Remote Sens., Volume 11, Issue 7 (April-1 2019) – 155 articles

Accurate estimation of above ground biomass (AGB) is very important for crop growth monitoring. The objective of this study was to estimate rice biomass by utilizing structural and meteorological features with widely used spectral features. Structural features were derived from the triangulated irregular network (TIN), which was directly built from structure from motion (SfM) point clouds. Growing degree days (GDD) was used as the meteorological feature. Three models were used to estimate rice AGB, including the simple linear regression (SLR) model, simple exponential regression (SER) model, and machine learning model (random forest). Compared to models that do not use structural and meteorological features (NDRE, R² = 0.64, RMSE = 286.79 g/m², MAE = 236.49 g/m²), models that include such features obtained better estimation accuracy (NDRE*Hcv/GDD, R² = 0.86, RMSE = 178.37 g/m², MAE = 127.34 g/m²). This study suggests that the estimation accuracy of rice biomass can benefit from the utilization of structural and meteorological features. Full article

Applications of stereo imagery acquired by cameras onboard unmanned aerial vehicles (UAVs) as practical forest inventory tools are hindered by the unavailability of ground surface elevation. It is still a challenging issue to remove the elevation of ground surface in leaf-on stereo imagery to extract forest canopy height without the help of lidar data. This study proposed a method for the extraction of forest canopy height through the synthesis of UAV stereo imagery of leaf-on and leaf-off, and further demonstrated that the extracted forest canopy height could be used for the inventory of deciduous forest aboveground biomass (AGB). The points cloud of the leaf-on and leaf-off stereo imagery was firstly extracted by an algorithm of structure from motion (SFM) using the same ground control points (GCP). The digital surface model (DSM) was produced by rasterizing the point cloud of UAV leaf-on. The point cloud of UAV leaf-off was processed by iterative median filtering to remove vegetation points, and the digital terrain model (DTM) was generated by the rasterization of the filtered point cloud. The mean canopy height model (MCHM) was derived from the DSM subtracted by the DTM (i.e., DSM-DTM). Forest AGB maps were generated using models developed based on the MCHM and sampling plots of forest AGB and were evaluated by those of lidar. Results showed that forest AGB maps from UAV stereo imagery were highly correlated with those from lidar data with R² higher than 0.94 and RMSE lower than 10.0 Mg/ha (i.e., relative RMSE 18.8%). These results demonstrated that UAV stereo imagery could be used as a practical inventory tool for deciduous forest AGB. Full article

The growing population in China has led to an increasing importance of crop area (CA) protection. A powerful tool for acquiring accurate and up-to-date CA maps is automatic mapping using information extracted from high spatial resolution remote sensing (RS) images. RS image information extraction includes feature classification, which is a long-standing research issue in the RS community. Emerging deep learning techniques, such as the deep semantic segmentation network technique, are effective methods to automatically discover relevant contextual features and get better image classification results. In this study, we exploited deep semantic segmentation networks to classify and extract CA from high-resolution RS images. WorldView-2 (WV-2) images with only Red-Green-Blue (RGB) bands were used to confirm the effectiveness of the proposed semantic classification framework for information extraction and the CA mapping task. Specifically, we used the deep learning framework TensorFlow to construct a platform for sampling, training, testing, and classifying to extract and map CA on the basis of DeepLabv3+. By leveraging per-pixel and random sample point accuracy evaluation methods, we conclude that the proposed approach can efficiently obtain acceptable accuracy (Overall Accuracy = 95%, Kappa = 0.90) of CA classification in the study area, and the approach performs better than other deep semantic segmentation networks (U-Net/PspNet/SegNet/DeepLabv2) and traditional machine learning methods, such as Maximum Likelihood (ML), Support Vector Machine (SVM), and RF (Random Forest). Furthermore, the proposed approach is highly scalable for the variety of crop types in a crop area. Overall, the proposed approach can train a precise and effective model that is capable of adequately describing the small, irregular fields of smallholder agriculture and handling the great level of details in RGB high spatial resolution images. Full article

This study proposes an effective method to map rice crops using the Sentinel-1 SAR (Synthetic Aperture Radar) time series over the Camargue region, Southern France. First, the temporal behavior of the SAR backscattering coefficient over 832 plots containing different crop types was analyzed. Through this analysis, the rice cultivation was identified using metrics derived from the Gaussian profile of the VV/VH time series (3 metrics), the variance of the VV/VH time series (one metric), and the slope of the linear regression of the VH time series (one metric). Using the derived metrics, rice plots were mapped through two different approaches: decision tree and Random Forest (RF). To validate the accuracy of each approach, the classified rice map was compared to the available national data. Similar high overall accuracy was obtained using both approaches. The overall accuracy obtained using a simple decision tree reached 96.3%, whereas an overall accuracy of 96.6% was obtained using the RF classifier. The approach, therefore, provides a simple yet precise and powerful tool to map paddy rice areas. Full article

