This is a list of reviewers who served the Journal of Applied Remote Sensing in 2021.

Guest editors Shuisen Chen, Chandrasekar Nainarpandian, and Ayad M. Fadhil Al-Quraishi summarize the Special Section on Coastal Zone Remote Sensing for Environmental Sustainability.

Using satellite remote sensing technology to observe sea surface wind fields has become an important instrument. The successful launch of Tiangong-2, the future Surface Water and Ocean Topography (SWOT) mission, and demonstrating GaoFen Project provides the future understanding for the generation of new wide-swath imaging altimeter. The constructed data sets and validated data sets of the model function were obtained by combining the observable conditions of imaging altimeter with the theoretical model of sea surface backscattering coefficient. A model function for wind speed retrieval at incidence angles from 1 deg to 7 deg and from 7 deg to 11 deg was proposed, which is suitable for wide-swath imaging altimeter, model coefficients were calculated using the constructed data sets, model functions were validated using the validated data sets. The validated results of model function show that the root-mean-squared error (RMSE) ranged from 0.005 to 0.72  m  /  s. Wind speed was retrieved with China France Oceanography Satellite (CFOSAT) Surface Waves Investigation and Monitoring instrument (SWIM) L2 data for incidence angles from 3 deg to 10 deg with 0.5 interval, and wind speed was validated with European Centre for Medium-Range Weather Forecasts (ECMWF) data. The results of wind speed validation show that the average of RMSE is 1.6  m  /  s. Therefore, the wind speed retrieval was validated with the wind speed from ECMWF, which showed good coincidence.

Ship detection and segmentation in optical remote sensing (ORS) images is a current concern actively addressed by the academic community owing to its vast applications. Most present methods only detect the location of the ship but do not segment it at the pixel level. Considering the complex background of ORS images, identifying ships from interferences, such as clouds, waves, and some land architectures similar to ships, proves to be difficult. To address this issue, we propose a high-resolution representation and multistage region-based network (HR-MSRN) for ship detection and segmentation from ORS images. HR-MSRN mainly consists of three parts: the high-resolution feature pyramid network (HRFPN), region proposal network (RPN), and multistage detection and segmentation network (MSDSN). First, HRFPN is built as a backbone network to extract and fuse multilevel image feature maps. Second, ship candidate boxes are generated by defining numerous anchors through RPN. Third, using the idea of a cascade mask R-CNN as the reference method, the MSDSN is proposed to obtain the ship localization and mask shape. We utilize the proposed framework to evaluate an Airbus-ship dataset, and the experiments indicate that (1) HRFPN provides better feature representation ability than the ResNet-FPN when maintaining the same detection framework, especially for small ships; (2) the direct flow between mask branches refines the mask information, and the semantic segmentation branch enhances context information, which indicates that MSDSN is effective and promotes further improvements in ship detection and segmentation from ORS images; (3) in comparison to other region-based methods, HR-MSRN obtains superior performance of ship detection and segmentation in the ORS imagery.

The frequent occurrence of harmful algal blooms (HABs) in China’s coastal areas has caused major losses to the economy and society and has seriously threatened the coastal marine environment. Based on 30 years of relevant statistical data and satellite remote sensing data from 1990 to 2019, we analyze spatial–temporal distribution of HABs in China’s coastal areas. Our results show that 1557 HABs occurred in China’s coastal areas from 1990 to 2019: (1) 961 HABs occurred in offshore China from 2001 to 2012, a period with many HABs, accounting for 62.00% of the HABs of the study period. The numbers of HABs (1998 to 2019) in the Bohai Sea, Yellow Sea, East China Sea, and South China Sea were 198, 133, 813, and 249, respectively. 536 HABs occurred from April to September, accounting for 88.00% of the total occurred from 2009 to 2019. (2) Spatially, based on the kernel density estimation, HABs were mainly concentrated in Bohai Bay, the Yangtze and Pearl River deltas, and other regions with relatively dense populations and relatively developed economies, especially in Zhejiang Province and in the Yangtze River Estuary. Imagery-derived information on the build-up of HABs provides coastal communities with objective information to plan and deal with the adverse environmental and health effects associated with an HAB.

This paper proposes the design of an ecological classification management method based on ecological sensitivity and disturbance of road network construction that reflects the characteristics of an island ecosystem using geographical information system and remote sensing technologies. This method evaluates the ecological impacts during the high-intensity development of the island, classifies different ecological management types, and shows the key areas and management strategies for improving the ecological and environmental quality of the island and implementing ecological restoration. The proposed method is applied to Haitan Island, which has been in a period of high-intensity development since 2012. The results show that (1) the ecological sensitivity evaluation method of the island established in this study fully considers the key ecological issues of an island ecosystem. The evaluation results of highly and moderately sensitive areas cover surface water and its conservation area, woodland, and beach, objectively reflecting the relative spatial heterogeneity of ecological sensitivity within Haitan Island. (2) The distribution of high disturbance areas of road network construction is in line with the overall planning pattern of Pingtan Comprehensive Experimental Zone, which matched with the objectivity reality. The high and moderate disturbance areas of the Haitan Island increased by 6.80 and 27.60  km², respectively, in the last 10 years, and the no disturbance area decreased by 73.70  km². The impact on highly ecologically sensitive areas is little, and the key ecological areas are strictly protected in the high-intensity development activities. (3) The areas of ecological protection priority area, ecological restoration priority area, general control zone, and ecological reconstruction area on Haitan Island are 92.74, 9.27, 113.57, and 34.42  km², respectively. The ecological classification management should be implemented for different areas, and targeted ecological protection and restoration strategies should be adopted, which is conducive to improving the overall ecological environment quality of Haitan Island.

To objectively and justly compare the results of remote-sensing image fusion, evaluate the fusion algorithm, and optimize the fusion process, a quality evaluation index system is constructed by considering both qualitative and quantitative evaluations. The quantitative evaluation indexes are divided into three types based on the analysis of evaluation methods of the fusion quality of primary remote-sensing images. The applicability and effectiveness of six pixel-level fusion methods of optical remote-sensing image fusion were verified based on the constructed quality evaluation index system used to evaluate those methods in offshore and archipelagic areas. The experimental results indicate that the constructed quality evaluation index system evaluates the fusion algorithm objectively, comprehensively, and accurately. The intensity–hue–saturation transformation-based method distorts spectral features of the original image and easily causes spectrum degradation, whereas the wavelet transformation-based method has an advantage in the preservation of spectral features but efficiently produces blockiness and image penumbrae. The principal component analysis transformation-based method preserves more detailed textures and structural characteristics of original images but loses some physical properties. The Brovey transformation-based method results in spectral distortion, the high pass filter-based method shows a clear boundary of surface features after image fusion, and the Gram–Schmidt-based method shows a high ability of spectral information retention, but the latter performs poorly in the preservation of spatial information. The proposed quality evaluation index system of remote-sensing image fusion can be used as the objective effect evaluation criteria for remote-sensing image fusion and offers a reference for future applications of fusion results.

Our study establishes general statistical methods for estimating the accuracy of diachronic low-resolution satellite instruments for shoreline monitoring using the most common satellite datasets such as Landsat and Sentinel-2. The study coast is extended from Cape Araxos to Rio Port, in the Gulf of Patras in Northwestern Peloponnese, Greece. Landsat archival data from 1996, 2008, 2016, and 2018 (30 m spatial resolution) covering the entire study area were used consisting of images acquired by the respective instruments due the time such as the Thematic Mapper (TM), the Enhanced Thematic Mapper Plus (ETM+), and the Operational Land Imager (OLI). Moreover, for the years of 2016 and 2018, Sentinel-2 imagery with 10 m spatial resolution was used. High-resolution datasets of the respective years were used as reference for the calculations. “Generate near table” and “mean center,” two simple and fast statistical tools of the ArcMap v5 toolbox, were applied, and a comparison due to the relative results of shoreline change envelope (SCE) rates was performed. It is the first time that such tools are used to evaluate the accuracy of shorelines derived from low-resolution images. It was proved that the generate near table tool produces very similar results to the SCE tool of the Digital Shoreline Analysis System. There is a strong correlation between the two tools strongly dependent on the dataset being used (Landsat or Sentinel). The analysis of the Sentinel-2 datasets showed that the specific sensor can discriminate the waterline with better accuracy than all Landsat instruments and could be useful to relative studies especially when the procedure follows the subpixel level algorithms. Finally, the results showed that the upgrades of the Landsat instruments from TM to ETM and to OLI did not improve the accuracy of land-water delineation.

