Remote Sensing — Open Access Journal

The Sentinel-2 satellite mission, developed by the European Space Agency (ESA) for the Copernicus program of the European Union, provides repetitive multi-spectral observations of all Earth land surfaces at a high resolution. The Level 2A product is a basic product requested by many Sentinel-2 users: it provides surface reflectance after atmospheric correction, with a cloud and cloud shadow mask. The cloud/shadow mask is a key element to enable an automatic processing of Sentinel-2 data, and therefore, its performances must be accurately validated. To validate the Sentinel-2 operational Level 2A cloud mask, a software program named Active Learning Cloud Detection (ALCD) was developed, to produce reference cloud masks. Active learning methods allow reducing the number of necessary training samples by iteratively selecting them where the confidence of the classifier is low in the previous iterations. The ALCD method was designed to minimize human operator time thanks to a manually-supervised active learning method. The trained classifier uses a combination of spectral and multi-temporal information as input features and produces fully-classified images. The ALCD method was validated using visual criteria, consistency checks, and compared to another manually-generated cloud masks, with an overall accuracy above 98%. ALCD was used to create 32 reference cloud masks, on 10 different sites, with different seasons and cloud cover types. These masks were used to validate the cloud and shadow masks produced by three Sentinel-2 Level 2A processors: MAJA, used by the French Space Agency (CNES) to deliver Level 2A products, Sen2Cor, used by the European Space Agency (ESA), and FMask, used by the United States Geological Survey (USGS). The results show that MAJA and FMask perform similarly, with an overall accuracy around 90% (91% for MAJA, 90% for FMask), while Sen2Cor’s overall accuracy is 84%. The reference cloud masks, as well as the ALCD software used to generate them are made available to the Sentinel-2 user community. Full article

This study investigates the statistical characteristics of raindrop size distributions (DSDs) in monsoon season with observations collected by the second-generation Particle Size and Velocity (Parsivel²) disdrometer located in Zhuhai, southern China. The characteristics are quantified based on convective and stratiform precipitation classified using the rainfall intensity and total number of drops. On average of the whole dataset, the DSD characteristic in southern China consists of a higher number concentration of relatively small-sized drops when compare with eastern China and northern China, respectively. In the meanwhile, the D_m and log₁₀N_w scatter plots prove that the convective rain in monsoon season can be identified as maritime-like cluster. The DSD is in good agreement with a three-parameter gamma distribution, especially for the medium to large raindrop size. Using filtered data observed by Parsivel² disdrometer, a new Z–R relationship, Z= 498R^1.3, is derived for convective rain in monsoon season in southern China. These results offer insights of the microphysical nature of precipitation in Zhuhai during monsoon season, and provide essential information that may be useful for precipitation retrievals based on weather radar observations. Full article

Tropical Rainfall Measurement Mission (TRMM) is one of the most popular global high resolution satellite-based precipitation products with a goal of measuring precipitation over the oceans and tropics. However, in recent years, the TRMM mission has come to an end. Its successor, Global Precipitation Measurement (GPM) mission was launched to measure the earth’s precipitation structure, with an aim to improve upon the TRMM project. Both of the precipitation products have their own strengths and weaknesses in resolution, accuracy, and availability. The aim of this study is to evaluate the hydrologic utilization of the TRMM and GPM products in a humid basin of China. The main findings of this study can be summarized as follows: (1) 3B42V7 generally outperforms 3B42V6 in terms of hydrologic performance. Meanwhile, 3B42RTV7 significantly outperforms 3B42RTV6, and showed close performance with the bias-adjusted TRMM Multi-satellite Precipitation Analysis (TMPA) products. (2) The GPM showed better agreement with gauge observation than the TMPA products with lower RB and higher correlation coefficient (CC) values at different time scales. (3) The VIC hydrological model generally outperformed the XAJ hydrological model with lower RB, higher Nash–Sutcliffe Coefficient of Efficiency (NSCE) and CC values; though the 3B42RTV6 and 3B42RTV7 showed higher CC values in simulating the streamflow hydrograph by using the VIC and XAJ hydrological models. It can be found that the conceptual hydrological model was enough for the hydrologic evaluation of TRMM and GPM IMERG satellite-based precipitation in a humid basin of China. This study provides a reference for the comparison of multiple models on watershed scale. Full article