This work presents a detailed analysis of building damage recognition, employing multi-source data fusion and ensemble learning algorithms for rapid damage mapping tasks. A damage classification framework is introduced and tested to categorize the building damage following the recent 2018 Sulawesi earthquake and tsunami. Three robust ensemble learning classifiers were investigated for recognizing building damage from Synthetic Aperture Radar (SAR) and optical remote sensing datasets and their derived features. The contribution of each feature dataset was also explored, considering different combinations of sensors as well as their temporal information. SAR scenes acquired by the ALOS-2 PALSAR-2 and Sentinel-1 sensors were used. The optical Sentinel-2 and PlanetScope sensors were also included in this study. A non-local filter in the preprocessing phase was used to enhance the SAR features. Our results demonstrated that the canonical correlation forests classifier performs better in comparison to the other classifiers. In the data fusion analysis, Digital Elevation Model (DEM)- and SAR-derived features contributed the most in the overall damage classification. Our proposed mapping framework successfully classifies four levels of building damage (with overall accuracy >90%, average accuracy >67%). The proposed framework learned the damage patterns from a limited available human-interpreted building damage annotation and expands this information to map a larger affected area. This process including pre- and post-processing phases were completed in about 3 h after acquiring all raw datasets. Full article

Population growth, economic development and rural-urban migration have caused rapid expansion of urban areas and metropolitan regions in Turkey. The structure of urban administration and planning has faced different socio-economic and political challenges, which have hindered the structured and planned development of cities and regions, resulting in an irregular and uneven development of these regions. We conducted detailed comparative analysis on spatio-temporal changes of the identified seven land-use/cover classes across different regions in Turkey with the use of Corine Land Cover (CLC) data of circa 1990, 2000, 2006 and 2012, integrated with Geographic Information System (GIS) techniques. Here we compared spatio-temporal changes of urban and non-urban land uses, which differ across regions and across different hierarchical levels of urban areas. Our findings have shown that peri-urban areas are growing more than rural areas, and even growing more than urban areas in some regions. A deeper look at regions located in different geographical zones pointed to substantial development disparities across western and eastern regions of Turkey. We also employed multiple regression models to explain any possible drivers of land-use change, regarding both urban and non-urban land uses. The results reveal that the three influencing factors-socio-economic characteristics, regional characteristics and location, and development constraints, facilitate land-use change. However, their impacts differ in different geographical locations, as well as with different hierarchical levels. Full article

Jointly using spectral and spatial information has become a mainstream strategy in the field of hyperspectral image (HSI) processing, especially for classification. However, due to the existence of noisy or correlated spectral bands in the spectral domain and inhomogeneous pixels in the spatial neighborhood, HSI classification results are often degraded and unsatisfactory. Motivated by the attention mechanism, this paper proposes a spatial–spectral squeeze-and-excitation (SSSE) module to adaptively learn the weights for different spectral bands and for different neighboring pixels. The SSSE structure can suppress or motivate features at a certain position, which can effectively resist noise interference and improve the classification results. Furthermore, we embed several SSSE modules into a residual network architecture and generate an SSSE-based residual network (SSSERN) model for HSI classification. The proposed SSSERN method is compared with several existing deep learning networks on two benchmark hyperspectral data sets. Experimental results demonstrate the effectiveness of our proposed network. Full article

Nowadays, 3-D convolutional neural networks (3-D CNN) have attracted lots of attention in the spectral-spatial classification of hyperspectral imageries (HSI). In this model, the feed-forward processing structure reduces the computational burden of 3-D structural processing. However, this model as a vector-based methodology cannot analyze the full content of the HSI information, and as a result, its features are not quite discriminative. On the other hand, convolutional long short-term memory (CLSTM) can recurrently analyze the 3-D structural data to extract more discriminative and abstract features. However, the computational burden of this model as a sequence-based methodology is extremely high. In the meanwhile, the robust spectral-spatial feature extraction with a reasonable computational burden is of great interest in HSI classification. For this purpose, a two-stage method based on the integration of CNN and CLSTM is proposed. In the first stage, 3-D CNN is applied to extract low-dimensional shallow spectral-spatial features from HSI, where information on the spatial features are less than that of the spectral information; consequently, in the second stage, the CLSTM, for the first time, is applied to recurrently analyze the spatial information while considering the spectral one. The experimental results obtained from three widely used HSI datasets indicate that the application of the recurrent analysis for spatial feature extractions makes the proposed model robust against different spatial sizes of the extracted patches. Moreover, applying the 3-D CNN prior to the CLSTM efficiently reduces the model’s computational burden. The experimental results also indicated that the proposed model led to a 1% to 2% improvement compared to its counterpart models. Full article

