Off-nadir Angles Research Articles

Comparison of forest canopy gap fraction measurements from drone-based video frames, below-canopy hemispherical photography, and airborne laser scanning

ABSTRACT The amount of gaps in forest canopy is related to the radiation interception for photosynthesis and visibility through the canopy. The dependence of forest canopy gap fraction determination on view zenith angle was calculated from polar-transformed sparse ( ≈ 4 m − 2 ) airborne laser scanning (ALS) point clouds for a Scots pine (Pinus sylvestris L.) stand growing on Kiriku Bog, Estonia. Visibility of ground targets was estimated from video image frames taken during drone (UAV) overpass at low altitude (40 m). Below-canopy digital hemispherical images (DHP) were taken in zenith direction as reference measurements. Angular grids of 3 ∘ and 5 ∘ were used to match the three data sources so as to decrease uncertainties in measurement geometries. The linear relationship between DHP and UAV data had R 2 = 0.67, with most of the deviations occurring at gap boundaries. Relationships over individual targets between DHP and polar-transformed ALS data had 0.3 < R 2 ≤ 0.8 . However, the simulation overestimated gap fraction at smaller zenith angles because of uncertainties in constructing lidar pulse footprints from point data. We conclude that regional coverage by means of sparse ALS point clouds shows potential for the assessment of forest canopy gaps at off-nadir angles.

Potential of the upcoming Biomass P-band radar mission for digital terrain modelling beneath dense tropical forests: first accuracy assessment

Abstract. P-band radar surveys can be used to produce digital elevation models (DEMs) at ground level in dense tropical forest areas. However, the previous performance studies were carried out on DEMs derived from high-resolution synthetic aperture radar (SAR) acquisitions, while the upcoming Biomass space mission should provide DEMs at a coarser resolution due to a frequency bandwidth of 6 MHz in range. The aim of this study is to predict the quality of the DEM products derived from Biomass data through dense forest. A DEM was produced in a slant range geometry at a 15-m mesh size using SAR tomography. It was compared to the reference LiDAR survey in terms of elevation and slope, and the statistical distribution of the corresponding error was analysed with regards to the radar off-nadir angle and to the local slope and aspect. The results show that the DEM elevation has no systematic error despite the presence of forest, and it has an RMS error of 3 meters. The DEM slope has a mean error of 2° and an RMS error of 6°. The behaviour of these errors for different terrain slopes and aspects will allow to predict the quality of Biomass DEMs on a variety of terrestrial landscapes.

Early Monitoring of Maize Northern Leaf Blight Using Vegetation Indices and Plant Traits from Multiangle Hyperspectral Data

Maize northern leaf blight (MNLB), characterized by a bottom-up progression, is a prevalent and damaging disease affecting maize growth. Early monitoring is crucial for timely interventions, thus mitigating yield losses. Hyperspectral remote sensing technology is an effective means of early crop disease monitoring. However, traditional single-angle vertical hyperspectral remote sensing methods face challenges in monitoring early MNLB in the lower part of maize canopy due to obstruction by upper canopy leaves. Therefore, we propose a multiangle hyperspectral remote sensing method for early MNLB monitoring. From multiangle hyperspectral data (−60° to 60°), we extracted and selected vegetation indices (VIs) and plant traits (PTs) that show significant differences between healthy and diseased maize samples. Our findings indicate that besides structural PTs (LAI and FIDF), other PTs like Cab, Car, Anth, Cw, Cp, and CBC show strong disease discrimination capabilities. Using these selected features, we developed a disease monitoring model with the random forest (RF) algorithm, integrating VIs and PTs (PTVI-RF). The results showed that PTVI-RF outperformed models based solely on VIs or PTs. For instance, the overall accuracy (OA) of the PTVI-RF model at 0° was 80%, which was 4% and 6% higher than models relying solely on VIs and PTs, respectively. Additionally, we explored the impact of viewing angles on model accuracy. The results show that compared to the accuracy at the nadir angle (0°), higher accuracy is obtained at smaller off-nadir angles (±10° to ±30°), while lower accuracy is obtained at larger angles (±40° to ±60°). Specifically, the OA of the PTVI-RF model ranges from 80% to 88% and the Kappa ranges from 0.6 to 0.76 at ±10° to ±30°, with the highest accuracy at −10° (OA = 88%, Kappa = 0.76). In contrast, the OA ranges from 72% to 80% and the Kappa ranges from 0.44 to 0.6 at ±40° to ±60°. In conclusion, this research demonstrates that PTVI-RF, constructed by fusing VIs and PTs extracted from multiangle hyperspectral data, can effectively monitor early MNLB. This provides a basis for the early prevention and control of MNLB and offers a valuable reference for early monitoring crop diseases with similar bottom-up progression.