Feature matching via local descriptors is one of the most fundamental problems in many computer vision tasks, as well as in the remote sensing image processing community. For example, in terms of remote sensing image registration based on the feature, feature matching is a vital process to determine the quality of transform model. While in the process of feature matching, the quality of feature descriptor determines the matching result directly. At present, the most commonly used descriptor is hand-crafted by the designer’s expertise or intuition. However, it is hard to cover all the different cases, especially for remote sensing images with nonlinear grayscale deformation. Recently, deep learning shows explosive growth and improves the performance of tasks in various fields, especially in the computer vision community. Here, we created remote sensing image training patch samples, named Invar-Dataset in a novel and automatic way, then trained a deep learning convolutional neural network, named DescNet to generate a robust feature descriptor for feature matching. A special experiment was carried out to illustrate that our created training dataset was more helpful to train a network to generate a good feature descriptor. A qualitative experiment was then performed to show that feature descriptor vector learned by the DescNet could be used to register remote sensing images with large gray scale difference successfully. A quantitative experiment was then carried out to illustrate that the feature vector generated by the DescNet could acquire more matched points than those generated by hand-crafted feature Scale Invariant Feature Transform (SIFT) descriptor and other networks. On average, the matched points acquired by DescNet was almost twice those acquired by other methods. Finally, we analyzed the advantages of our created training dataset Invar-Dataset and DescNet and gave the possible development of training deep descriptor network. Full article

Monitoring surface movement near highways over soft clay subgrades is fundamental for understanding the dynamics of the settlement process and preventing hazards. Earlier studies have demonstrated the accuracy and cost-effectiveness of using time series radar interferometry (InSAR) technique to measure the ground deformation. However, the accuracy of the advanced differential InSAR techniques, including short baseline subset (SBAS) InSAR, is limited by the temporal deformation models used. In this study, a comparison of four widely used time series deformation models in InSAR, namely Multi Velocity Model (MVM), Permanent Velocity Model (PVM), Seasonal Model (SM) and Cubic Polynomial Model (CPM), was conducted to measure the long-term ground deformation after the construction of road embankment over soft clay subgrade. SBAS-InSAR technique with TerraSAR-X satellite imagery were conducted to generate the time series deformation data over the studied highway. In the experiments, three accuracy indices were applied to show the residual phase, mean temporal coherence and the RMS of high-pass deformation, respectively. In addition, the derived time series deformation maps of the highway based on the four selected models and 17 TerraSAR-X images acquired from June 2014 to November 2015 were compared. The leveling data was also used to validate the experimental results. Our results suggested the Seasonal Model is the most suitable model for the selected study site. Consequently, we analyzed two bridges in detail and three single points distributed near the highway. Compared with the ground leveling deformation measurements and results of other models, SM showed better consistency, with the accuracy of deformation to be ±7 mm. Full article

The time-varying characteristic of the bias in the GPS code observation is investigated using triple-frequency observations. The method for estimating the combined code bias is presented and the twelve-month (1 January–31 December 2016) triple-frequency GPS data set from 114 International GNSS Service (IGS) stations is processed to analyze the characteristic of the combined code bias. The results show that the main periods of the combined code bias are 12, 8, 6, 4, 4.8 and 2.67 h. The time-varying characteristic of the combined code bias, which is the combination of differential code bias (DCB) (P1–P5) and DCB (P1–P2), shows that the real satellite DCBs are also time-varying. The difference between the two sets of the computed constant parts of the combined code bias, with the IGS DCB products of DCB (P1–P2) and DCB (P1–P2) and the mean of the estimated 24-h combined code bias series, further show that the combined code bias cannot be replaced by the DCB (P1–P2) and DCB (P1–P5) products. The time-varying part of inter-frequency clock bias (IFCB) can be estimated by the phase and code observations and the phase based IFCB is the combinations of the triple-frequency satellite uncalibrated phase delays (UPDs) and the code-based IFCB is the function of the DCBs. The performances of the computed the IFCB with different methods in single point positioning indicate that the accuracy for the constant part of the combined code bias is reduced, when the IGS DCB products are used to compute. These performances also show that the time-varying part of IFCB estimated with phase observation is better than that of code observation. The predicted results show that 98% of the predicted constant part of the combined code bias can be corrected and the attenuation of the predicted accuracy is much less evident. However, the accuracy of the predicted time-varying part decreases significantly with the predicted time. Full article