Methane (${\mathrm{CH}}_4$) is a potent greenhouse gas with a large temporal variability. To increase the spatial coverage, methane observations are increasingly made from satellites that retrieve the column-averaged dry air mole fraction of methane (${\mathrm{XCH}}_4$). To understand and quantify the spatial differences of the seasonal cycle and trend of ${\mathrm{XCH}}_4$ in more detail, and to ultimately help reduce uncertainties in methane emissions and sinks, we evaluated and analyzed the average ${\mathrm{XCH}}_4$ seasonal cycle and trend from three Greenhouse Gases Observing Satellite (GOSAT) retrieval algorithms: National Institute for Environmental Studies algorithm version 02.75, RemoTeC ${\mathrm{CH}}_4$ Proxy algorithm version 2.3.8 and RemoTeC ${\mathrm{CH}}_4$ Full Physics algorithm version 2.3.8. Evaluations were made against the Total Carbon Column Observing Network (TCCON) retrievals at 15 TCCON sites for 2009–2015, and the analysis was performed, in addition to the TCCON sites, at 31 latitude bands between latitudes 44.43°S and 53.13°N. At latitude bands, we also compared the trend of GOSAT ${\mathrm{XCH}}_4$ retrievals to the NOAA’s Marine Boundary Layer reference data. The average seasonal cycle and the non-linear trend were, for the first time for methane, modeled with a dynamic regression method called Dynamic Linear Model that quantifies the trend and the seasonal cycle, and provides reliable uncertainties for the parameters. Our results show that, if the number of co-located soundings is sufficiently large throughout the year, the seasonal cycle and trend of the three GOSAT retrievals agree well, mostly within the uncertainty ranges, with the TCCON retrievals. Especially estimates of the maximum day of ${\mathrm{XCH}}_4$ agree well, both between the GOSAT and TCCON retrievals, and between the three GOSAT retrievals at the latitude bands. In our analysis, we showed that there are large spatial differences in the trend and seasonal cycle of ${\mathrm{XCH}}_4$. These differences are linked to the regional ${\mathrm{CH}}_4$ sources and sinks, and call for further research. Full article

Spaceborne radar altimeters record echo waveforms over all Earth surfaces, but their interpretation and quantitative exploitation over the Arctic Ocean is particularly challenging. Radar returns may be from all ocean, all sea ice, or a mixture of the two, so the first task is the determination of which surface and then an interpretation of the signal to give range. Subsequently, corrections have to be applied for various surface and atmospheric effects before making a comparison with a reference level. This paper discusses the drivers for improved altimetry in the Arctic and then reviews the various approaches that have been used to achieve the initial classification and subsequent retracking over these diverse surfaces, showing examples from both LRM (low resolution mode) and SAR (synthetic aperture radar) altimeters. The review then discusses the issues concerning corrections, including the choices between using other remote-sensing measurements and using those from models or climatology. The paper finishes with some perspectives on future developments, incorporating secondary frequency, interferometric SAR and opportunities for fusion with measurements from laser altimetry or from the SMOS salinity sensor, and provides a full list of relevant abbreviations. Full article

One of the intrinsic properties of conventional terrestrial laser scanning technology is the unevenness of its point density over the scene where objects rendered closer to the scanner are more densely covered than the ones far away. This uneven distribution can be amplified as the working range of a laser scanner gets longer. In such case a higher pulse repetition rate (PRR) is applied to the whole scanning area and the scanning time will be dramatically increased. To improve the efficiency of the conventional laser scanning technology, a prototype of adaptive scanning technology, the HRS3D-AS scanner has been developed by Blackmore Sensors and Analytics, Inc. This paper briefly describes the working principles of the adaptive scanner and presents a thorough evaluation on the distributions of the point density in comparison to the conventional scanning. Based on this study, we show that such a new technology can produce a point cloud of more uniform density and less data volume. The overall field scanning time can be reduced by several times compared to the conventional, PRR-fixed scanning. Such properties are expected to significantly simplify the algorithmic development and increase the productivity in data acquisition and processing. The limitations of this new adaptive scanning technology are also discussed in terms of redundant and unresolved details. Finally, recommendations related to the practicing of such adaptive scan are discussed. Full article

Recent flood events have demonstrated a demand for satellite-based inundation mapping in near real-time (NRT). Simulating and forecasting flood extent is essential for risk mitigation. While numerical models are designed to provide such information, they usually lack reference at fine spatiotemporal resolution. Remote sensing techniques are expected to fill this void. Unlike optical sensors, synthetic aperture radar (SAR) provides valid measurements through cloud cover with high resolution and increasing sampling frequency from multiple missions. This study reviews theories and algorithms of flood inundation mapping using SAR data, together with a discussion of their strengths and limitations, focusing on the level of automation, robustness, and accuracy. We find that the automation and robustness of non-obstructed inundation mapping have been achieved in this era of big earth observation (EO) data with acceptable accuracy. They are not yet satisfactory, however, for the detection of beneath-vegetation flood mapping using L-band or multi-polarized (dual or fully) SAR data or for urban flood detection using fine-resolution SAR and ancillary building and topographic data. Full article