Analysis of Cassini Altimetric Crossovers on Titan

The Cassini spacecraft performed several flybys of Saturn’s largest moon, Titan, collecting valuable data. During several passes, altimetric data were acquired. Here, we focus on altimetric measurements collected by Cassini’s radar when flying over the same region at different epochs in order to correlate such measurements (crossovers) and investigate differences in altimetry. In our study, we assess altimetric errors associated with three distinct methods for extracting topography from Cassini’s radar data: the maximum likelihood estimator (MLE), the threshold method, and the first moment technique. Focusing on crossover events, during which Cassini revisited specific areas of Titan’s surface, we conduct a detailed examination of the consistency and accuracy of these three topography extraction methods. The proposed analysis involves closely examining altimetric data collected at different epochs over identical geographical regions, allowing us to investigate potential errors due to the variations in off-nadir angle, relative impact, uncertainties, and systematic errors inherent in the application of these methodologies. Our findings reveal that the correction applied for the off-nadir angle to the threshold and first moment methods significantly reduces the dispersion in the delta difference at the crossover, resulting in a dispersion of the order of 60 m, even lower than what is achieved with the MLE (~70 m). Additionally, an effort is made to assess the potential of Cassini for estimating the tidal signal on Titan. Considering the altimetric errors identified in our study and the relatively low number of crossovers performed by Cassini, our assessment indicates that it is not feasible to accurately measure the tidal signal on Titan using the currently available standard altimetry data from Cassini. Our assessment regarding the accuracy of the Cassini altimeter provides valuable insights for future planetary exploration endeavors. Our study advances the understanding of Titan’s complex landscape and contributes to refining topographical models derived from Cassini’s altimetry observations. These insights not only enhance our knowledge of Saturn’s largest moon but also open prospects for Titan surface and interior exploration using radar systems.

Building Detection from SkySat Images with Transfer Learning: a Case Study over Ankara

AbstractThe detection and continuous updating of buildings in geodatabases has long been a major research area in geographic information science and is an important theme for national mapping agencies. Advancements in machine learning techniques, particularly state-of-the-art deep learning (DL) models, offer promising solutions for extracting and modeling building rooftops from images. However, tasks such as automatic labelling of learning data and the generalizability of models remain challenging. In this study, we assessed the sensor and geographic area adaptation capabilities of a pretrained DL model implemented in the ArcGIS environment using very-high-resolution (50 cm) SkySat imagery. The model was trained for digitizing building footprints via Mask R‑CNN with a ResNet50 backbone using aerial and satellite images from parts of the USA. Here, we utilized images from three different SkySat satellites with various acquisition dates and off-nadir angles and refined the pretrained model using small numbers of buildings as training data (5–53 buildings) over Ankara. We evaluated the buildings in areas with different characteristics, such as urban transformation, slums, regular, and obtained high accuracies with F‑1 scores of 0.92, 0.94, and 0.96 from SkySat 4, 7, and 17, respectively. The study findings showed that the DL model has high transfer learning capability for Ankara using only a few buildings and that the recent SkySat satellites demonstrate superior image quality.

Overlap correction function based on multi-angle measurements for an airborne direct-detection lidar for atmospheric sensing.