More and more human activities have caused varying degrees of interference with the ecosystem. As an effective quantitative means of eco-environmental quality, remote sensing has been widely used. The remote sensing ecological index (RSEI) is the most popular ecological evaluation model at present. However, the model is affected by the direction of the eigenvector, and two completely opposite results will appear in the calculation of the model. In the past, researchers often manually judged whether the two results met the expectations according to a priori knowledge to select the appropriate calculation results. With the development of remote sensing big data, applying this artificial discrimination method to eco-environmental monitoring under the background of big data is difficult. Therefore, we test the evolution of the RSEI model eigenvector in time series through large samples. It is found that the results of the two models are completely opposite in space. For any model, changing the order of input bands may result in RSEI also being the opposite. Through the large sample test in different time periods, the eigenvectors of each ecological factor affect the direction of the corresponding RSEI. Therefore, we propose an improved model to judge the characteristic contribution direction of the first principal component by selecting the wet factor, which is less affected by seasonal changes. The model direction can be automatically modified without the intervention of researchers according to subjective experience. The improved model can adapt different periods of RSEI calculation. At the same time, no matter how the input band order changes, the final result direction is correct. The improved model makes it possible to monitor and calculate the eco-environmental quality of remote sensing big data and provides a solid scientific basis for the development of the model through research on the mechanism of the model.

Groundwater is an important fresh water resource, which is very important for human survival. We analyzed groundwater potential assessment (GPA) by using back propagation (BP) neural network model and analytic hierarchy process (AHP) model in remote sensing and geographic information system in the north of Zhuhai City, Guangdong Province, China. We have used a variety of factors related to groundwater. These models were based on the relationship between groundwater supply potential and hydrogeological factors. GaoFen-6 remote sensing image was first collected, processed and got lithology, relief, slope, flow accumulation (FA), land temperature, soil humidity and vegetation coverage. Water yield data were collected from 36 well locations. We determined the weight of each index by using BP neural network and AHP. The results were fitted with the actual well water yield. The R² of BP neural network and AHP models for GPA were 0.89125 and 0.85946, respectively. The results show that BP neural network model can eliminate the influence of subjective factors on the results. It is BP neural network model that more suitable for the development and utilization of groundwater resources than AHP model.

Ecosystem services are the conditions in which human existence can be maintained and satisfied by natural ecosystems and their constituent species. Meanwhile, island cities have limited natural resources and fragile environments. So, more attention should be paid to the evaluation of their ecosystem service value (ESV). We propose an approach for estimation of ecosystem services value using land-use and land-cover change (LUCC) information obtained by multitemporal remote sensing images based on the method of equivalent value factor. The results in Zhoushan Island, China, from 1984 to 2020 showed that (1) land use and land cover significantly changed, with all LUCC decreasing in general except for the categories of construction land and water bodies. The total area increased by 67.96  km², and an average annual increase rate of 0.37%, which was related to the land reclamation measures taken by people in the Zhoushan Islands. (2) Changes in ESV were significant during the last 37 years. The total annual ESV in the study area decreased from 937.98 million Yuan in 1984 to 899.13 million Yuan in 2020, with a decrease of 4.14%. It is mostly attributable to the 5.51% decrease of forest land, 128.69% growth of water bodies, 28.64% decrease of mudflats, 3.37% decrease of cropland/grassland, and 68.38% decrease of bare land. (3) The sensitivity index of land-use types is <1, which means that the ESV is inelastic to adjusted coefficient of ESV and the research results are reliable. We can provide theoretical methods and a basis for decision-making in support of realizing regional ecological security and the sustainable use of land resources.

Beach placer deposits are economically enriched with Fe-Ti oxide, Zr, and rare earth minerals such as garnet, rutile, zircon, monazite, ilmenite, and sillimanite. Landsat 8-OLI image provides the distinct spectral signature of placer deposits in its multispectral bands that can be used as a key source for mapping placer mineral deposits. GIS-based machine learning algorithms (MLAs) are used for executing the subpixel-based classification of high-dimensional data products (multispectral images, i.e., Landsat 8-OLI image) based on its spectral signatures for mapping the placer mineral deposits. The Landsat 8-OLI images are executed using the four widely used MLAs such as maximum likelihood classification, random forest classifier (RFC), support vector machine, and artificial neural network for mapping placer mineral deposits in the southwest coastal stretches of Thiruvananthapuram district in the Kerala state of India. The resultant map shows the occurrence of placer deposits with a major concentration of ilmenite. The accuracy assessment reveals that RFC shows better performance for mapping beach minerals with an overall accuracy of 84.4925% and kappa coefficient of 0.8184. Our study proves that the MLAs are the vital tools for mapping the placer mineral deposits in the sandy beaches, and thereby, this approach is being used as a scientific key tool for opening the gateways for beach mineral mapping using multispectral satellite images.

SNAP-6 DInSAR analysis of Sentinel-1 SAR images acquired during a 2-month time frame when no significant uplift/subsidence events occurred provides the first known baseline DInSAR data for the Charleston, South Carolina, area. DInSAR measurements during a 72-day time frame had substantial bias throughout the region, with variability a strong function of the level of vegetation present. Specifically, results for (a) urban-built areas show low variability (<0.27  cm) with a bias/offset ranging from −25 to 16 cm, (b) nearby Francis Marion National Forest (FMNF) shows much high variability (1.67 to 10.52 cm) with bias/offset from −44 to 22 cm, and (c) mixed urban-forest and other vegetated areas in the region have variability (1.5 to 2.5 cm) and bias/offset (−15  cm to 15 cm) that fall between these ranges. Taken together, the satellite-based measurements indicate the need for integration of available local GPS and DInSAR measurements, as well as placement of additional persistent scatterer sites in the outlying regions.

Sea surface temperature (SST) is one of the physical parameters of the ocean, which plays an important role in meteorology, navigation, and fishing. SST retrieval from thermal infrared remote sensing has become a research focus because of its advantages of high efficiency and wide observation range. However, more accurate retrieval algorithms and more complete data processing procedures are needed for the satellite data with higher spatial–temporal resolution. High-precision SST retrieval models based on split-window algorithm were established using Gaofen-5 Visual and Infrared Multispectral Imager (VIMS) data, some of which introduced the quadratic term of brightness temperature difference into the models. The highest accuracy is better than 0.15 K. Moreover, we proposed a relatively complete cross-calibration approach to solve the problem of unstable calibration coefficients in bands 11 and 12 of VIMS. Therefore, the accuracy of the models was verified by satellite images. At last, the models were applied to the retrieval of SST in the vicinity of Fuqing Nuclear Power Plant. We expand the practical value of VIMS in SST retrieval and thermal discharge monitoring and provide technical support for the application of high-resolution remote sensing data.

We investigate the potential vulnerability to seawater intrusion into the aquifers along the Kozhikode coastal stretch using the modified GALDIT (GALDIT-U) model from an urbanization perspective. Urban growth impact (expansion of the impervious surface) was added as an additional input parameter in the GALDIT model, and the analyses were performed using GIS. The results indicated that the area of 16.84  km², found within the proximity of urban settlements and eroded shorelines along with the Kozhikode, Beypore, Kadalundi, and Faroke, falls under very high vulnerability to seawater intrusion with vulnerability index values ranging 24.82 to 29.82. In the southern sector, the area of 16.84  km² between Elathur and Beypore also falls within very high vulnerability zones. Similarly, the area of 37.01  km² in Quilandi, Faroke, Puthiyankadi, Panniyankara, and the east sector of Kozhikode municipality is seen as high vulnerability zones with the vulnerability index values ranging 19.84 to 24.82. Seawater intrusion estimated under moderate vulnerability zones is found to occupy 47.85  km² (vulnerability index values 14.88 to 19.84). The low and very low vulnerability zones cover 65.07 and 48.07  km², with the vulnerability index values ranging 9.92 to 14.88 and 4.96 to 9.92, respectively. When cross-validated using major hydrochemical parameters, the results indicate a strong correlation between the vulnerability index classes and groundwater physicochemical parameters. The sensitivity analysis carried out indicated that the distance from the shoreline (D), impacts of urban growth (U), and depth to groundwater table above sea level (L) to be highly influencing parameters for seawater intrusion vulnerability in the coastal stretches. The modified GALDIT-U model shows reliable accuracy for estimating and demarcating seawater intrusion vulnerability zones and could act as an efficient tool in the sustainable management of coastal aquifers in an urban environment.

When developing modern unmanned aerial vehicle-based onboard small-sized radar systems, which are used for monitoring of coastal zones for their operational and environmental purposes, adequate mathematical models are needed of reflections from the corresponding resolution elements of the underlying surfaces, located on the water–land interface. We consider the development of some mathematical models for such applications. We analyze two modern approaches used in developing mathematical models for input signals reflected both separately from the terrestrial and marine surfaces and simultaneously from two aforementioned surfaces at once. A statistical approach and two families of distributions are chosen for modeling the signals, which enable synthesis of efficient algorithms for modeling the fluxes of random variables. The input signals of sea surface reflections are approximated by the log-normal law, while the reflections from the land surface are approximated by the Weibull distribution. For edge-coastal modeling, it is proposed to use the law of distribution of the vector sum of the flux of random variables comprising the input signals reflected from elementary land and sea areas that simultaneously fall within the corresponding resolution element. A highlight of the models developed is the consideration of both the probability distribution laws of the reflected signals and their correlation-spectral characteristics between resolution elements, as well as the anisotropy of reflections when monitoring surfaces from several angles. The presented mathematical models are consistent with experimental data sets of reflections, obtained from coastal zones. The development of coastal reflection models and corresponding modeling algorithms reduces the required number of actual field tests and saves time for implementation of monitoring systems, and hence, limits the overhead of this implementation.