WorldView-3 satellite is providing images with an unprecedented combination of high spatial and spectral resolution. The stereo capabilities and the very high resolution of the panchromatic band (0.31 m) have been fostering new applications in urban areas, where the complexity of the morphology requires a higher level of detail. The present technical note aims to test the accuracy of digital elevation models that can be obtained by WorldView-3 stereo-pairs in these particular contexts, with an operational state-of-the-art algorithm. Validation is performed using check points and existing models of the area (from LiDAR data and oblique aerial images). The experiments, conducted over the city of Bologna (Italy) with six images, proved that roof surfaces and open spaces can be reconstructed with an average error of 1–2 pixels, but severe discrepancies frequently occur in narrow roads and urban canyons (up to several metres in average). The level of completeness achievable with only one pair is extremely variable (ranging from 50% to 90%), due to the combined effect of the geometry of acquisition and the specific urban texture. Better results can be obtained by using more than one pair. Furthermore, smaller convergence angles can be beneficial for the reconstruction of specific urban structures, such as soaring towers. Full article

Pan-sharpening (PS) is a method of fusing the spatial details of a high-resolution panchromatic (PAN) image with the spectral information of a low-resolution multi-spectral (MS) image. Visual inspection is a crucial step in the evaluation of fused products whose subjectivity renders the assessment of pansharpened data a challenging problem. Most previous research on the development of PS algorithms has only superficially addressed the issue of qualitative evaluation, generally by depicting visual representations of the fused images. Hence, it is highly desirable to be able to predict pan-sharpened image quality automatically and accurately, as it would be perceived and reported by human viewers. Such a method is indispensable for the correct evaluation of PS techniques that produce images for visual applications such as Google Earth and Microsoft Bing. Here, we propose a new image quality assessment (IQA) measure that supports the visual qualitative analysis of pansharpened outcomes by using the statistics of natural images, commonly referred to as natural scene statistics (NSS), to extract statistical regularities from PS images. Importantly, NSS are measurably modified by the presence of distortions. We analyze six PS methods in the presence of two common distortions, blur and white noise, on PAN images. Furthermore, we conducted a human study on the subjective quality of pristine and degraded PS images and created a completely blind (opinion-unaware) fused image quality analyzer. In addition, we propose an opinion-aware fused image quality analyzer, whose predictions with respect to human perceptual evaluations of pansharpened images are highly correlated. Full article

The TanDEM-X (TDX) mission launched by the German Aerospace Center delivers unprecedented global coverage of a high-quality digital elevation model (DEM) with a pixel spacing of 12 m. To examine the relationships of terrain, vegetation, and building elevations with hydrologic, geologic, geomorphologic, or ecologic factors, quantification of TDX DEM errors at a local scale is necessary. We estimated the errors of TDX data for open ground, forested, and built areas in a coastal urban environment by comparing the TDX DEM with LiDAR data for the same areas, using a series of error measures including root mean square error (RMSE) and absolute deviation at the 90% quantile (LE90). RMSE and LE90 values were 0.49 m and 0.79 m, respectively, for open ground. These values, which are much lower than the 10 m LE90 specified for the TDX DEM, highlight the promise of TDX DEM data for mapping hydrologic and geomorphic features in coastal areas. The RMSE/LE90 values for mangrove forest, tropical hardwood hammock forest, pine forest, dense residential, sparse residential, and downtown areas were 1.15/1.75, 2.28/3.37, 3.16/5.00, 1.89/2.90, 2.62/4.29 and 35.70/51.67 m, respectively. Regression analysis indicated that variation in canopy height of densely forested mangrove and hardwood hammock was well represented by the TDX DEM. Thus, TDX DEM data can be used to estimate tree height in densely vegetated forest on nearly flat topography next to the shoreline. TDX DEM errors for pine forest and residential areas were larger because of multiple reflection and shadow effects. Furthermore, the TDX DEM failed to capture the many high-rise buildings in downtown, resulting in the lowest accuracy among the different land cover types. Therefore, caution should be exercised in using TDX DEM data to reconstruct building models in a highly developed metropolitan area with many tall buildings separated by narrow open spaces. Full article

Pseudo invariant calibration sites (PICS) have been extensively used for the radiometric calibration and temporal stability monitoring of optical satellite sensors. Due to limited knowledge about the radiometric stability of North Africa, only a limited number of sites in the region are used for this purpose. This work presents an automated approach to classify North Africa for its potential use as an extended PICS (EPICS) covering vast portions of the continent. An unsupervised classification algorithm identified 19 “clusters” representing distinct land surface types was used; three clusters were identified with spatial uncertainties within approximately 5% in the shorter wavelength bands and 3% in the longer wavelength bands. A key advantage of the cluster approach is that large numbers of pixels are aggregated into contiguous homogeneous regions sufficiently distributed across the continent to allow multiple imaging opportunities per day, as opposed to imaging a typical PICS once during the sensor’s revisit period. This potential increase in temporal resolution could result in increased sensitivity for the quicker identification of changes in sensor response. Full article