Affected by the complexity of the indoor environment, accurate indoor positioning is challenging in many localization based services (LBS). Recently, it has been recognized that, visible light communication (VLC) is promising for indoor navigation and positioning, due to the low implementation cost with marginal modification to the existing infrastructure and the possibility to achieve high accurate positioning results. Provided that the positions of the light emitting diodes (LEDs) are known to the receiver, the angle of arrival (AOA) of the light signal is able to be estimated by a camera embedded in a smart phone, and thus the position of the smart phone can be derived based on the triangulation. In this paper, the performance of the positioning accuracy is analyzed based on indoor positioning with VLC, and the analytical upper bound of location error is derived. Extensive simulation results have verified the theoretical analysis on the VLC-based localization approach in different indoor scenarios. In order to obtain better location performance, the principles of choosing reference LED and localization LED are also given. Full article

The Selenga-Baikal Basin, a transboundary river basin between Mongolia and Russia, warmed at nearly twice the global rate and experienced enhanced human activities in recent decades. To understand the vegetation response to climate change, the dynamic spatial-temporal characteristics of the vegetation and the relationships between the vegetation dynamics and climate variability in the Selenga-Baikal Basin were investigated using the Normalized Difference Vegetation Index (NDVI) and gridded temperature and precipitation data for the period of 1982 to 2015. Our results indicated that precipitation played a key role in vegetation growth across regions that presented multiyear mean annual precipitation lower than 350 mm, although its importance became less apparent over regions with precipitation exceeding 350 mm. Because of the overall temperature-limited conditions, temperature had a more substantial impact on vegetation growth than precipitation. Generally, an increasing trend was observed in the growth of forest vegetation, which is heavily dependent on temperature, whereas a decreasing trend was detected for grassland, for which the predominant growth-limiting factor is precipitation. Additionally, human activities, such as urbanization, mining, increased wildfires, illegal logging, and livestock overgrazing are important factors driving vegetation change. Full article

Surface downward longwave radiation (DLR) is a crucial component in Earth’s surface energy balance. Yu et al. (2013) developed a parameterization for retrieving clear-sky DLR at high spatial resolution by combined use of satellite thermal infrared (TIR) data and column integrated water vapor (IWV). We extended the Yu2013 parameterization to Moderate Resolution Imaging Spectroradiometer (MODIS) data based on atmospheric radiative simulation, and we modified the parameterization to decrease the systematic negative biases at large IWVs. The new parameterization improved DLR accuracy by 1.9 to 3.1 W/m² for IWV ≥3 cm compared to the Yu2013 algorithm. We also compared the new parameterization with four algorithms, including two based on Top-of-Atmosphere (TOA) radiance and two using near-surface meteorological parameters and water vapor. The algorithms were first evaluated using simulated data and then applied to MODIS data and validated using surface measurements at 14 stations around the globe. The results suggest that the new parameterization outperforms the TOA-radiance based algorithms in the regions where ground temperature is substantially different (enough that the difference between them is as large as 20 K) from skin air temperature. The parameterization also works well at high elevations where atmospheric parameter-based algorithms often have large biases. Furthermore, comparing different sources of atmospheric input data, we found that using the parameters interpolated from atmospheric reanalysis data improved the DLR estimation by 7.8 W/m² for the new parameterization and 19.1 W/m² for other algorithms at high-altitude sites, as compared to MODIS atmospheric products. Full article