A methodology and framework were presented to detect the seawalls accurately and efficiently in low coastal areas and were then evaluated in the study area of Hallandale Beach City, Broward County, Florida. Aerial images collected by the Florida Department of Transportation were processed using eCognition Developer software for multiresolution segmentation and classification of objects. Two classification approaches, pixel-based image analysis and the object-based image analysis method, were applied for image classification. However, pixel-based classification was discarded for having less accuracy in output. Three techniques within object-based classification-machine learning (ML) technique, knowledge-based (KB) technique, and ML followed by KB technique were used to compare the most efficient method of classification. While performing the ML technique, three algorithms, such as random forest, support vector machine, and decision tree, were applied to test the best algorithm. Of all the approaches used, the combination of ML and a KB method was able to map the seawall effectively with an overall accuracy of 94%.

Recent advancements in remote sensing facilitate the mapping of mineral occurrences but lack quantitative estimation of mineral abundances. In the present study, EO-1 Hyperion data are used for mapping and estimating beach placer minerals in the Cuddalore coastal stretch of Tamil Nadu, India, using spectral angle mapper (SAM) algorithm, continuum removed band depth analysis, and random forest (RF) regression. Strategic beach placer minerals are exposed in the image due to their specific spectral reflection/absorption characteristics; these unique settings are used to explore their spatial distribution from associated features. EO-1 Hyperion images are analyzed for (i) atmospheric correction; (ii) selection and transformation of optimum bands; (iii) fixing the input pixels from selected bands; (iv) generating n-D angle—for the preparation of end-members with unique spectra of mineral classes; (v) comparative analysis between true end-member and reference spectra of minerals using SAM, spectral feature fitting, and binary encoding algorithms; and (vi) image classification using SAM. The result of SAM shows the occurrence of placer mineral deposits such as zircon with an overall accuracy of 81.06% and Kappa coefficient of 0.71. The zircon shows strong absorption in spectral geometric parameters of EO-1 Hyperion data, such as band depth and band area at the spectral range of 1075 to 1150 nm. The positive correlation between geometric parameters of mineral spectra in its absorption characteristics and the mineral concentration of in situ samplings is used for developing an RF prediction model for estimating the placer minerals from beach sand deposits. Significantly, the lower RMSE values estimated at 0.082 and 0.156 for reference spectra (in situ) and image spectra respectively proved the ability of EO-1 Hyperion data for quantifying mineral resources.

Biomass is a critical biophysical parameter used to monitor agricultural and terrestrial ecosystems. Operational difficulties in measuring biomass on the field scale led to use of remote sensing techniques. Numerous vegetation indices (VIs) have been developed for this purpose, most of which utilize bands in the VIS-NIR spectral region. However, a significant number of them exhibit saturation at leaf area index (LAI) higher than 2. We propose an innovative vegetation index red-edge ratio normalized difference vegetation index (RERNDVI), that combines a VI sensitive to moderate to high LAI with VI sensitive to low to moderate LAI. An empirical assessment of the performance of our new index compared with nine existing spectral indices was conducted using hyperspectral and biomass data collected by Center for Advanced Land Management Information Technologies for maize and soybean fields, and Sentinel-2 data for semi-arid shrubland sites for which biomass was estimated based on the allometric model. The coefficient of determination (r²) of RERNDVI with green biomass for maize and soybean is high, but equally important, its noise equivalence is significantly lower than that obtained for all other indices for the full range of biomass levels. Nevertheless, RERNDVI-biomass relationships vary for different crops and shrubs, suggesting that generalizing these relationships will require information regarding canopy structure parameters as well.

The NASA Clouds and the Earth’s Radiant Energy System (CERES) project provides observed flux and cloud products for the climate science community. Geostationary satellite (GEO) imager measured clouds and broadband derived fluxes are incorporated in the CERES SYN1deg product to provide regional diurnal information in between Sun-synchronous Terra and Aqua CERES measurements. The recently launched GEO imagers with onboard calibration systems have active calibration teams that incrementally update the calibration in order to mitigate calibration drifts. However, short-term L1B radiance anomalies and calibration adjustment discontinuities may still exist in the record. To avoid any GEO cloud and flux artifacts in the CERES SYN1deg product, these calibration events must be addressed while scaling the GEO imagers to the Aqua-moderate resolution imaging spectroradiometer (MODIS) calibration reference. All-sky tropical ocean ray-matching (ATO-RM) and deep convective cloud invariant target (DCC-IT)-based monitoring algorithms are presented to detect calibration-driven daily anomalies in the GOES-16 Advanced Baseline Imager L1B visible (0.65  μm) radiance measurements. Sufficient daily ATO-RM sampling was obtained both by ray-matching GOES-16 with multiple MODIS and visible-infrared imaging radiometer suite imagers as well as by increasing the grid resolution. Optimized angular matching and outlier filtering were most effective in reducing the ATO-RM daily gain algorithm noise. The DCC-IT daily calibration algorithm utilized a larger domain and included more GOES-16 scan times. The DCC-IT daily gain uncertainty was reduced by normalizing the DCC regional reflectance on a regional, seasonal, and diurnal basis. The combination of ATO-RM and DCC-IT daily monitoring algorithms is shown to detect, with a high degree of confidence, daily GOES-16 L1B calibration-driven radiance anomalies >2.4  %  , while keeping false positives at a minimum. Remarkably, the ATO-RM and DCC-IT daily gains are mostly within 0.5%. The ATO-RM and DCC-IT daily monitoring algorithms can be easily adapted to other GEO imagers and visible channels.

Landslides are natural calamities that happen all over the world. Every year, landslides claim the lives of thousands of people and cause significant damage worldwide. The rapid and unpredictable expansion of many cities has a significant impact on the physical environment, with rapidly urbanizing regions of uneven topography where environmental circumstances make the construction of structures and infrastructure impossible or susceptible to instability. Our work proposes the integration of remote sensing, a knowledge-based numerical rating scheme, and multiple overlay analysis methodologies for landslide susceptibility mapping. We evaluate binary overlay, weighted overlay, and fuzzy overlay to map landslide vulnerability in the Ambegaon taluka of Pune district, Maharashtra. The input information used for the prediction and valuation of the landslide susceptibility map (LSM) includes an inventory of 39 active landslide points and 10 potential causality features of landslides: rainfall, slope, aspect, curvature, elevation, Euclidean distance to streamline, normalized difference vegetation index, topographic wetness index, Euclidean distance to roadline, and lithology. The area of consideration was categorized into five main susceptibility groups based on the computed landslide susceptibility scale, ranging from extremely low to severe. The LSM created using the fuzzy gamma operator (k  =  0.95) has an overall forecast and prediction accuracy of 89.74%. The optimum value of the gamma operator (k) for this research is 0.95. The LSM was verified by correlating the frequency of landslides on different hazard classes. Implementation of this LSM at the regional level offers a foundation for expanding the approach to other geographies.

Drought is one of the most complex and threatening disasters, defined as a period of drier-than-normal conditions which results in water-related problems. In addition to climate variation, human activities can cause or intensify drought through groundwater overextraction, unsupervised expansion of agricultural lands, etc. One prime example of such drought has been observed in the Zayandehrud River basin in central Iran. We investigate the long-term (1986 to 2019) drought of the Zayandehrud River basin using meteorological, hydrological, and agricultural drought indices. The main focus of this research is to explore the impact of the irrigated area expansion in the upstream sub-basins on the agricultural drought of the downstream. To this end, different drought indices are employed over the whole basin and its sub-basins. The results indicate that the long-term precipitation index and snow reserves do not show a long-term trend during the study period, while the combined precipitation–evapotranspiration index and groundwater storage decline. The research concludes that the expansion of the agricultural lands in the river upstream, rising air temperature, and increasing population and industrial activities in the basin are most likely the driving forces to intensify the agricultural drought in the midstream and downstream parts of the basin.

Shanghai Pilot Free Trade Zone (SPFTZ) is the first free trade zone in China, consisting of eight independent zones and located in Pudong New District. With the rapid development in the SPFTZ, frequent human activities may induce severe land subsidence. Based on time-series interferometric synthetic aperture radar (InSAR) technique and multitemporal C-band images, the temporal and spatial evolution of subsidence in Pudong New District is obtained from May 2016 to October 2020. The achieved results show that severe subsidence occurred in the east of the study area, where the maximum subsidence rate was −40  mm  /  year. Most of the land in the SPFTZ remained stable, but small-scale subsidence with a maximum subsidence rate of −18.1  mm  /  year occurred in two zones. InSAR-derived results are in good agreement with the manual measurements, with a root mean square error of 1.4 mm. Time-series analysis reveals that land subsidence in the SPFTZ mainly originates from urban constructions, including building construction and vehicle load. Meanwhile, the most severe subsidence occurred in the east of Pudong International Airport, which may be attributed to the airport renovation project. In a reclaimed area, land subsidence can be classified into different construction periods, which gradually weakens from north to south.