Remotely sensed vegetation indices (RSVIs) can be used to efficiently estimate terrestrial primary productivity across space and time. Terrestrial productivity, however, has many facets (e.g., spatial and temporal variability, including seasonality, interannual variability, and trends), and different vegetation indices may not be equally good at predicting them. Their accuracy in monitoring productivity has been mostly tested in single-ecosystem studies, but their performance in different ecosystems distributed over large areas still needs to be fully explored. To fill this gap, we identified the facets of terrestrial gross primary production (GPP) that could be monitored using RSVIs. We compared the temporal and spatial patterns of four vegetation indices (NDVI, EVI, NIR_V, and CCI), derived from the MODIS MAIAC data set and of GPP derived from data from 58 eddy-flux towers in eight ecosystems with different plant functional types (evergreen needle-leaved forest, evergreen broad-leaved forest, deciduous broad-leaved forest, mixed forest, open shrubland, grassland, cropland, and wetland) distributed throughout Europe, covering Mediterranean, temperate, and boreal regions. The RSVIs monitored temporal variability well in most of the ecosystem types, with grasslands and evergreen broad-leaved forests most strongly and weakly correlated with weekly and monthly RSVI data, respectively. The performance of the RSVIs monitoring temporal variability decreased sharply, however, when the seasonal component of the time series was removed, suggesting that the seasonal cycles of both the GPP and RSVI time series were the dominant drivers of their relationships. Removing winter values from the analyses did not affect the results. NDVI and CCI identified the spatial variability of average annual GPP, and all RSVIs identified GPP seasonality well. The RSVI estimates, however, could not estimate the interannual variability of GPP across sites or monitor the trends of GPP. Overall, our results indicate that RSVIs are suitable to track different facets of GPP variability at the local scale, therefore they are reliable sources of GPP monitoring at larger geographical scales. Full article

The main objective of this work was to study the feasibility of using the green red vegetation index (GRVI) and the red edge ratio (RE/R) obtained from UAS imagery for monitoring the effects of soil water deficit and for predicting fibre quality in a surface-irrigated cotton crop. The performance of these indices to track the effects of water stress on cotton was compared to that of the normalised difference vegetation index (NDVI) and crop water stress index (CWSI). The study was conducted during two consecutive seasons on a commercial farm where three irrigation frequencies and two nitrogen rates were being tested. High-resolution multispectral images of the site were acquired on four dates in 2017 and six dates in 2018, encompassing a range of matric potential values. Leaf stomatal conductance was also measured at the image acquisition times. At harvest, lint yield and fibre quality (micronaire) were determined for each treatment. Results showed that within each year, the N rates tested (> 180 kg N ha⁻¹) did not have a statistically significant effect on the spectral indices. Larger intervals between irrigations in the less frequently irrigated treatments led to an increase (p < 0.05) in the CWSI and a reduction (p < 0.05) in the GRVI, RE/R, and to a lesser extent in the NDVI. A statistically significant and good correlation was observed between the GRVI and RE/R with soil matric potential and stomatal conductance at specific dates. The GRVI and RE/R were in accordance with the soil and plant water status when plants experienced a mild level of water stress. In most of the cases, the GRVI and RE/R displayed long-term effects of the water stress on plants, thus hampering their use for determinations of the actual soil and plant water status. The NDVI was a better predictor of lint yield than the GRVI and RE/R. However, both GRVI and RE/R correlated well (p < 0.01) with micronaire in both years of study and were better predictors of micronaire than the NDVI. This research presents the GRVI and RE/R as good predictors of fibre quality with potential to be used from satellite platforms. This would provide cotton producers the possibility of designing specific harvesting plans in the case that large fibre quality variability was expected to avoid discount prices. Further research is needed to evaluate the capability of these indices obtained from satellite platforms and to study whether these results obtained for cotton can be extrapolated to other crops. Full article

Knowledge of the aboveground biomass (AGB) of large pasture fields is invaluable as it assists graziers to set stocking rate. In this preliminary evaluation, we investigated the response of Sentinel-1 (S1) Synthetic Aperture Radar (SAR) data to biophysical variables (leaf area index, height and AGB) for native pasture grasses on a hilly, pastoral farm. The S1 polarimetric parameters such as backscattering coefficients, scattering entropy, scattering anisotropy, and mean scattering angle were regressed against the widely used morphological parameters of leaf area index (LAI) and height, as well as AGB of pasture grasses. We found S1 data to be more responsive to the pasture parameters when using a 1 m digital elevation model (DEM) to orthorectify the SAR image than when we employed the often-used Shuttle Radar Topography 30 m and 90 m Missions. With the 1m DEM analysis, a significant quadratic relationship was observed between AGB and VH cross-polarisation (R² = 0.71), and significant exponential relationships between polarimetric entropy and LAI and AGB (R² = 0.53 and 0.45, respectively). Similarly, the mean scattering angle showed a significant exponential relationship with LAI and AGB (R² = 0.58 and R² = 0.83, respectively). The study also found a significant quadratic relationship between the mean scattering angle and pasture height (R² = 0.72). Despite a relatively small dataset and single season, the mean scattering angle in conjunction with a generalised additive model (GAM) explained 73% of variance in the AGB estimates. The GAM model estimated AGB with a root mean square error of 392 kg/ha over a range in pasture AGB of 443 kg/ha to 2642 kg/ha with pasture LAI ranging from 0.27 to 1.87 and height 3.25 cm to 13.75 cm. These performance metrics, while indicative at best owing to the limited datasets used, are nonetheless encouraging in terms of the application of S1 data to evaluating pasture parameters under conditions which may preclude use of traditional optical remote sensing systems. Full article