We estimate the overlap function (accounting for near-field effects) for an airborne nadir-mounted lidar, based on multi-angle measurements of an atmospheric scene obtained during two flights. For each atmospheric layer, a regression on the logarithm of the range-corrected signal versus the secant of the off-nadir angle allowed evaluation of the optical depth and the backscattering coefficient multiplied by the lidar constant. These quantities allow for computation of the lidar signal unaffected by the overlap effect, and then for determination of the overlap correction function. Its evolution over time can also help to detect changes in the alignment. The method is easy to implement as long as a scanning capability is available, and it can be applied in aerosol-free or aerosol-laden conditions, the requirement being a constant and horizontally homogeneous atmosphere during the measurements. For multichannel lidars, the method can be applied separately for each channel.

Digital Beamforming Features in Subaperture Antenna Arrays of Highly Informative Space Radars with Aperture Synthesis

It is shown that during digital beamforming in a subaperture antenna array of highly informative space radars with synthetic aperture, it is necessary to provide a procedure for equalizing the time delays of the subaperture output signals for transmission and reception. It has been established that losses in the signal-to-noise ratio on a radar image of a point target for a typical aperture size at an elevation angle of 0.7 m and a signal spectrum width of 600 MHz reach 2.1 and 7.5 dB for off-nadir angles of 21° and 50°, respectively. Recommendations are formulated for equalizing time delays for transmission and reception based on spectral processing. Simulation results are presented.

MCHA-Net: A multi-end composite higher-order attention network guided with hierarchical supervised signal for high-resolution remote sensing image change detection

Change detection (CD) is the main way to detect changes in the Earth’s surface features in a timely and accurate manner and to understand the interactions between humans and nature, CD is also an important scientific tool for supporting decision-making. Many convolution-based methods and Transformer-based methods have gained remarkable success in the field of CD with high-resolution remote sensing images (HRSIs) due to their powerful feature extraction capability and global information modeling capability, respectively. However, the diversity and complexity of HRSIs render CD methods constantly challenging, e.g., off-nadir angle imaging, seasonal turnover, and simultaneous changes in multiple feature types. Common convolution-based or Transformer-based encoding–decoding networks have a single way of data modeling and a low degree of feature fusion, resulting in poor applicability to different remote sensing data, poor recognition of different semantic targets, and limited accuracy. To improve the generalizability and detection accuracy of the network, we developed a composite higher-order attention network with multiple encoding paths named MCHA-Net. MCHA-Net has four encoding backbones, namely, the Siamese-learning path, Residual-learning path, and Transformer-learning path. Different encoding ways are equivalent to different thinking ends, and the integration can make the network have a stronger feature representation capability, forming a local–global-cross domain data modeling approach and making the network have powerful data sensing and mining capability. The decoding end aggregates the semantic information of each layer and integrates them into a unified linearized feature mapping module to achieve full modeling of the separability of information in the target domain. In addition, we propose a new higher-order attention mechanism to perform adaptive feature refinement for each layer of each encoding end, to guide the network in focusing on the targeted region and to guarantee the boundary integrity and internal compactness of the detection results. Moreover, we design a hierarchical network supervision schema that adds supervision signals at different feature abstraction levels to impose differentiated soft constraints on each layer of the network, ensuring high semantic consistency of features across layers and facilitating fast network fitting. Experimental results on three benchmark datasets (S2Looking, SVCD, and SYSU-CD) and one transfer application dataset (Google Dataset) show that MCHA-Net outperforms state-of-the-art methods in both visual interpretation and quantitative evaluation, and exhibits strong generalization and robustness against few-shot learning.

Convolutional Neural Networks with Mask Shift Loss for Improving Building Outlines Detection

In the development of supervised semantic segmentation models for building outline detection, a significant challenge is obtaining sufficient accurately labelled training data. Old building outline datasets captured against historic imagery often exist covering the same locations as new imagery. By training Deep Convolutional Neural Networks to recognise buildings in these areas we can detect new and changed buildings. Ideally, historic building outlines as training labels paired with new orthophotography could be used to train models to detect buildings in new imagery. However, historic imagery may be orthorectified to a different standard, or have different off-nadir angles resulting in an apparent displacement of buildings between historic and current imagery sets and in labelled buildings that do not align accurately with the building image. This displacement problem presents difficulties in training accurate models. This paper proposes a novel loss function that searches the prediction segmentation masks for the best matching crop to a central crop of the training segmentation mask, thus reducing error due to linear displacements between an image and mask. According to the Intersection-over-Union metric, a ground truth test set for the building outlines detection problem improved from 0.751 to 0.802 on the Christchurch Buildings dataset when using the mask shift loss versus binary cross-entropy.