As one of the driest and lowest regions in Asia, the vegetation variation in the Aydingkol Lake Basin in Xinjiang, China, reveals the status of its ecological recovery. To more accurately retrieve the fractional vegetation index (FVC) to investigate the vegetation variation in this basin, we proposed an innovative method—the random forest (RF) method—based on multiple remote sensing indices. The retrieval shows a satisfactory accuracy with a mean error, mean square error, mean relative error (RE) percentage, and determination coefficient of 0.01,0.04,14.61% and 0.95, respectively. In the Aydingkol Lake Basin, the average annual FVC peaked in 2015 with a value of 0.08. The annual average value of FVC decreases at a rate of 0.06  *  10^−  2  /  a during these eight years. The reduction of FVC is mainly distributed in the central-western part of the basin. In exploring the influence on FVC, average annual precipitation has the most significant impact on the FVC. Its correlation coefficient with FVC is 0.87. With increasing elevation and slope, most FVC showed a decreasing trend. In recent years, the conversion of cropland to barren land and cropland to grassland had the greatest impact on decreasing of the FVC, and the mean FVC decreases from 0.19 to 0.17 before and after the land-use change. Moreover, human impact had a significant influence on the variation in FVC. Due to the establishment of the Aydingkol Lake Basin Wetland Reserve (the SAIC Volkswagen Test Site), the annual average FVC in the corresponding place increases from 0.06 to 0.13 (decreases from 0.148 to 0.005). This study is helpful to guide ecological protection in similar arid regions.

Many large-scale energy, transportation, water conservancy, and hydropower engineering projects take place in a rock mass environment. It is still difficult to accurately extract and describe rock mass structural information, which is crucial to analyzing rock mass deformation and stability. As a major rock mass surface exposure type, the structural surface has an irregular and rough shape, for which it is not appropriate to use traditional image processing methods. An intelligent extraction method is proposed for rock structural surfaces based on the integration of multimodal semantic features and a full convolutional neural network. The main contents of the proposed method are as follows: (1) generation of a dense point cloud model of the rock mass surface using multiview images and 3D geological semantic feature expression; (2) establishment of the mapping relationship among multimodal semantic features; (3) homogeneous unit extraction through integration of a full convolution neural network model; and (4) homogeneous unit merging and clustering. The method’s feasibility is proven through experiments, and its completeness and accuracy are verified by comparing results with traditional field measurements. Overall, the method provides a concept for intelligent extraction of rock mass structure information and has important theoretical value and practical significance.

Taft-Dehshir region is located in the central part of the Urmia-Dokhtar Magmatic Belt (UDMB) in central Iran. Some Cu mineralizations that occurred in the UDMB are definable by surface patterns of hydrothermal alterations and thus are detectable by advanced spaceborne thermal emission and reflection radiometer (ASTER)-based remote sensing. The final product of ASTER-based remote sensing was a prepared unique map that defined the type and pattern of alterations. Then, this map was integrated with complementary petrographic studies, XRD analysis, and careful field surveys, which were meticulously prepared based on the existing literature. The results of this research show that the hydrothermal alterations around the stockworks and skarnization in carbonate host rocks caused the formation of different Cu mineralizations such as porphyry (especially in Aliabad area), porphyry-skarn (in Darre-Zereshk area), Skarn (in Khezrabad and Khut areas), and vein (in Nasrabad area) types. This research determined the 18 decisive occurrences and five promising places that occurred in Oligocene-Miocene granitoid bodies.

We evaluated the feasibility for operational snow drought monitoring over Europe based on the near-real-time snow water equivalent (SWE) satellite product from the EUMETSAT Satellite Application Facility on Support to Operational Hydrology and Water Management (H-SAF). To do so, the consistency of this dataset with the consolidated dataset of the Canadian Meteorological Centre (CMC), as well as with the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts, was tested in terms of both spatial snow coverage and detection of anomalies from the long-term climatology. The analysis confirms a general good agreement among the three products as well as substantial differences over mountainous terrains, with the H-SAF product capturing only about 30% of the areas identified by CMC as snow-covered in those areas, while a better match between the ERA5 and the CMC spatial coverage is observed. However, significant inconsistencies in the correlation between all three SWE anomalies are observed over mountain areas. Due to the lack of a reliable reference dataset, the observed inconsistencies and the coarse spatial resolution (0.25 deg) of all three products limit the possibility for snow drought monitoring over key European regions such as the Alps.

This study proposes three clustered adaptive neuro-fuzzy inference system (ANFIS) models for the prediction of net radiation flux over climatic regions in Nigeria. Five temperature indices data obtained from the Nigerian Meteorological Agency, Lagos covering a period 1983 to 2013 were used as input variables to train and test the ANFIS models. Comparative analyses of the three ANFIS models—grid partitioning clustering (GPC), the subtractive clustering (SC), and the fuzzy c-means clustering (FCM) using the minimum values of coefficient of uncertainty (UII)—showed that ANFIS-GPC performed best in the semi-arid and sub-humid humid regions with values of 0.089 and 0.092, respectively. ANFIS-FCM performed best in the sub-humid dry region with a value of 0.074. ANFIS-SC performed best in the humid region with a value of 0.080. It can be concluded that the proposed models are regional dependent, and they are suitable for the prediction of net radiation flux over respective tropical regions for practical purposes.

Water hyacinth (Eichhornia crassipes) has become a threat to many aquatic environments worldwide. This aquatic weed presents a rapid reproductive capacity and outcompetes other aquatic plant species, forming dense, free-floating mats, which in many instances completely cover fresh-water surfaces. The infestation leads to several serious environmental (including ecological and socioeconomic) impacts that are hazardous to aquatic systems, disables human uses of surface waters, and affects hydraulic infrastructures (e.g., waterways and pumping stations). Our study explores the use of remote sensing tools to monitor and categorize the spread of water hyacinth, aiming at new insights into the detection, observation, and mapping of this invasive plant using vegetation indices and water indices calculated from multispectral data from satellite Sentinel-2, such as normalized difference vegetation index and normalized difference water index. The approach uses spatiotemporal information and has the potential to contribute to inform planners and decision-makers that are concerned with managing the plant by applying integrated measures. The case study deals with a small water course located in the downstream part of the Mondego river valley in Portugal, a country where water hyacinth is widely spread and constitutes a major problem, mainly in the irrigated agricultural areas of the Tagus, Sado, Mondego, and Lis rivers’ basins.

As one of the most important research directions in the field of computer vision, target detection has made great progress. However, in the field of remote sensing image processing, target detection still has the problem of insufficient feature extraction about the small targets and the low accuracy caused by large-scale difference. We propose a multiscale remote sensing image target detection method based on frequency attention and feature fusion enhancement to achieve better detection results. This method can effectively learn the features extracted by multifrequency components through the frequency channel attention network, realize the interaction of diversified information from various channels, and improve the representation ability of the network. A unique feature pyramid network is designed using interactive up-and-down sampling and skip connection, which sufficiently fuses the features, effectively captures multiscale context information, and improves the detection performance of the model. A feature enhancement network combining dilated convolution and residuals is designed to mine deeper semantic information. The experimental results show that the proposed method achieves higher detection accuracy on the two datasets NWPU VHR-10 and RSOD, when compared with some state-of-the-art target detection methods, and effectively improves the detection performance.

Very high spatial resolution (VHR) images are accessible for land-cover classification. Because of the complex information and deep features associated with increased spatial resolution, it is difficult to find an effective method for improving the classification accuracy of finer-resolution remote sensing images. We propose a scheme for sample selection and sample autoannotation to establish a new labeled benchmark using vector maps. In addition, a suitable multiscale object-driven convolutional neural networks (multi-OCNNs) were developed for the new benchmark. We trained multi-OCNNs to capture the deep and contextual information contained in samples of the VM-RSI benchmark to predict the land cover categories from target region images with no labels. Furthermore, an object-driven voting strategy applied in multi-OCNNs provides clear boundary information of ground objects and less noise in CNN predictions. Experiments on multisource VHR images, including SPOT-6, Gaofen-2, and ALOS satellite images, show that the proposed approach outperforms other object-based algorithms by achieving good performance results, verifying its feasibility, and effectiveness for land cover classification.

Lithological studies and geological unit mappings are generally applicable to many fields of natural resource management. Relatively suitable aquifers have been formed in complex formations in northwest Shahrood due to the presence of carbonate rocks as well as erosion and tectonic forces in the region. This study aims to identify and separate the calcareous formations that can form karst aquifers in the study area. As a result of erosion and tectonic forces, the rocks of the region exhibit spectral fluctuations, making it difficult for mapping geological formations using multispectral images. Therefore, Landsat 8 satellite-based images were processed by adopting the minimum noise fraction (MNF), independent component analysis (ICA), and band ratio (BR). The indices of calcareous formations and shale formations were created by the BR through the spectral behavior of pure pixels. Moreover, the support vector machine (SVM) and maximum likelihood (ML) were employed for classification. The SVM classifier proved more capable of classification than the ML classifier, and the transform ICA outperformed the MNF in the separation of formations. The lithological maps were extracted using the SVM with an overall accuracy (OA) of 68.04%. Furthermore, a method decision tree (DT) was employed to improve the classification accuracy. The DT classifier was then utilized to reclassify lithological maps that were classified by SVM and ML through morphological characteristics and indices of formations The DT classifier improved the lithological map accuracy by 10%. The boundaries of calcareous formations were extracted from non-calcareous formations with an accuracy of 93%, and the regional constructions were separated with an accuracy of ∼80  %  . Finally, the lithological map was developed with a kappa of 0.734 and an OA of 78.59%.