Surface albedo is an important physical attribute of the climate system and satellite retrievals are useful for understanding how it varies in time and space. Surface albedo is sensitive to land cover and structure, which can vary considerably within the area comprising the effective spatial resolution of the satellite-based retrieval. This is particularly true for MODIS products and for topographically complex regions, such as Norway, which makes it difficult to separate the environmental drivers (e.g., temperature and snow) from those related to land cover and vegetation structure. In the present study, we employ high resolution datasets of Norwegian land cover and structure to spectrally unmix MODIS surface albedo retrievals (MCD43A3 v6) to study how surface albedo varies with land cover and structure. Such insights are useful for constraining land cover-dependent albedo parameterizations in models employed for regional climate or hydrological research and for developing new empirical models. At the scale of individual land cover types, we found that the monthly surface albedo can be predicted at a high accuracy when given additional information about forest structure, snow cover, and near surface air temperature. Such predictions can provide useful empirical benchmarks for climate model predictions made at the land cover level, which is critical for instilling greater confidence in the albedo-related climate impacts of anthropogenic land use/land cover change (LULCC). Full article

Remote sensing techniques provide data on the spatial–temporal distribution of environmental parameters over regions with sparse ground observations. However, the resolution of satellite precipitation data is too coarse to be applied to hydrological and meteorological research at basin scales. Downscaling research using coarse remote sensing data to obtain high-resolution precipitation data is significant for the development of basin-scale research. Here, we propose improvements to a spatial–temporal method for downscaling satellite precipitation. The improved method uses a nonlinear regression model and introduces longitude and latitude based on processed normalized difference vegetation index (NDVI) and a digital elevation model (DEM) to stimulate precipitation in the Qilian Mountains during 2006–2015. The final downscaled annual precipitation (FDAP) results are corrected by observed data to obtain corrected final downscaled annual precipitation (CFDAP) datasets. For temporal downscaling, monthly downscaled data are the corrected monthly ratio multiplied by the corresponding downscaled annual datasets. The results indicated that processed NDVI (PNDVI) reflected spatial precipitation patterns more accurately than the original NDVI. The accuracy was significantly improved when the final downscaled annual precipitation data were corrected by observed data. The average annual root mean square error (RMSE) from 2006 to 2015 of CFDAP was 66.48 and 83.07 mm less than that of FDAP and original Tropical Rainfall Measuring Mission (TRMM) data, respectively. Compared with previous methods, which use NDVI and/or DEM to downscale TRMM, the accuracy of FDAP and CFDAP from the improved method was higher, and the RMSE decreased on average by 13.63 and 80.11 mm. The RMSE of monthly data from corrected monthly ratio (CMR) decreased on average by 4.93 mm over monthly data from previous monthly ratio (PMR). In addition, the accuracy of the original satellite data affected the initial downscaling results but had no significant effects on the corrected downscaling results. Full article

Remote sensing archaeology in recent years has emphasized the use of high-precision and high-accuracy tools to achieve the detailed documentation of archaeological elements (drones, LIDAR, etc.). Satellite remote sensing has also benefited from an increase in the spatial and spectral resolution of the sensors, which is enabling the discovery and documentation of new archaeological features and sites worldwide. While there can be no doubt that a great deal is being gained via such “site detection” approaches, there still remains the possibility of further exploring remote sensing methods to analyse archaeological problems. In this paper, this issue is discussed by focusing on one common archaeological topic: the mapping of environmental resources used in the past and, in particular, the procurement of lithic raw material by hunter-gatherer groups. This is illustrated by showing how the combined use of Landsat 8 images and “ground-truthing” via focused field studies has allowed the identification of a number of potential chert sources, the major lithic resource used by coastal groups between 11,500–1,500 cal. BP, in a vast area of the Atacama Desert covering 22,500 km². Besides discussing the case study, the strength of remote sensing techniques in addressing archaeological questions comprising large spatial scales is highlighted, stressing the key role they can play in the detection and study of specific environmental resources within challenging physical settings. Full article

