-PDRD.

SAR (synthetic aperture radar) and PolSAR (polarimetric synthetic aperture radar) images fulfill a fundamental role in Earth observation, due to their advantages over optical images. However, the presence of speckle noise hinders their automatic interpretation and unsupervised use, rendering traditional segmentation tools ineffective. For this reason, advanced image segmentation models are sought to overcome the limitations that make an adequate treatment of speckled images difficult. We propose a procedure for SAR and PolSAR image classification, based on texture descriptors, that combines fractal dimension, a specific probability distribution function, Tsallis entropy, and the entropic index. A vector of local texture features is built using a set of reference regions, then a support vector machine classifier is applied. The proposed algorithm is tested with synthetic and actual monopolarimetric and polarimetric SAR imagery, exhibiting visually remarkable and robust results in coincidence with quantitative quality metrics as accuracy and F1-score.

In remote sensing images, the complex and changeable background often interferes with the detection of cloud and its shadow, which leads to the phenomenon of missing detection in areas with similar background colors. In addition, most methods cannot achieve the real-time detection effect because of too much calculation. To solve the above problems, the parallel asymmetric network with double attention is proposed. The algorithm adopts a parallel method, which allows two branches to participate in the calculation at the same time. One branch is used to fuse different information from two branches at different levels, and another branch is responsible for extracting deeper context information. At the same time, double attention module and asymmetric dilated block are used for two branches, respectively. Double attention module can help the algorithm pay more attention to the category and spatial information of cloud and its shadow, thus reducing the interference caused by background information in images. Asymmetric dilated block can extract two levels of receptive fields, and it can help the network to obtain enough receptive fields in cloud and its shadow images, thus reducing the cases of missed detection and false detection. Moreover, these two modules are lightweight in their parameters and calculation. Compared with some previous methods, our method can guarantee the accuracy and the detection speed is fast.

The issue of limited labeled samples is still grave in hyperspectral image (HSI) classification. Collaborative learning promotes a solution to this issue by combining active learning (AL) and semisupervised learning. However, it has been found that the performance of wrong pseudolabels added into the iteration may seriously deteriorate classification performance. To tackle this problem, we propose a reverification of pseudolabels algorithm based on superpixels segmentation, which is tripartite, including two AL selection strategies, the first-verification procedure based on three classifiers, and reverification procedure improving the correctness of pseudolabeled samples based on superpixels segmentation. Specifically, two AL strategies mainly generate training samples sets for two check classifiers, and three classifiers are main forces to implement the first-verification for unlabeled samples with predictive labels. Subsequently, a reverification procedure based on local similarity in superpixels is applied to reverify these unlabeled samples with predictive labels passing the first-verification procedure. The proposed algorithm is tested on three widely used HSI datasets and compared with three state-of-the-art collaborative learning algorithms with onefold verification. Experimental results illustrate numerically and visually the significantly superior performance of our proposed algorithm considering the spatial information to reverify the correctness of pseudolabels to unlabeled samples.

X-band dual-polarization weather radar parameters characterized by clear-air turbulence and precipitation over Mount Everest are introduced. As the X-band radar wavelength is short, data quality control is exercised over the obtained differential phase (Φ_dp), reflectivity (Z_h), differential reflectivity (Z_dr), and other polarization physical quantities. Based on the X-band dual-polarization weather radar, FY4 satellite, and EC reanalysis data, the 3D structural characteristics of clear-air turbulence, weak precipitation, and meso-microscale strong convective precipitation on the northern slope of Mount Everest from June 2019 to July 2019 are analyzed. Our results reveal that clear-air turbulence on Mount Everest is mainly wet turbulence, and Z_h and Z_dr are significantly greater than those in the plain area. The ground clutter in some locations exhibits low Z_dr and high ρ_hv characteristics, opposite to those in the plain area. Weak precipitation on Mount Everest has a typical three-layer structure similar to that in the plain area, and when Z_h  >  25  dBZ, Z_h and Z_dr display a positive linear correlation. Alternatively, Mount Everest’s atmospheric environment is conducive to triggering strong convective weather, consisting of common isolated convection cells at the microscale. The echo top height and maximum precipitation intensity of strong convection weather on Mount Everest are greater than those in the plain area. However, the whole layer convective thickness is much smaller than that in the plain area. Our preliminary results reveal the raindrop size distribution, quantitative precipitation estimation, and microphysical parameterization scheme of cloud precipitation in the Mount Everest region.

Wet path delay (WPD) is an important source of satellite altimeters’ transmission error, which must be fully corrected to ensure data quality. After decades of study, WPD can now be accurately negated by measuring the brightness temperature with a microwave radiometer (MWR). However, due to hardware and financial limitations, many satellites are not equipped with MWRs, and WPD can only be corrected by mathematical models. Currently available empirical models, which are mostly established by traditional fitting methods, are unable to reach the accuracy required by altimeters, so it is necessary to further improve the correction accuracy of the models. Based on three machine learning algorithms, we established three high-precision wet tropospheric correction (WTC) models suitable for satellite altimeters. We analyzed the global navigation satellite system dataset and three MWR datasets, i.e., Jason-3, Sentinel-3A, and Sentinel-3B, for data quality and consistency and combined these datasets for modeling. In addition, several input features were extracted from existing knowledge on tropospheric delay. Through a comparative analysis of different feature combinations, temperature (T), water vapor pressure (e), and total content of water vapor were selected as the input features. Then we established three WTC models based on multilayer perceptron, random forest, and the adaptive network-based fuzzy inference system (ANFIS), respectively. With a dataset from Jason-2 that was not used during modeling as the verification set, we verified the generalizability of these WTC models from the global, latitudinal, and seasonal perspectives. Moreover, three traditional WTC models were compared with the ones developed here. The results show that the WTC models established based on machine learning algorithms outperform traditional WTC models. The WTC model based on ANFIS boasts the best calculation accuracy in all aspects, over 50% higher than that of traditional WTC models.