Remote sensing images contain complex feature information, and traditional convolutional networks cannot effectively model the contextual relationships. To address this problem, we propose a semantic segmentation network for remote sensing images based on semantic relationship aware. We construct the semantic relationship aware module to obtain the global semantic information of remote sensing images by self-attention. In addition, the separable space convergence pyramid module was constructed to effectively utilize the feature information in the high-level feature maps. By separable convolution with different dilation rates, the network can acquire multiscale semantic information. Our semantic relation aware network (SRANet) improves the overall accuracy by 0.33% over the benchmark network in the Vaihingen dataset and by 0.42% in the Potsdam dataset. The class activation maps show that the SRANet has ideal activation responses for targets at different scales in images. Furthermore, our SRANet can produce competitive segmentation performance compared with other state-of-the-art segmentation networks.

Accurate information about land-cover change (LCC) is important in many different fields. GlobeLand30 is a widely used global land-cover product that can support change analysis. However, since the data sources used in GlobeLand30 products are acquired by different sensors, it is hard to achieve accurate geometric registration between classified images; as a result, spurious changes may be detected. For this reason, we aimed to investigate the effects of geometric misregistration on LCC detection using the 2010 and 2020 GlobeLand30 products for the whole of China to quantitatively understand the impact of geometric misregistration in GlobeLand30 on the analysis of LCC. First, we quantified the effect of the misregistration using a block-statistical approach that calculates the difference in the change ratio based on a pixel-by-pixel comparison method together with a method that estimates the proportional change in classes within 1-km blocks (a block-by-block comparison). The results obtained using the pixel-by-pixel comparison method and the block-by-block method give land-change ratios for China as 10.6% and 6.95%, respectively, which indicate that about 3.65% of the changes may be spurious changes caused by geometric misregistration. We then analyzed the relationship between the landscape heterogeneity and the spurious changes using the Geodetector model. The results show that landscape heterogeneity explains the spatial distribution of spurious changes due to the misregistration effect to a certain extent and show a positive correlation. In terms of the Shannon diversity index, the explanatory ability reaches more than 50% in most subregions. Our work provides a quantitative analysis of the influence of geometric misregistration on LCC measurements and has an important reference value for the updating of land-cover mapping used in change detection.

In mountainous terrain, the synthetic aperture radar (SAR) interferometry technology causes serious spatial decorrelation due to the shadows and overlaps in the range direction. We propose a minimum cost flow phase unwrapping method considering range direction slope (RDS). The RDS is calculated in the SAR imaging direction, and the RDS is used to divide the high- and low-quality regions of the interferometric phase. In the high-quality area, the residual error between the estimated value of the wrapped phase and the true value is greatly affected by the RDS. The quality map based on the slope in range direction is introduced to limit the number of residuals, at the same time, the quality map is used as the unit cost of the flow to balance the residuals of each node in the minimum cost flow, achieve the minimum cost of the minimum cost flow network, so as to achieve the minimum residual estimation and complete the unwrapping. The phase unwrapping experiments are carried out with the simulated data and the real data from Sentinal-1A in Qinling Mountain in China, and the experimental results show that the root-mean-square error of this method is lower than that of the path tracking method, the least squares method, and the traditional minimum cost flow method. The proposed method is suitable for phase unwrapping in complex mountainous areas.

Change detection is a significant issue for understanding the changes occurring on the land surface. We propose a change detection approach based on a semantic segmentation network from multispectral (MS) images. Different from the traditional approaches that learn deep features from the change index or establish mapping relations from patches, the proposed approach employs the semantic segmentation network UNet++ for end-to-end change detection. Nevertheless, in UNet++, the deep feature is directly upsampled from the node in the lower level and does not involve much information from the nodes in the other levels. To cope with this problem and further enhance its robustness, the zigzag UNet++ (ZUNet++) is developed. In ZUNet++, the zigzag connection between nodes can be found, so the inputs of the node involve not only the upsampled deep feature but also the downsampled shallow feature, i.e., the network fuses multiple feature information. In addition, as few MS training datasets are available, we designed a strategy in which each MS image is transferred into several pseudo-RGB images; thus the network is trained by available RGB training sets and can be applied to the testing MS datasets. In the experiment, three real testing MS datasets that reflect different types of changes in Xi’an City are used. Experimental results show that, upon determining the appropriate parameter, the proposed ZUNet++ outperforms the other state-of-the-art approaches, demonstrating its feasibility and effectiveness.

Hail weather is a strong weather phenomenon produced by a system of severe convective weather. Furthermore, hail weather has the characteristics of short duration, strong locality, and great destructive capacity and is a type of disastrous weather. Therefore, the recognition and tracking of hail weather have always been significant research topics in the field of meteorology. Doppler weather radar can obtain high precision and high frequency data by observing strong weather information with high temporal and spatial resolution. However, weather radar cannot directly recognize and inform forecasters of strong weather conditions; thus it takes considerable time and energy to identify and monitor the weather artificially. Based on the continuous development of artificial intelligence (AI) technology, combining the advantages of different observation data and using AI to process it are important ways to improve the strong weather monitoring and early warning systems. Our study presents a hail weather recognition method based on AI. In the proposed method, faster region-based convolutional neural network deep learning is applied to recognize hail weather areas, and the accuracy and precision on the test set are more than 86% and 97%, respectively. The reliability of the proposed method is verified by theoretical analysis, the actual hail situation, and comparison with the traditional hail recognition method. The experimental results provide a significant reference for the recognition and early warning of hail weather.

More and more high-resolution synthetic aperture radar (SAR) image datasets have been available, which promote the applications of SAR images in land cover classification, such as vegetation monitoring, land cover, land use, cartography, etc. A novel algorithm based on the encoder-decoder structure is proposed in this paper. We add the channel attention and spatial attention modules to the algorithm for SAR images land cover classification. And these two modules, which interact with each other instead of being completely independent, can maximize the use of extracted information. Experiments on the high-resolution Gaofen-3 dataset have been carried out. Some classical semantic segmentation algorithms, including SegNet, HR-Net, Deeplabv3+, and GF3-baseline, are utilized to compare with the proposed algorithm. The experimental results demonstrate that the encoder–decoder network with the attention mechanism can get higher accuracy than other encoder–decoder networks.

High sensitivity uncooled microbolometers are necessary to meet the needs of the next generation of infrared detectors, and vanadium oxide thin films are the potential candidates for uncooled microbolometers due to their high temperature coefficient of resistance (TCR) at room temperature. It is, however, very difficult to deposit vanadium oxide thin films having a high temperature coefficient of resistance because of the process limits in microbolometer fabrication. We present a fabrication method for vanadium oxide thin films. Through the formation of a vanadium oxide V_xO_y thermometer thin films with thickness of 95 nm prepared by sputter-depositing nine alternative layers of V₂O₅ and V of thicknesses of 15 and 5 nm, respectively. Two samples of vanadium oxide mixed phase are characterized in this work and compared to a not annealed sample, the first sample is annealed at 300°C for 30 min in O₂ and the second in N₂ atmosphere. The results show that annealing atmosphere has an effect on microstructure, optical, and thermal properties of the mixed phase. Comparing to our previous work, we have reached in this work successful results with vanadium oxide sample having both high TCR and low resistivity.

For small- or medium-sized river basins that suffer from floods, the lack of river gauging data is an important reason why flood warning at the appropriate time is difficult for local stakeholders. As a valuable dataset observed from spaceborne or airborne platforms, remote sensing images can capture slices of time-series information to cover the whole river basin. We consider the Nilwara Ganga tributaries in southern Sri Lanka for our research. The maximum flood extents and the corresponding elevations of flooded areas were determined in the post-flood Sentinel-1 images. Relying only on commonly available remote sensing datasets, the rainfall-triggered flood inundation models were proposed to simulate the maximum inundation extent of a potential flood. The results show that the models proposed here can be well applied in the determination of peak stage, especially for the non-tributary river reach (root mean square error = 0.59, R-square  =  0.76). The average of Intersection over Union (IoU) accuracy of the maximum flood extent predictions was 0.75. Validations of two flood events had the IoU accuracy of 0.79 and 0.69, respectively. For gauging data-scarce areas, this approach has great potential to provide a different perspective for flood early warning and will benefit the subsequent risk assessment and disaster mitigation.