An in situ soil moisture observation network at pixel scale is constructed in cropland in the northeast of China for accurate regional soil moisture evaluations of satellite products. The soil moisture products are based on the Japan Aerospace Exploration Agency (JAXA) algorithm and the Land Parameter Retrieval Model (LPRM) from the Advanced Microwave Scanning Radiometer 2 (AMSR2), and the products from the FengYun-3B (FY3B) satellite are evaluated using synchronous in situ data collected by the EC-5 sensors at the surface in a typical cropland in the northeast of China during the crop-growing season from May to September 2017. The results show that the JAXA product provides an underestimation with a bias (b) of -0.094 cm³/cm³, and the LPRM soil moisture product generates an overestimation with a b of 0.156 cm³/cm³. However the LPRM product shows a better correlation with the in situ data, especially in the early experimental period when the correlation coefficient is 0.654, which means only the JAXA product in the early stage, with an unbiased root mean square error (ubRMSE) of 0.049 cm³/cm³ and a b of -0.043 cm³/cm³, reaches the goal accuracy (±0.05 cm³/cm³). The FY3B has consistently obtained microwave brightness temperature data, but its soil moisture product data in the study area is seriously missing during most of the experimental period. However, it recovers in the later period and is closer to the in situ data than the JAXA and LPRM products. The three products show totally different trends with vegetation cover, soil temperature, and actual soil moisture itself in different time periods. The LPRM product is more sensitive and correlated with the in situ data, and is less susceptible to interferences. The JAXA is numerically closer to the in situ data, but the results are still affected by temperature. Both will decrease in accuracy as the actual soil moisture increases. The FY3B seems to perform better at the end of the whole period after data recovery. Full article

Measurement campaigns in wind energy research are becoming increasingly complex, which has exacerbated the difficulty of taking optimal measurements using light detection and ranging (LiDAR) systems. Compromises between spatial and temporal resolutions are always necessary in the study of heterogeneous flows, like wind turbine wakes. Below, we develop a method for space-time conversion that acts as a temporal fluid-dynamic interpolation without incurring the immense computing costs of a 4D flow solver. We tested this space-time conversion with synthetic LiDAR data extracted from a large-eddy-simulation (LES) of a neutrally stable single-turbine wake field. The data was synthesised with a numerical LiDAR simulator. Then, we performed a parametric study of 11 different scanning velocities. We found that temporal error dominates the mapping error at low scanning speeds and that spatial error becomes dominant at fast scanning speeds. Our space-time conversion method increases the temporal resolution of the LiDAR data by a factor 2.4 to 40 to correct the scan-containing temporal shift and to synchronise the scan with the time code of the LES data. The mean-value error of the test case is reduced to a minimum relative error of 0.13% and the standard-deviation error is reduced to a minimum of 0.6% when the optimal scanning velocity is used. When working with the original unprocessed LiDAR measurements, the space-time-conversion yielded a maximal error reduction of 69% in the mean value and 58% in the standard deviation with the parameters identified with our analysis. Full article

Understanding the causes of inter-satellite biases in climate data records from observations of the Earth is crucial for constructing a consistent time series of the essential climate variables. In this article, we analyse the strong scan- and time-dependent biases observed for the microwave humidity sounders on board the NOAA-16 and NOAA-19 satellites. We find compelling evidence that radio frequency interference (RFI) is the cause of the biases. We also devise a correction scheme for the raw count signals for the instruments to mitigate the effect of RFI. Our results show that the RFI-corrected, recalibrated data exhibit distinctly reduced biases and provide consistent time series. Full article

The general topic here is the application of very high-resolution satellite imagery to the study of ocean phenomena having horizontal spatial scales of the order of 1 kilometer, which is the realm of the ocean submesoscale. The focus of the present study is the use of WorldView-2 along-track stereo imagery to probe a submesoscale feature in the Baltic Sea that consists of an apparent inward spiraling of surface aggregations of algae. In this case, a single pair of images is analyzed using an optical-flow velocity algorithm. Because such image data generally have a much lower dynamic range than in land applications, the impact of residual instrument noise (e.g., data striping) is more severe and requires attention; we use a simple scheme to reduce the impact of such noise. The results show that the spiral feature has at its core a cyclonic vortex, about 1 km in radius and having a vertical vorticity of about three times the Coriolis frequency. Analysis also reveals that an individual algal aggregation corresponds to a velocity front having both horizontal shear and convergence, while wind-accelerated clumps of surface algae can introduce fine-scale signatures into the velocity field. Overall, the analysis supports the interpretation of algal spirals as evidence of a submesoscale eddy and of algal aggregations as indicating areas of surface convergence. Full article

This paper proposes novel autoencoders for unsupervised feature-learning from hyperspectral data. Hyperspectral data typically have many dimensions and a significant amount of variability such that many data points are required to represent the distribution of the data. This poses challenges for higher-level algorithms which use the hyperspectral data (e.g., those that map the environment). Feature-learning mitigates this by projecting the data into a lower-dimensional space where the important information is either preserved or enhanced. In many applications, the amount of labelled hyperspectral data that can be acquired is limited. Hence, there is a need for feature-learning algorithms to be unsupervised. This work proposes unsupervised techniques that incorporate spectral measures from the remote-sensing literature into the objective functions of autoencoder feature learners. The proposed techniques are evaluated on the separability of their feature spaces as well as on their application as features for a clustering task, where they are compared against other unsupervised feature-learning approaches on several different datasets. The results show that autoencoders using spectral measures outperform those using the standard squared-error objective function for unsupervised hyperspectral feature-learning. Full article