Fully polarimetric synthetic aperture radar (PolSAR) can transmit and receive electromagnetic energy on four polarization channels (HH, HV, VH, VV). The data acquired from four channels have both similarities and complementarities. Utilizing the information between the four channels can considerably improve the performance of PolSAR image classification. Convolutional neural network can be used to extract the channel-spatial features of PolSAR images. Self-paced learning has been demonstrated to be instrumental in enhancing the learning robustness of convolutional neural network. In this paper, a novel classification method for PolSAR images using self-paced convolutional neural network (SPCNN) is proposed. In our method, each pixel is denoted by a 3-dimensional tensor block formed by its scattering intensity values on four channels, Pauli’s RGB values and its neighborhood information. Then, we train SPCNN to extract the channel-spatial features and obtain the classification results. Inspired by self-paced learning, SPCNN learns the easier samples first and gradually involves more difficult samples into the training process. This learning mechanism can make network converge to better values. The proposed method achieved state-of-the-art performances on four real PolSAR dataset. Full article

Investigation of microwave scattering mechanisms is extremely important for developing methods for ocean remote sensing. Recent studies have shown that a common two-scale scattering model accounting for resonance (Bragg) scattering has some drawbacks, in particular it often overestimates the vertical-to-horizontal polarization radar return ratio and underestimates the radar Doppler shifts if the latter are assumed as associated with quasi linear resonance surface waves. It is supposed nowadays that radar backscattering at moderate incidence angles is determined not only by resonance Bragg mechanism but also contains non polarized (non Bragg) component which is associated supposedly with wave breaking but which is still insufficiently studied. Better understanding of the scattering mechanisms can be achieved when studying variations of radar return due to long wind waves. In this paper, results of experiments from an Oceanographic Platform on the Black Sea using dual co-polarized X-band scatterometers working at moderate incidence are presented and variations of Bragg and non-Bragg components (BC and NBC, respectively) and radar Doppler shifts are analysed. It is established that BC and NBC are non-uniformly distributed over profile of dominant (decametre-scale) wind waves (DWW). Variations of BC are characterized by some “background” return weakly modulated with the dominant wind wave periods, while NBC is determined mostly by rare and strong spikes occurred near the crests of the most intense individual waves in groups of DWW. We hypothesize that the spikes are due to intensification of nonlinear structures on the profile of short, decimetre-scale wind waves when the latter are amplified by intense DWW. Bragg scattering in slicks under the experimental conditions was suppressed stronger than NBC and spikes dominated in total radar return. It is obtained that radar Doppler shifts at HH-polarization are larger than at VV-polarization, particularly in slicks, the same relation is for NBC and BC Doppler shifts, thus indicating different scattering mechanisms for these components. Full article

In this paper, a system-specific phased-array radar system simulator was developed, based on a time-domain modeling and simulation method, mainly for system performance evaluation of the future Spectrum-Efficient National Surveillance Radar (SENSR). The goal of the simulation study was to establish a complete data quality prediction method based on specific radar hardware and electronics designs. The distributed weather targets were modeled using a covariance matrix-based method. The data quality analysis was conducted using Next-Generation Radar (NEXRAD) Level-II data as a basis, in which the impact of various pulse compression waveforms and channel electronic instability on weather radar data quality was evaluated. Two typical weather scenarios were employed to assess the simulator’s performance, including a tornado case and a convective precipitation case. Also, modeling of some demonstration systems was evaluated, including a generic weather radar, a planar polarimetric phased-array radar, and a cylindrical polarimetric phased-array radar. Corresponding error statistics were provided to help multifunction phased-array radar (MPAR) designers perform trade-off studies. Full article