In a whisk-broom sensor, such as the Advanced Baseline Imager (ABI) onboard the Geostationary Operational Environmental Satellite (GOES)-R series of geostationary satellites, the sensor optics projects the observed scene onto the sensor focal plane arrays (FPA), with the detectors arranged in a column perpendicular to the scan direction. An assessment system compares the images made by the sensor with known landmarks and calculates the image deviations from the landmark positions. These deviations are called navigation residuals, and in this work they are interpreted in a broad sense as deviations of the actual from the ideal optical paths. The deviations can be caused by several issues, such as a sensor pointing error, deficiency in sensor optics, and inaccuracy of detector locations on FPA. As the ABI sensor scans the Earth to make a full-disk image in 22 parallel swaths, we refer the landmark residuals from all swaths to the observing detector positions in the ABI detector column. This interpretation inverts the retrieval of navigation residuals and gives deviations of the image from the expected ideal position on an ideal FPA. As the errors of pointing and sensor optics can be assessed independently, the remaining deviations can be interpreted as these with respect to the ideal detector position map (in other words, to the pre-launch assessment). We call these deviations the apparent detector positions. Knowledge of these deviations will greatly simplify the navigation commissioning for future ABI sensors and similar devices. Using our archive of ABI image navigation data for ABI sensor on the GOES-16 satellite, we show the wealth of additional information about the sensor on-orbit condition that may be extracted with the proposed method. This algorithm can use any image navigation system that is sufficiently precise.

Total suspended sediments (TSS) are an important water quality indicator. The spatiotemporal distribution patterns of TSS concentrations in summer in the Peace-Athabasca Delta (PAD) remain unclear between 2000 and 2020. Among empirical and machine learning models, the random forest (RF) model exhibits the best performance for the PAD with a coefficient of determination, normalized root-means-square error, and relative root-mean-square error of 0.92, 7.51%, and 27.36%, respectively. Our study analyzes the spatiotemporal TSS distribution based on 21 years of Landsat remote sensing data and the RF model. The results showed that the TSS concentrations of the Peace River were elevated in 2000, 2016, 2019, and 2020. The long-term TSS concentration distribution in the PAD showed a decreasing trend from west to east, and the TSS concentrations ranged from 6 to 80 mg/L level and remained stable in Lake Athabasca during the investigated period. Moreover, 57.14% of the PAD area exhibited no significant change between 2000 and 2020. The results of our study provide detailed information for the local government on the TSS concentrations in the PAD.

We propose a polarimetric synthetic aperture radar (PolSAR) image terrain classification algorithm based on complete local binary patterns (CLBP) feature integrated convolutional neural network (CNN) (CLBP-CNN). Traditional CNN has a powerful high-level deep features extraction ability, which can effectively improve the terrain classification accuracy in PolSAR images. However, most traditional CNN-based methods only focus on the high-level deep feature extraction of the synthetic aperture radar (SAR) terrains; they ignore the low-level texture features, resulting in incomplete feature representation and poor classification accuracy. In fact, low-level texture features also play an important role in PolSAR terrain classification. To solve the problem that traditional CNN-based terrain classification methods easily lose the low-level texture features in the process of feature extraction, the proposed method uses the CLBP descriptor to extract multi-level texture features under different receptive fields, and it adaptively combines the high-level deep features and the low-level texture features for better SAR terrain feature description. CLBP-CNN greatly alleviates the shortcomings of traditional CNN in missing the low-level texture features; it improves the feature representation completeness, so it can achieve better terrain classification results. The superiority of CLBP-CNN is verified on the data sets of Flevoland, San Francisco, and Oberpfaffenhofen.

Several studies have assessed forest disturbance in tropical forests using Landsat imagery. However, the spatial resolution (30 m) of Landsat images has often been considered too coarse to accurately detect the extent and impacts of selective logging. The Sentinel-2 satellite launched in 2015 has been providing images at spatial resolutions of 10 to 20 m and those images have shown an improved potential for detecting forest disturbances in tropical regions. We compared Landsat-8 and Sentinel-2 imagery for detecting selective logging in a rain forest site in the Brazilian Amazon. The aerosol-free modified soil adjusted vegetation index (MSAVI_af) was retrieved from the satellite images acquired in August 2020 immediately following logging. A robust reference dataset of very-high-resolution imagery (0.5 m) acquired using a complementary metal oxide semiconductor sensor (visible bands) onboard of an unmanned aerial vehicle was used to image the area of interest and a map derived from it was used to assess the classification accuracies made using satellite-derived data. The overall accuracy of the classified Sentinel-2 and Landsat-8 images varied between 54% and 83%, depending on the applied classification parameters for distinguishing undisturbed from disturbed forest canopy. Images acquired using the UAV allowed us to detect subtle impacts of canopy openings by selective logging activities. Images acquired using the UAV allowed the detection of small canopy openings, but not Sentinel-2 or Landsat-8. Sentinel-2 provided more details of canopy disturbances than Landsat image. Our classification approach is fully implementable on the Google Earth Engine platform and is a promising technique to monitor selective logging impacts in tropical forests.

We present MangoGAN, a general adversarial network (GAN)-based deep learning semantic segmentation model for the detection of mango tree crowns in remotely sensed aerial images. The aerial images are acquired by low-altitude remote sensing carried out using a quadrotor unmanned aerial vehicle in a mango orchard. Aerial images are acquired with a vision spectrum optical sensor, also popularly known as RGB images as the payload. MangoGAN is trained on 1430 images patches of size 240  ×  240  pixels. The testing was carried out on 160 images. Results are analyzed using the precision, recall, F1 parameters derived from contingency matrix and by visualization using Gradcam method. The performance of the MangoGAN is compared with peer architectures trained on the same data. MangoGAN outperforms its peer architectures

Desert use and its management play a significant role in the conservation of land resources. The premise of developing and using deserts based on local conditions includes managing the physical characteristics of sandy desert land. We explore the various achievements of remote sensing of bare soil moisture and, specifically, employ two methods involving the remote sensing of surface physical characteristics, dynamic measurements, and Global Navigation Satellite System-Reflectometry (GNSS-R) sensing by satellite. Some observation parameters of the Earth’s surface emissivity are summarized, which were attained through remote observations. In addition, the bright temperature inversion model on the soil temperature, moisture content, and texture, and the GNSS-R soil moisture sensor and inversion model were analyzed. The influence of horizontal polarization of the surface roughness on the results of dedicated satellite measurement of emissivity was also analyzed. Further, a sensing separation model of soil moisture particle size for GNSS-R was derived. Soil moisture is related to soil reflectivity. The soil particle size determines soil porosity, which is directly related to the evaporation rate of water and affects the surface temperature, thereby affecting soil emissivity. We determined that a level of complementarity exists between the enhancement of emissivity and attenuation of reflectivity by roughness as well as between the emissivity of dedicated satellite measurement line polarization and the reflectivity of GNSS-R measurement with circular polarization. Based on this complementarity, a method of GNSS-R and dedicated satellite collaborative measurement of soil texture using satellite data is proposed. The experimental results show that the two methods are sensitive to the soil moisture content, but the sensitivity to the mixed components of sandy soil is very poor. Both the sensitivity of GNSS-R sensing to sand particle size and sand surface temperature are better than those of dedicated satellite sensing.

Convolutional neural networks (CNNs) have shown excellent performance for hyperspectral image (HSI) classification due to their characteristics of both local connectivity and sharing weights. Nevertheless, with the in-depth study of network architecture, merely manual empirical design can no longer meet the current scenario needs. In addition, the existing CNN-based frameworks are heavily affected by the redundant three-dimensional cubes of the input and result in inefficient description issues of HSIs. We propose an image-based neural architecture automatic search framework (I-NAS) as an alternative to CNN. First, to alleviate the redundant spectral–spatial distribution, I-NAS feeds a full image into the framework via a label masking fashion. Second, an end-to-end cell-based structure search space is considered to enrich the feature representation. Then, it determined the optimal cells by employing a gradient descent search algorithm. Finally, the well-trained CNN architecture is automatically constructed by stacking the optimal cells. The experimental results from two real HSI datasets indicate that our proposal can provide a competitive performance in classification.

Point cloud data with high accuracy and high density is an important data source for the depiction of real ground objects, and there is a broad research prospect of using point cloud data directly for 3D object detection and recognition using deep learning methods. However, many deep learning models in previous research ignored the point cloud structure information and the sampling randomness. To overcome this limitation, we proposed an innovative 3D point cloud deep learning model, namely, the minimum bounding box over-segmentation–graph convolution 3D point cloud deep learning network model (MBBOS-GCN) for enhancing the structural information perception capability of the model and reduce the sampling randomness. In MBBOS-GCN, the number of points sampled is used as the scale, and a modified graph convolution model is used to collect point cloud structure information from different scales. The point cloud is divided into several small regions by the minimum bounding box algorithm, and the farthest point sampling (FPS) algorithm is used to sample within each small region to reduce sampling randomness. The experiments on object classification and semantic scene data segmentation show that: (1) the MBBOS-GCN model has high classification and segmentation accuracy, which is up to 91.87% and 89.5% on the ModelNet40 dataset and ScanNet dataset, respectively; (2) the MBBOS-GCN model is provided has good stability and robustness with a little change in accuracy under the altering density of input point cloud data, and slight classification loss value; (3) the MBBOS-GCN model can be adapted to real complex scenes when the classification accuracy reaches up to 97.53%. These superior performance of the MBBOS-GCN model can provide an effective support for the construction of digital twin city background data and the calibration of multimode satellite feature inversion algorithm validation.