Underestimation of LiDAR heights is widely known but has never been evaluated for several sensors and for diverse types of ecological conditions. This underestimation is mainly linked to the probability of the pulse to reach the ground and the top of vegetation. Main causes of this underestimation are pulse density, pattern of scan (sensors), scan angles, specific contract parameters (flying altitude, pulse repetition frequency) and characteristics of the territory (slopes, stand density and species composition). This study, carried out at a resolution of 1 × 1 m, first assessed the possibility of making an adjustment model to correct the bias of the digital terrain model (DTM), and then proposed a global adjustment model to correct the bias on the canopy height model (CHM). For this study, the bias of both DTM and CHM were calculated by subtracting two LiDAR datasets: high-density pixels with 21 pulses/m² (first return) and more (DTM or CHM reference value pixels) and low-density pixels (DTM or CHM value to correct). After preliminary analyses, it was concluded that the DTM did not need specific adjustment. In contrast, the CHM needed adjustments. Among the variables studied, three were selected for the final CHM adjustment model: the maximum height of the pixel (H2Corr); the density of first returns by m² (D_first); and the standard deviation of nine maximum heights of the neighborhood cells (H_STD9). The modeling occurred in three steps. The first two steps enabled the determination of significant variables and the shape of the equation to be defined (linear mixed model and non-linear model). The third step made it possible to propose an empirical equation using a non-linear mixed model that can be applied to a 1 × 1 m CHM. The CHM underestimation correction could be used for a preliminary step to several uses of the CHM such as volume calculation, forest growth models or multi-temporal analysis. Full article

With the warming of the global climate, the mass loss of the Greenland ice sheet is intensifying, having a profound impact on the rising of the global sea level. Here, we used Gravity Recovery and Climate Experiment (GRACE) RL06 data to retrieve the time series variations of ice sheet mass in Greenland from January 2003 to December 2015. Meanwhile, the spatial changes of ice sheet mass and its relationship with land surface temperature are studied by means of Theil–Sen median trend analysis, the Mann–Kendall (MK) test, empirical orthogonal function (EOF) analysis, and wavelet transform analysis. The results showed: (1) in terms of time, we found that the total mass of ice sheet decreases steadily at a speed of −195 ± 21 Gt/yr and an acceleration of −11 ± 2 Gt/yr² from 2003 to 2015. This mass loss was relatively stable in the two years after 2012, and then continued a decreasing trend; (2) in terms of space, the mass loss areas of the Greenland ice sheet mainly concentrates in the southeastern, southwestern, and northwestern regions, and the southeastern region mass losses have a maximum rate of more than 27 cm/yr (equivalent water height), while the northeastern region show a minimum rate of less than 3 cm/yr, showing significant changes as a whole. In addition, using spatial distribution and the time coefficients of the first two models obtained by EOF decomposition, ice sheet quality in the southeastern and northwestern regions of Greenland show different significant changes in different periods from 2003 to 2015, while the other regions showed relatively stable changes; (3) in terms of driving factors temperature, there is an anti-phase relationship between ice sheet mass change and land surface temperature by the mean XWT-based semblance value of −0.34 in a significant oscillation period variation of 12 months. Meanwhile, XWT-based semblance values have the largest relative change in 2005 and 2012, and the smallest relative change in 2009 and 2010, indicating that the influence of land surface temperature on ice sheet mass significantly varies in different years. Full article

More than 90% of the sugar production in China comes from sugarcane, which is widely grown in South China. Optical image time series have proven to be efficient for sugarcane mapping. There are, however, two limitations associated with previous research: one is that the critical observations during the sugarcane growing season are limited due to frequent cloudy weather in South China; the other is that the classification method requires imagery time series covering the entire growing season, which reduces the time efficiency. The Sentinel-1A (S1A) synthetic aperture radar (SAR) data featuring relatively high spatial-temporal resolution provides an ideal data source for all-weather observations. In this study, we attempted to develop a method for the early season mapping of sugarcane. First, we proposed a framework consisting of two procedures: initial sugarcane mapping using the S1A SAR imagery time series, followed by non-vegetation removal using Sentinel-2 optical imagery. Second, we tested the framework using an incremental classification strategy based on S1A imagery covering the entire 2017–2018 sugarcane season. The study area was in Suixi and Leizhou counties of Zhanjiang city, China. Results indicated that an acceptable accuracy, in terms of Kappa coefficient, can be achieved to a level above 0.902 using time series three months before sugarcane harvest. In general, sugarcane mapping utilizing the combination of VH + VV as well as VH polarization alone outperformed mapping using VV alone. Although the XGBoost classifier with VH + VV polarization achieved a maximum accuracy that was slightly lower than the random forest (RF) classifier, the XGBoost shows promising performance in that it was more robust to overfitting with noisy VV time series and the computation speed was 7.7 times faster than RF classifier. The total sugarcane areas in Suixi and Leizhou for the 2017–2018 harvest year estimated by this study were approximately 598.95 km² and 497.65 km², respectively. The relative accuracy of the total sugarcane mapping area was approximately 86.3%. Full article