An Efficient Algorithm for Infrared Earth Sensor with a Large Field of View

Infrared Earth sensors with large-field-of-view (FOV) cameras are widely used in low-Earth-orbit satellites. To improve the accuracy and speed of Earth sensors, an algorithm based on modified random sample consensus (RANSAC) and weighted total least squares (WTLS) is proposed. Firstly, the modified RANSAC with a pre-verification step was used to remove the noisy points efficiently. Then, the Earth's oblateness was taken into consideration and the Earth's horizon was projected onto a unit sphere as a three-dimensional (3D) curve. Finally, the TLS and WTLS were used to fit the projection of the Earth horizon. With the help of TLS and WTLS, the accuracy of the Earth sensor was greatly improved. Simulated images and on-orbit infrared images obtained via the satellite Tianping-2B were used to assess the performance of the algorithm. The experimental results demonstrate that the method outperforms RANSAC, M-estimator sample consensus (MLESAC), and Hough transformation in terms of speed. The accuracy of the algorithm for nadir estimation is approximately 0.04° (root-mean-square error) when Earth is fully visible and 0.16° when the off-nadir angle is 120°, which is a significant improvement upon other nadir estimation algorithms.

Comparison between Different TomoSAR Imaging Models for Airborne Platform Flying at Low Altitude

The classical planar-wavefront-based TomoSAR imaging model suffers from the problem that the effective integration interval is not enough to cover the target distribution region in the low-altitude airborne case. It will lead to a deterioration of the performance of tomogram reconstruction and inaccuracy of estimated scatterers. This paper reviews the exact and approximate forms of the aforementioned inaccurate model based on planar wavefront and points out the problem with the conventional model. To solve this problem, we propose spherical wavefront models with the exact form or an approximate form of the slant range formula. The estimated variable for the scatterer’s location is converted from elevation to off-nadir angle, and the effective integration interval has been extended. In addition, we explore relationships between the exact form of the conventional model and the exact form of the proposed model, and the relationship between the approximate form of the conventional model and the approximate form of the proposed model. This provides a basis for modifying the inversion algorithm that is designed based on the conventional model to adapt to the low-altitude airborne case. Eventually, through experiments based on simulated data and measured data, the imprecise reconstructions obtained with the conventional model are demonstrated, and the correctness of spherical wavefront models and the effectiveness of transformation between models are proved.

Multicriteria Accuracy Assessment of Digital Elevation Models (DEMs) Produced by Airborne P-Band Polarimetric SAR Tomography in Tropical Rainforests

The penetration capability of P-band radar waves through dense vegetation, along with the ability of tomography to separate the contributions of different layers in a vertical reflectivity profile, make P-band radar tomography a promising tool for digital terrain modeling in forested areas, specifically in dense tropical forests under which terrain topography remains poorly known. This paper aims to assess the overall quality of a digital terrain model (DTM) produced using tomographic processing of airborne P-band SAR imagery acquired during the TropiSAR campaign in French Guiana. Many quality descriptors are used to evaluate the quality of this DTM. Position and slope accuracies are computed based on a lidar DTM considered as the reference, and the impact of several parameters on these accuracies is studied, namely, slope, slope orientation, off-nadir angle and local incidence angle. The realism of the landforms is also studied according to geomorphological criteria. The results of this multicriteria accuracy assessment show the high potential of P-band SAR tomography in depicting the topography under forests, despite the intrinsic limitations related to the slant range geometry: the absolute elevation error is around 2 m; the slope is overestimated with an error of about 16°, mainly due to a processing artifact for which easy and direct solutions exist. Indeed, this error is equal to about 3° in flat artifact-free areas. These errors vary depending on the acquisition parameters and the local topography. The shapes are globally well preserved. These results are also discussed in the frame of the upcoming BIOMASS mission developed by the European Space Agency (ESA) and expected to be launched in 2024.