Change detection is an important research direction in the field of remote sensing technology. However, for hyperspectral images, the nonlinear relationship between the two temporal images will increase the difficulty of judging whether the pixel is changed or not. To solve this problem, a hyperspectral change detection method is proposed in which the transformation matrices are obtained by using the constraint formula based on the minimum spectral angle, which uses both spectral and spatial information. Further, a kernel function is used to handle the nonlinear points. There are three main steps in the proposed method: first, the two temporal hyperspectral images are transformed into new dimensional space by a nonlinear function; second, in the dimension of observation, all the observations are combined into a vector, and then the two transformation matrices are obtained by using the formula of spectral angle constraint; and third, each pixel is given weight with a spatial weight map, which combined the spectral information and spatial information. Study results on three data sets indicate that the proposed method performs better than most unsupervised methods.

Aircraft detection is notoriously challenging owing to the orientation and size variations of aircraft objects. Existing detection pipelines compromise with efficiency or accuracy to deal with the large visual variations. We present a novel cascaded framework that joins object detection and orientation prediction through multi-task learning. The cascaded framework consists of three stages and operates in a coarse-to-fine manner. Each stage simultaneously rejects false targets, regresses the locations of object candidates, and calibrates the orientations of the candidates to upright gradually. After each stage, the range of orientation angles continuously decreases, which contributes to accurately distinguishing aircraft from non-aircraft. In addition, we design a generative-adversarial architecture in the cascaded framework. Specifically, its generator learns to cheat the discriminator by producing the feature representation that is invariant to the change of aircraft size. Meanwhile, the discriminator competes with the generator and aims to accurately discriminate the feature representations of small objects from the representations of large ones. Through competition between the generator and the discriminator, the feature distributions over small and large objects are made similar, thus resulting in more accurate detection. Experimental results demonstrate our method achieves significant improvements over existing algorithms on the challenging detection dataset, while still keeps real-time performance.

Change detection comprises a very important application in environmental studies involving multitemporal data obtained by remote sensing. Developing more accurate change detection methods is an ongoing challenge. Our study presents a new, unsupervised change detection method based on the concepts of stochastic distances and thresholding. To prove the effectiveness of the method, a study was carried out involving a region in southeastern Brazil, from 1999 to 2018, which underwent a high rate of environmental degradation caused by urban, industrial, and sand mining expansion. In this investigation, images obtained by thematic mapper and operational land imager sensors aboard the Landsat-5 and -8 satellites were used. Comparisons with the change vector analysis (CVA) method are included in the analyses. Results showed that the proposed method is capable of providing more accurate results in relation to the CVA method, after adequate parameterization, providing more realistic mappings with greater precision.

Recently, convolutional neural networks have greatly improved the performance of hyperspectral image (HSI) classification. However, these methods mainly use local spatial–spectral information for the HSI classification and require a large number of labeled samples to ensure high classification accuracy. In our study, we propose a multiscale nested U-Net (MsNU-Net) to capture the global context information and improve the HSI classification accuracy with a small number of labeled samples. We took an HSI as the input and constructed a nested U-Net to complete the classification. Because scale is very important for image recognition, we propose a simple but effective multiscale loss function. Apart from introducing multiscale features into the network, this method uses Gaussian filters to construct multiscale data, inputs the multiscale data into the nested U-Net with shared parameters, and calculates the sum of loss functions of different scales as the final loss function. Furthermore, it introduces different scales of global context information, thus improving the classification accuracy. To demonstrate its effectiveness, we carried out classification experiments on four widely used HSIs. The results show that this method could achieve a higher classification accuracy than the compared methods when only a small number of labeled samples is available. Furthermore, the codes of the proposed method will be made available freely at https://github.com/liubing220524/MsNU-Net.

The segmentation and extraction of buildings in high-resolution remote sensing images has good application prospects in military, civil, and other fields. With a depth encoder–decoder structure, U-Net is a frequently used model for high-precision image segmentation. However, the design of U-Net makes it hard to retain the detailed information of edges when processing the building segmentation. Specifically, the low-level features extracted from the shallow layer and the abstract features extracted from the deep layer cannot be completely merged, resulting in inaccurate segmentation. In response to this problem, we design a new multiscale feature extraction module that extracts target information through three convolution kernels of different scales. Taking U-Net as the baseline, by replacing skip connections with this module, we propose a multiscale feature extraction U-Net. This method can perform secondary feature extraction on the shallow feature information in the skip connection, refine the detailed information, and narrow the semantic gap between the low-level features and high-level features. It can not only improve the ability of the network to extract multiscale feature information, from a larger range to more layers to extract the edge detail information of the building in the remote sensing image, but also increase the number of skip connections to reduce network overfitting. Experimental results on Massachusetts remote sensing data and Massachusetts building data show that the method proposed offers significant improvement in terms of precision and accuracy compared with the methods full convolutional network, U-Net, SegNet, and high-resolution network, with an F1 score of 88.73%, mean IoU of 91.15%, precision of 89.74%, accuracy of 97.36%, and recall of 87.74%.

Low-spatial-resolution hyperspectral images have rich spectral information and poor spatial resolution, whereas high-spatial-resolution multispectral images have high spatial resolution and limited spectral information. Current methods cannot simultaneously consider the structure of spatial and spectral domains of an hyperspectral image cube. To solve this problem, we propose a fusion model based on spatial nonlocal similarity and low-rank prior learning. The proposed method first extracts a series of fixed-size, full-band tensor blocks, select their corresponding reference blocks to search similar full-band blocks in its search window, list similar blocks to form tensor groups, and then use a four-dimensional tensor structure to extract local full-band blocks. Next, a low-rank regularization term is introduced to the fusion model. Finally, the fusion problem is transformed to a low-rank regularization optimization problem, which is solved by the alternating direction multiplier method. Comparison with state-of-the-art methods demonstrates the method’s effectiveness.

In recent years, researchers have addressed the problem of using noncoherent approaches to estimate pulse width and pulse repetition interval. Since the measured transmitter is noncooperative, and noncoherent integration gain can be realized, the input signal-to-noise ratio (SNR) for these estimators becomes critical. We examine multiple edge detectors that exploit moving sums calculated as part of a Haar filtering of the received signal magnitudes. Two different ratio tests are considered in addition to the Haar filtering (or “difference of boxes”) approach, and a binary hypothesis test is designed based on a “smallest of” constant false alarm rate formulation. Probability arguments are then invoked to derive readily evaluated expressions for the detection thresholds. Tests are conducted, indicating that performance of the ratio-based approaches is comparable in terms of processed peak-to-background ratio. However, comparisons of root mean-squared (RMS) error indicate that the difference-based (Haar) approach produces lower error than both ratio-based approaches. The Haar filter approach is further demonstrated to remain effective (100% detection, 0% false alarm, RMS estimation errors of <3  %  ) at low SNRs of ∼0  dB.

Prolonging the integration time is a widely used method to enhance the detection performance for weak targets, whereas range migration (RM) and Doppler frequency migration (DFM) might occur when the target moves with high speed and high maneuverability. Although various algorithms have been proposed to solve the migration problem and realize coherent integration, they are based on the assumption that the target is moving radially relative to radar, i.e., the instantaneous slant range from radar to target is expressed as a finite order polynomial with respect to slow time. However, the target in the practical scene is probably moving obliquely relative to radar, thus modeling error will exist between the polynomial range expression and the target’s actual instantaneous slant range from radar. We focus on the accurate modeling and coherent integration problems for the weak maneuvering target with oblique movement, where the oblique movement model (OMM) of target echo is established. Subsequently, the OMM-generalized Radon–Fourier transform method is proposed to coherently accumulate the target’s energy via correcting the complex RM and DFM. Both simulation and real data processed results have been provided to demonstrate the effectiveness of the proposed method.

Grain size distribution is accepted as the basic control parameter in production processes in fields such as chemistry, mining, metallurgy, materials, construction, as well as in natural processes such as in the field of geology. Depending on the size of the grain, the methods used to determine the grain size are also varied. Laser scanning is a measurement technique in which 3D point information is obtained at very high speed compared to conventional measurement techniques. In our study, limestone aggregate samples were prepared by crushing and sieving into mixtures with different size distributions. Scaled digital photographs of these samples having known grain size distribution were taken, image analysis was applied on them, and the size distributions were determined. After that, the surfaces of the prepared mixtures were scanned using a terrestrial laser scanner and transferred to the computer environment as point cloud data. These data were analyzed by two different methods and first the D80, D50, and D30 characteristic sizes, representing the sieve aperture or the corresponding grain diameter through which 80%, 50%, and 30% of the sample passes, respectively, in the cumulative grain size distribution graph, were measured by roughness analysis method and then the size distribution were determined by analyzing point cloud data with a new algorithm. The validity of the methods was evaluated by comparing the D80, D50, and D30 sizes and size distributions determined by different methods with each other and with the sieve analysis results. According to the results obtained in our study, the analysis of LiDAR point cloud data can be successfully applied to determine the D80, D50, and D30 sizes and size distribution of aggregate piles.