Change detection on bitemporal synthetic aperture radar (SAR) images is a key branch of SAR image interpretation. However, it is challenging due to speckle and unavoidable registration errors within bitemporal SAR images. A key issue is whether and how despeckling and structural features can improve accuracy. In this paper, we investigate how despeckling and structural features can benefit change detection for SAR images. Several change detection methods were performed on both input images and the corresponding despeckled images, where despeckling was achieved by different methods. The comparisons demonstrate that despeckling methods that preserve the structures can suppress noise in difference images and can improve the accuracy of change detection. We also developed a sparse model to exploit structural features from the difference images while reducing the influence of misalignment between bitemporal SAR images. The comparisons were performed on five datasets of bitemporal SAR images, and the experimental results show that our proposed sparse model outperforms other traditional methods, demonstrating the advantages of change detection. Full article

Recent open data initiatives allow free access to a vast amount of light detection and ranging (LiDAR) data in many cities. However, most open LiDAR data of cities are acquired by airborne scanning, where points on building façades are sparse or even completely missing due to occlusions in the urban environment, leading to the absence of façade details. This paper presents an approach for improving the LiDAR data coverage on building façades by using point cloud generated from ground images. A coarse-to-fine strategy is proposed to fuse these two-point clouds of different sources with very limited overlaps. First, the façade point cloud generated from ground images is leveled by adjusting the facade normal to perpendicular to the upright direction. Then leveling façade point cloud is geolocated by alignment between images GPS data and their structure from motion (SfM) coordinates. Next, a modified coherent point drift algorithm with (surface) normal consistency is proposed to accurately align the façade point cloud to the LiDAR data. The significance of this work resides in the use of 2D overlapping points on the building outlines instead of the limited 3D overlap between the two-point clouds. This way we can still achieve reliable and precise registration under incomplete coverage and ambiguous correspondence. Experiments show that the proposed approach can significantly improve the façade details in open LiDAR data, and achieve 2 to 10 times higher registration accuracy, when compared to classic registration methods. Full article

As an important part of maritime traffic, ships play an important role in military and civilian applications. However, ships’ appearances are susceptible to some factors such as lighting, occlusion, and sea state, making ship classification more challenging. This is of great importance when exploring global and detailed information for ship classification in optical remote sensing images. In this paper, a novel method to obtain discriminative feature representation of a ship image is proposed. The proposed classification framework consists of a multifeature ensemble based on convolutional neural network (ME-CNN). Specifically, two-dimensional discrete fractional Fourier transform (2D-DFrFT) is employed to extract multi-order amplitude and phase information, which contains such important information as profiles, edges, and corners; completed local binary pattern (CLBP) is used to obtain local information about ship images; Gabor filter is used to gain the global information about ship images. Then, deep convolutional neural network (CNN) is applied to extract more abstract features based on the above information. CNN, extracting high-level features automatically, has performed well for object classification tasks. After high-feature learning, as the one of fusion strategies, decision-level fusion is investigated for the final classification result. The average accuracy of the proposed approach is 98.75% on the BCCT200-resize data, 92.50% on the original BCCT200 data, and 87.33% on the challenging VAIS data, which validates the effectiveness of the proposed method when compared to the existing state-of-art algorithms. Full article

The extraction of high-accuracy co- and post-seismic deformation fields and inversions of seismic slip distributions is significant in the comprehension of seismogenic mechanisms. On 12 November 2017, a M_W 7.3 earthquake occurred on the border between Iran and Iraq. To construct the co-seismic deformation field, Sentinel-1A synthetic aperture radar (SAR) images from three tracks were used. Based on a prior knowledge, least-squares iterative approximation was employed to construct the three-dimensional (3D) co-seismic deformation field. to derive a time series of 2D post-seismic deformation, the multidimensional small baseline subset (MSBAS) technique was use. Co-seismic deformation fields were asymmetric; the maximum relative displacement was nearly 90cm in the radar line-of-sight between two centers of co-seismic deformation. The 3D co-seismic deformation field showed southwestward horizontal motion and continuous subsidence-to-uplift variation from northeast to southwest. The two-dimensional (2D) post-seismic deformation time series showed a gradual decaying trend and good correspondence with the aftershock distribution. The main mechanism of post-seismic deformation was an afterslip of the post-seismic faults. We used the elastic half-space model to invert co-seismic deformation fields and obtain source parameters of the slip model. The maximum and average slips were 2.5 and 0.72 m, respectively. The average slip angle was 126.38° and the moment magnitude was M_W 7.34. The results of this study will contribute to research on regional tectonic activities. Full article