An evaluation of PlanetScope images for 3D reconstruction and change detection – experimental validations with case studies

ABSTRACT The PlanetScope satellite constellation has over 130 Dove satellites running 24/7, which collect weekly and even daily images globally at 3–5 m resolution. It has global data coverage and high temporal resolution, which constitute the most attractive features for medium-resolution 3D reconstruction and change detection in remote sensing applications. One shortfall of the PlanetScope images is that they are often captured at very small off-nadir angles to minimize relief differences for 2D time-series image analysis, which is not intended for classic stereo 3D reconstruction due to the very small base-to-height ratios of stereo pairs. However, considering the abundant PlanetScope images, the multi-view stereo 3D reconstruction approach leveraging a large number of images may drive the possibility of achieving more accurate 3D reconstruction, and consequently, 3D change detection on a global scale. In this paper, a multi-view stereo 3D reconstruction pipeline was adopted to comprehensively evaluate the 3D potential of PlanetScope images by performing accuracy analysis for both 3D reconstruction and change detection in semi-randomly selected regions with ground truth data. Three case studies using the PlanetScope images were performed: (1) a case study on multi-view stereo 3D reconstruction, (2) a case study on 3D change detection of buildings and trees, and (3) a case study on volumetric estimation for natural disaster monitoring. Our experiments showed that the PlanetScope images provided sufficient coverage for multi-view stereo 3D reconstruction given an area of interest. It could achieve a reasonably acceptable accuracy with root-mean-square errors of 4–6 m in our test regions and detect significant 3D changes. The capability of estimating the volumetric changes was also evaluated for the recent avalanche in Chamoli, India, and the estimated volume favorably matched the results from existing studies using data with higher resolution.

An Accurate Calculation of the Atmospheric Refraction Error of Optical Remote Sensing Images Based on the Fine-Layered Light Vector Method

Atmospheric refraction is an important factor that affects the positioning accuracy of optical high-resolution satellite imaging. Current calculation methods do not consider regional change characteristics in the atmosphere; the atmospheric division model is rough and does not conform to reality. This article studies atmospheric refraction in optical remote sensing satellite observations, provides a formula for calculating the strict refraction of a multilayer atmosphere, and shows a new method for calculating atmospheric refraction error. We use a ray-tracing method to calculate the light propagation path in the atmosphere and continuously layer the atmosphere according to the height difference of 100 m. Different from general calculation methods, we prove that a regional atmospheric model can calculate the atmospheric refraction index more accurately than using the empirical model. As the off-nadir angle gradually increases, the calculation results in this article are better than the current commonly used methods. We use WorldView-2 images of the Qinghai&#x2013;Tibet Plateau region in China for the experiments. When the off-nadir angle is less than 32&#x00B0;, the positioning accuracy improves by 6&#x2013;47&#x0025;. Compared with the standard atmospheric model, the regional atmospheric model improves positioning accuracy by 2&#x2013;18&#x0025;. This method reflects the continuous variation in the atmospheric refractive index with the vertical distribution of the atmosphere and amended regional meteorological conditions. Model errors caused by overly simple atmospheric division are avoided, andthe positioning accuracy of optical remote sensing images is increased.