Light detection and ranging (LiDAR) point clouds are sparse, unstructured, and disordered; hence, traditional convolutional neural networks are unsuitable for direct application in point-cloud data processing. Graph convolution neural networks (GCNs) can be used to process point-cloud data having the aforementioned characteristics; however, they are inefficient when the adjacent relationship of the point cloud is uncertain and adjacency-matrix elements are abundant. To solve these problems, a ground filtering method based on a superpoint graph convolution neural network (SPGCN) for vehicle LiDAR is proposed. With this method, the point-cloud data are transformed into a superpoint graph, and a GCN is employed for ground filtering. First, the points are projected on the range view and divided into grids, and the points in each grid are pooled in an average to form superpoints. Subsequently, a two-layer GCN is used to extract the features of superpoints, which are then assigned to each point. Finally, two fully connected layers are used to classify the ground and off-ground points. The SPGCN is trained and tested using the SemanticKITTI dataset, and the results confirm its effectiveness.

The use of remote sensing images for land cover analysis has broad prospects. At present, the resolution of aerial remote sensing images is getting higher and higher, and the span of time and space is getting larger and larger, therefore segmenting target objects enconter great difficulties. Convolutional neural networks are widely used in many image semantic segmentation tasks, but existing models often use simple accumulation of various convolutional layers or the direct stacking of interfeature reuse of up- and downsampling, the network very heavy. To improve the accuracy of land cover segmentation, we propose a multichannel feature fusion lozenge network. The multichannel feature fusion lozenge network (MLNet) is a three-sided network composed of three branches: one branch uses different levels of feature indexes to sample to maintain the integrity of high-frequency information; one branch focuses on contextual information and strengthens the compatibility of information within and between classes; and the last branch uses feature integration to filter redundant information based on multiresolution segmentation to extract key features. Compared with FCN, UNet, PSP, and other serial single road computing models, the MLNet, which performs feature fusion after three-way parallelism structure, can significantly improve the accuracy with only small increase in complexity. Experimental results show that the average accuracy of 85.30% is obtained on the land cover data set, which is much higher than that of 82.98% of FCN, 81.87% of UNet, 77.52% of SegNet, and 83.09% of EspNet, which proves the effectiveness of the model.

High-resolution near-field three-dimensional (3D) radar imaging techniques with millimeter and terahertz wave have played an increasingly significant role in applications including hidden object detection and nondestructive testing. Currently, imaging methods suffer from low efficiency when imaging defects in stratified media. This is caused by the inadaptability of the fast Fourier transform in the frequency domain due to the alteration of wave numbers in different media. We propose an imaging method based on the nonuniform fast Fourier transform to overcome the problem. Numerical simulations and experiments were carried out to verify our idea. Results show that our proposed method can greatly improve the imaging efficiency while maintaining the imaging quality. This work is of great significance for future development of nondestructive testing technology based on millimeter wave/terahertz imaging.

In radar imaging, the beamforming segment plays an important role in interference mitigation.

A beamforming algorithm based on frequency diverse array multiple-input multiple-output radar is proposed for suppressing false targets in imaging. A beampattern null spreading method based on weight vector design is presented by analyzing the composition of the weight vector. Specifically, the covariance matrix is reconstructed based on the artificial interference points. Then, the optimal weight vector can be solved by the algorithm to achieve the purpose of null spreading. As a result, the beamforming method effectively suppresses the false targets in the imaging scene. Simulation experiments are provided to illustrate the effectiveness and superiority of the proposed method.

We show an improvement in extraction of remotely sensed phenology in the Qinghai–Tibet Plateau (QTP). The improvement includes multiple preprocessing, newly proposed absolute growth rate and relative growth rate models based on botany and phenology, and selection of the appropriate growth rate model at different growing stages. We refer to this model as the tempo-differentially selected growth rate model (TDSGM). The newly developed TDSGM is a comprehensive and accurate remote sensing phenological extraction model without manual intervention, which is better than the current mainstream methods. Results show that: (1) influence of elevation and longitude on the QTP phenology has been more effective than that of latitude and vegetation types; (2) length of the growing season (LOS) was in a shortening trend in the long term, even it showed a slight extension in recent years in the short term; (3) the shortest LOS was identified in the area with an altitude of 3000 to 5500 m, which accounts for 90% of the total area of QTP. High instability of LOS was concentrated in meadows at 4000 to 5000 m; and (4) most previous studies on remotely sensed phenology showed that LOS was more related to the start of the growing season. However, we show that the LOSs of four out of seven QTP vegetation types were closely related to the end of the growth season, consistent with 1455 ground observations.

The northern Bay of Bengal (BoB) has been traditionally understudied and undersampled. Satellite and modeling products could compensate for the scarcity of in situ measurements, but this requires evaluating the accuracy of satellite and modeling products first. We present a comparison of sea surface temperature (SST) and sea surface salinity (SSS) products (satellite and model output) with 46 in situ observations in the northern BoB. We used satellite and modeled SST (daily) and SSS (weekly and daily) in this comparison. The results are as follows. (1) Both model and satellite-derived SSTs agreed well with in situ observations and with each other, with small biases (<1  °  C) and large correlation coefficients (r  >  0.77). (2) Neither model nor satellite SSSs agreed well with in situ observations (biases  >  0.5  PSU, r  <  0.54). (3) Calculations of the d-index support the argument that model and satellite SSTs agreed well with in situ observations (d-index values of 0.68 and 0.65, respectively), while the model and satellite SSSs did not agree well with observations (d-index values of 0.31 and 0.40, respectively). The results suggest that additional work is needed to improve both model prediction and satellite retrieval algorithms for SSS in the northern BoB.

Deep learning method has been extensively applied to ground penetrating radar two-dimensional profile (GPR B-SCAN) hyperbola detection recently. We propose a B-SCAN image feature extraction method based on the constraints of the GPR physical model, and further detect the weak boundary feature curve of the target in the local space. A deep convolutional neural network (DCNN) is first designed to extract high-level semantic features from B-SCAN images to remove direct wave. Next, a multiscale feature fusion DCNN is used to extract the features of the B-SCAN image with the direct wave removed, and the classifier network is used to identify the hyperbola of the upper boundary feature of the target. Finally, according to the hyperbola, the local space corresponding to the target in the B-SCAN image is determined. On this basis, the amplitude and phase information of the scattered electric field are used to segment the lower boundary characteristic curve of the target through convolution operation. Experimental results on simulation and field data show that feature information of the buried target in the GPR B-SCAN image can be efficiently extracted when the proposed method is adopted.

Due to the difficulty of acquiring labeled hyperspectral image (HSI), an unsupervised classification method known as deep adapted features alignment (DAFA) was proposed. First, deep adapted features from two domains are extracted by a convolutional deep adaptation network. Then, in order to align the distribution of two domains more accurately, the class-wise covariance and centroid alignment between deep adapted features from two domains are realized by matrix transformation and translation operation while maintaining the manifold structure of source features. Finally, source covariance and centroid aligned features and labels are used to train a base classifier, and target predictions are obtained by this classifier. Experiments conducted on several real HSI dataset pairs demonstrate the proposed DAFA method can effectively reduce distribution discrepancy between different HSIs and obtain good classification results.

The groundnut, also known as peanut, is a legume crop that is classified as both a grain legume and an oil crop (due to its high oil content). About 85% of the total groundnut in India is sown in the kharif season under rainfed conditions. This makes it extremely difficult to apply the optical remote sensing methodologies for groundnut discrimination and acreage estimation. We develop a remote sensing strategy based on timeseries synthetic aperture radar (SAR) imagery, which is immune to the rainy conditions of the observation period. Further, our work is distinct from the traditional multitemporal SAR backscatter analysis as it incorporates polarimetric parameters such as the alpha angle (α) and the polarimetric radar vegetation index to aid discrimination. These parameters were shortlisted on the basis of maximal sensitivity to groundnut growth stages. Random forest classifier, an ensemble learning-based classifier, is used to carryout the crop classification using these parameters. This approach is contrasted with SAR backscatter-based analysis (using the same RF classifier) and with a Euclidean distance (ED)-based model matching method (with the same polarimetric parameters). A detailed comparative analysis of classification accuracy of groundnut discrimination is carried out using extensive ground truth data over the agricultural fields of Junagadh district in the Gujarat state of Western India. The backscatter-based analysis with RF achieves 86.2% overall accuracy in discriminating groundnut (with the competing cotton crop), whereas polarimetric parameters with ED model matching achieves 76%. Using the proposed method, with dual-pol parameters derived out of Sentinel-1 SAR data, we report 92.7% overall accuracy with a kappa coefficient of 0.81. The major insight is that combining timeseries polarimetric information together with a good machine learning classifier improves the classification accuracy of short crops compared to nonpolarimetric approach.