Ku- and C-band spaceborne scatterometer sigma nought (σ°) backscatter data of snow covered landfast first-year sea ice from the Canadian Arctic Archipelago are acquired during the winter season with coincident in situ snow-thickness observations. Our objective is to describe a methodological framework for estimating relative snow thickness on first-year sea ice based on the variance in σ° from daily time series ASCAT and QuikSCAT scatterometer measurements during the late winter season prior to melt onset. We first describe our theoretical basis for this approach, including assumptions and conditions under which the method is ideally suited and then present observational evidence from four independent case studies to support our hypothesis. Results suggest that the approach can provide a relative measure of snow thickness prior to σ° detected melt onset at both Ku- and C-band frequencies. We observe that, during the late winter season, a thinner snow cover displays a larger variance in daily σ° compared to a thicker snow cover on first-year sea ice. This is because for a given increase in air temperature, a thinner snow cover manifests a larger increase in basal snow layer brine volume owing to its higher thermal conductivity, a larger increase in the dielectric constant and a larger increase in σ° at both Ku- and C bands. The approach does not apply when snow thickness distributions on first-year sea ice being compared are statistically similar, indicating that similar late winter σ° variances likely indicate regions of similar snow thickness. Full article

The ground surface soil heat flux (G₀) quantifies the energy transfer between the atmosphere and the ground through the land surface. However; it is difficult to obtain the spatial distribution of G₀ in permafrost regions because of the limitation of in situ observation and complication of ground surface conditions. This study aims at developing an improved G₀ parameterization scheme applicable to permafrost regions of the Qinghai-Tibet Plateau under clear-sky conditions. We validated several existing remote sensing-based models to estimate G₀ by analyzing in situ measurement data. Based on the validation of previous models on G₀; we added the solar time angle to the G₀ parameterization scheme; which considered the phase difference problem. The maximum values of RMSE and MAE between “measured G₀” and simulated G₀ using the improved parameterization scheme and in situ data were calculated to be 6.102 W/m² and 5.382 W/m²; respectively. When the error of the remotely sensed land surface temperature is less than 1 K and the surface albedo measured is less than 0.02; the accuracy of estimates based on remote sensing data for G₀ will be less than 5%. MODIS data (surface reflectance; land surface temperature; and emissivity) were used to calculate G₀ in a 10 x 10 km region around Tanggula site; which is located in the continuous permafrost region with long-term records of meteorological and permafrost parameters. The results obtained by the improved scheme and MODIS data were consistent with the observation. This study enhances our understanding of the impacts of climate change on the ground thermal regime of permafrost and the land surface processes between atmosphere and ground surface in cold regions. Full article

Polarimetric synthetic aperture radar (PolSAR) image classification has become more and more widely used in recent years. It is well known that PolSAR image classification is a dense prediction problem. The recently proposed fully convolutional networks (FCN) model, which is very good at dealing with the dense prediction problem, has great potential in resolving the task of PolSAR image classification. Nevertheless, for FCN, there are some problems to solve in PolSAR image classification. Fortunately, Li et al. proposed the sliding window fully convolutional networks (SFCN) model to tackle the problems of FCN in PolSAR image classification. However, only when the labeled training sample is sufficient, can SFCN achieve good classification results. To address the above mentioned problem, we propose adversarial reconstruction-classification networks (ARCN), which is based on SFCN and introduces reconstruction-classification networks (RCN) and adversarial training. The merit of our method is threefold: (i) A single composite representation that encodes information for supervised image classification and unsupervised image reconstruction can be constructed; (ii) By introducing adversarial training, the higher-order inconsistencies between the true image and reconstructed image can be detected and revised. Our method can achieve impressive performance in PolSAR image classification with fewer labeled training samples. We have validated its performance by comparing it against several state-of-the-art methods. Experimental results obtained by classifying three PolSAR images demonstrate the efficiency of the proposed method. Full article