S2Looking: A Satellite Side-Looking Dataset for Building Change Detection

Building-change detection underpins many important applications, especially in the military and crisis-management domains. Recent methods used for change detection have shifted towards deep learning, which depends on the quality of its training data. The assembly of large-scale annotated satellite imagery datasets is therefore essential for global building-change surveillance. Existing datasets almost exclusively offer near-nadir viewing angles. This limits the range of changes that can be detected. By offering larger observation ranges, the scroll imaging mode of optical satellites presents an opportunity to overcome this restriction. This paper therefore introduces S2Looking, a building-change-detection dataset that contains large-scale side-looking satellite images captured at various off-nadir angles. The dataset consists of 5000 bitemporal image pairs of rural areas and more than 65,920 annotated instances of changes throughout the world. The dataset can be used to train deep-learning-based change-detection algorithms. It expands upon existing datasets by providing (1) larger viewing angles; (2) large illumination variances; and (3) the added complexity of rural images. To facilitate the use of the dataset, a benchmark task has been established, and preliminary tests suggest that deep-learning algorithms find the dataset significantly more challenging than the closest-competing near-nadir dataset, LEVIR-CD+. S2Looking may therefore promote important advances in existing building-change-detection algorithms.

Multi-Hazard and Spatial Transferability of a CNN for Automated Building Damage Assessment

Automated classification of building damage in remote sensing images enables the rapid and spatially extensive assessment of the impact of natural hazards, thus speeding up emergency response efforts. Convolutional neural networks (CNNs) can reach good performance on such a task in experimental settings. How CNNs perform when applied under operational emergency conditions, with unseen data and time constraints, is not well studied. This study focuses on the applicability of a CNN-based model in such scenarios. We performed experiments on 13 disasters that differ in natural hazard type, geographical location, and image parameters. The types of natural hazards were hurricanes, tornadoes, floods, tsunamis, and volcanic eruptions, which struck across North America, Central America, and Asia. We used 175,289 buildings from the xBD dataset, which contains human-annotated multiclass damage labels on high-resolution satellite imagery with red, green, and blue (RGB) bands. First, our experiments showed that the performance in terms of area under the curve does not correlate with the type of natural hazard, geographical region, and satellite parameters such as the off-nadir angle. Second, while performance differed highly between occurrences of disasters, our model still reached a high level of performance without using any labeled data of the test disaster during training. This provides the first evidence that such a model can be effectively applied under operational conditions, where labeled damage data of the disaster cannot be available timely and thus model (re-)training is not an option.

MAPPING OF WATER SURFACE LEVELS AND SLOPES WITH SINGLE PHOTON LIDAR – A CASE STUDY AT THE RIVER RHINE

Abstract. Precise knowledge of water surface level heights is crucial for safe ship navigation and as basis for calibration of hydrodynamic-numerical models. While Airborne Laser Scanning (ALS) is a well established technique for topographic mapping, ALS-based water surface mapping using conventional infrared lasers suffers from the high degree of specular reflection which leads to data voids for off-nadir angles beyond 5–7 degrees. The advent of single photon sensitive ALS systems using green laser sources presents the prospect of large-area, high-resolution water surface mapping due to the high receiver sensitivity and measurement rate of such systems. Building on previous studies on subject matters, we present the results of a pilot project initiated and conducted by the German Federal Institute of Hydrology (BfG, Koblenz) at the Rhine River. Three specific test sites with varying water surface and flow velocity properties were captured on October 30th and 31th, 2019 with the Leica SPL100 from flying altitudes of 3000 m, 2500 m, 1600 m, and 800 m, respectively. As anticipated, the water surface laser pulse density was high and exhibited 20–145 points/m2 depending on flying altitude. After quality control, strip adjustment, and point cloud analysis, three water surface classification methods were implemented based on: (i) height quantiles, (ii) point cloud segmentation, and (iii) inverse DTM filtering. All approaches featured relative and absolute water level height accuracies better than 10 cm. We conclude that Single Photon LiDAR based high resolution mapping of water surface levels and tilts is feasible when employing application specific data acquisition parameters, i.e., off-nadir angle &amp;amp;leq;10° and flying altitude &amp;amp;leq;3000 m.