The Advanced Baseline Imager (ABI) is the primary instrument onboard NOAA’s current generation of Geostationary Operational Environment Satellites R-series (GOES-R) satellites, measuring the reflected and emitted energy from the Earth. It consists of 16 channels, 10 in the thermal infrared (IR) and 6 in the solar reflective spectrum. Being the first in the GOES-R series satellites, GOES-16 was launched on November 19, 2016, and became operational as GOES-East at 75.2°W since December 18, 2017. We examine the radiometric calibration accuracy and stability of GOES-16 ABI IR radiance since its first light in January 2017, including the effects of two major updates of the GOES-R Ground Segment processing for the IR channels in October 2017, and June 2018. Using measurements by multiple hyperspectral radiometers from low Earth orbit satellites as references, it is found that, when converted to scene brightness temperature of 300 K, the calibrated ABI IR radiance is accurate within 0.13 K for Ch16 (13.3  μm), within 0.06 K for channel 12 (9.6  μm), and within 0.05 K for the other IR channels. This is an order of magnitude better than the requirement of 1 K. Since June, 2018, the radiometric calibration of GOES-16 ABI IR channels has been temporally stable, spatially uniform within the ABI full-disk field of view, absent of diurnal and seasonal variations, and invariant within various timelines. Other than short term disruptions as noted in the Calibration Event Log, GOES-16 ABI IR Level 1b products since June 19, 2018, is a reliable reference for satellite intercomparison or intercalibration studies.

The glacier snowline can be used as an indicator of a glacier’s equilibrium line, which is a pivotal parameter for studying the effect of climate change on glaciers. However, the relationship between snowline altitudes (SLAs) and climatic regime, as well as the comparison between different glacier types, has received less attention. Using Google Earth Engine, we first developed an automated algorithm that employs the Otsu thresholding method to distinguish snow-covered areas from clean ice on a near-infrared band of Landsat imagery available from 1995 to 2016 and further to delineate glacier SLAs in the three regions of the eastern Tibetan Plateau (TP). The three study regions, Sepu Kangri (maritime), Bu’Gyai Kangri (continental), and western Qiajajima (continental), in the eastern TP have different climate regimes and are on a latitudinal transect from south to north. We then investigated the impacts of climatic factors on the SLA and its variability over the period studied. The results over the eastern TP indicate that (1) the SLA increased by 94, 55, and 49 m from south to north during the 22-year period, with the SLA variations of maritime glaciers being the most pronounced; (2) the southern maritime glaciers were mainly affected by precipitation, whereas the northern continental glaciers were influenced by temperature. Owing to the difference in primary climatic factors affecting snowlines, continental glaciers were found to have higher SLAs on the south slope, whereas maritime glaciers had higher SLAs on the north slope.

As one of the fastest growing economies in the world, the Great Mekong Subregion (GMS) has experienced dramatic changes in impervious surface areas (ISA) during the last few decades, which have a profound impact on human society and ecosystems. To quantify the impacts of rapid urbanization, it is essential to examine spatial distribution and dynamic change of ISA in the GMS. Thus we the applied object-based image analysis (OBIA) method to map ISA in the GMS from 1992 to 2019 using nighttime light data and Landsat images. Meanwhile, spatiotemporal changes of ISA across the GMS during the study period were analyzed. Our results show that (1) the OBIA algorithm was effective for mapping ISA in the GMS, with the overall accuracy reaching 92.28%. (2) The ISA in the GMS reached 17671.47  km² in 2019, which is 4.36 times more than that in 1992. (3) The increase of ISA in the GMS mainly occurred in 10 metropolitan cities with the highest population densities, accounting for ∼93  %   of the GMS’s total in 2019. There are different possible major driving factors for each city and member state. Monitoring the spatiotemporal changes of ISA in rapidly urbanized areas is crucial for promoting sustainable urban development and reducing the impact of urbanization.

The observation of polar clouds is of great significance to studying climate change in Antarctica. The modified Klett–Fernald method and Mie scattering light detection and ranging (Lidar) were used to study the characteristics of low-altitude clouds from February 15 to March 2, 2017 at the Great Wall Station. The temporal and spatial variations in the cloud extinction coefficient were obtained, and the optical and physical characteristics of the low-altitude clouds were statistically analyzed. The analysis shows that the near-ground atmosphere is very clean and the atmosphere at low heights is covered by low-altitude clouds. The meteorological and Lidar data show that the polar cyclone may be the main driving force of the cloud-height changes at this station.