Biomass Burning in Africa: An Investigation of Fire Radiative Power Missed by MODIS Using the 375 m VIIRS Active Fire Product

Biomass burning plays a key role in the interaction between the atmosphere and the biosphere. The nearly two-decade-old Moderate Resolution Imaging Spectroradiometer (MODIS) active fire product provides critical information (e.g., fire radiative power or FRP) for characterizing fires and estimating smoke emissions. Due to limitations of sensing geometry, MODIS fire detection capability degrades at off-nadir angles and the sensor misses the observation of fires occurring inside its equatorial swath gaps. This study investigates missing MODIS FRP observations using the 375 m Visible Infrared Imaging Radiometer Suite (VIIRS) active fire data across Africa where fire occurs in the majority of vegetation-covered areas and significantly contributes to global biomass-burning emissions. We first examine the FRP relationship between the two sensors on a continental scale and in grids of seven different resolutions. We find that MODIS misses a considerable number of low-intensity fires across Africa, which results in the underestimation of daily MODIS FRP by at least 42.8% compared to VIIRS FRP. The underestimation of MODIS FRP varies largely with grid size and satellite view angle. Based on comparisons of grid-level FRP from the two sensors, adjustment models are established at seven resolutions from 0.05°–0.5° for mitigating the underestimation of MODIS grid FRP. Furthermore, the investigation of the effect of equatorial swath gaps on MODIS FRP observations reveals that swath gaps could lead to the underestimation of MODIS monthly summed FRP by 12.5%. The quantitative information of missing MODIS FRP helps to improve our understanding of potential uncertainties in the MODIS FRP based applications, especially emissions estimation.

The Impact of Pan-Sharpening and Spectral Resolution on Vineyard Segmentation through Machine Learning

Precision viticulture benefits from the accurate detection of vineyard vegetation from remote sensing, without a priori knowledge of vine locations. Vineyard detection enables efficient, and potentially automated, derivation of spatial measures such as length and area of crop, and hence required volumes of water, fertilizer, and other resources. Machine learning techniques have provided significant advancements in recent years in the areas of image segmentation, classification, and object detection, with neural networks shown to perform well in the detection of vineyards and other crops. However, what has not been extensively quantitatively examined is the extent to which the initial choice of input imagery impacts detection/segmentation accuracy. Here, we use a standard deep convolutional neural network (CNN) to detect and segment vineyards across Australia using DigitalGlobe Worldview-2 images at ∼50 cm (panchromatic) and ∼2 m (multispectral) spatial resolution. A quantitative assessment of the variation in model performance with input parameters during model training is presented from a remote sensing perspective, with combinations of panchromatic, multispectral, pan-sharpened multispectral, and the spectral Normalised Difference Vegetation Index (NDVI) considered. The impact of image acquisition parameters—namely, the off-nadir angle and solar elevation angle—on the quality of pan-sharpening is also assessed. The results are synthesised into a ‘recipe’ for optimising the accuracy of vineyard segmentation, which can provide a guide to others aiming to implement or improve automated crop detection and classification.

Validation of Wind Speeds From Brown-Peaky Retracker in the Gulf of Mexico and East Coast of North America

The performance of Brown-Peaky (BP) retracker on retrieving the backscatter coefficients, and then wind speeds, in coastal zones has been examined in this article. Eight years of Jason-2 waveforms are reprocessed by the BP retracker. An empirical wind speed model is used to obtain the altimeter wind speeds based on the BP-derived backscatter coefficients. The validation of BP-derived wind speeds is conducted through comparisons with wind speeds from the in situ anemometer and Jason-2 official altimeter product. The along-track root mean square errors (RMSEs) and biases between altimeter and anemometer wind speed time series are calculated pointwise to evaluate the data quality. The orthogonal regression analysis is computed to evaluate the overall data quality. The validation results show that the BP-derived backscatter coefficients are highly correlated with the square of the off-nadir angles. By removing this correlation-induced error, the quality of BP-derived backscatter coefficients is comparable to that obtained from the three-parameter maximum likelihood estimator (MLE3). The improved backscatter coefficients are beneficial for retrieving reliable along-track wind speed observations, which allows for lower and smoother RMSEs and biases. The 1- and 2-Hz BP-derived wind speeds achieve similar performance, implying that the application of high rate altimeter wind speeds in coastal zones is possible. We also find that the altimeter wind speeds are significantly dependent on the sea state in the last 20-km distance to the coast, where the bias between altimeter and anemometer wind speeds increases remarkably and varies inversely with the offshore distance.