Search

Accurate crop classification with synthetic aperture radar (SAR) data is a significant area of research and translating into practice from local to regional scale crop inventory mapping. With the growing accessibility to abundant data sources from both current and upcoming dual-polarimetric SAR missions, the capability to generate precise crop maps is set to enhance substantially. The geometric and dielectric properties of targets highly influence radar backscatter. Especially for agricultural crops, which exhibit dynamic changes in target properties and physiological structure throughout their phenology, discriminating between crops using SAR data remains a significant challenge. This study introduces a novel Gaussian process classifier model called the evidence modified Gaussian process classifier (EM-GPC) that employs a regression approach for multiclass crop classification with a modified evidence. We utilize dual-polarimetric SAR data acquired over two study sites to perform crop classification utilizing EM-GPC. At first, we perform a phenology-based single-date crop classification using ESAR C- and L-band SAR data acquired over the DEMMIN site in Germany. The performance evaluation of the EM-GPC model revealed robust accuracy during various crop phenological stages, showcasing its adaptability to temporal variations. Further, we perform a multidate crop classification using C-band RADARSAT-2 and L-band simulated NISAR product from UAVSAR data acquired over Manitoba, Canada. In this study, EM-GPC successfully classifies major crop types (wheat, canola, soybeans, corn, barley, oats, and rye). The efficacy of the proposed model establishes its capability for crop classification utilizing dual-polarimetric SAR data in operational settings.

An algorithm for classifying crop areas using multi-frequency Synthetic Aperture Radar (SAR) and optical data is evaluated for the upcoming NASA ISRO SAR (NISAR) mission and its active crop area products. Two time-series of L-band ALOS-2 and C-band Sentinel-1A images over an agricultural region in the Southern United States are used as the input, as well as high-resolution Planet optical data. To overcome the delay by at least one year of existing landcover maps, training and validation sets of crop/non-crop polygons are derived with the contemporary Planet images. The classification results show that the 80% requirement on the NISAR science accuracy is achievable only with L-band HV input and with a resolution of 100 m. In comparison, HH polarized images do not meet this target. The spatial resolution is a key factor: 100 m is necessary to accomplish the 80% goal, while 10 m do not produce the desired accuracy. Unlike the previous study reporting that C-band performs better than L-band, we found otherwise in this study. This suggests that the performance likely depends on the site of interest and crop types. Alternative to the SAR images, the Normalized Difference Vegetation Index (NDVI) from the Planet data is not effective either as an input to the classification algorithm or as ground truth for training the algorithm. The reason is that NDVI becomes saturated and temporally static, thus rendering crop pixels to be misclassified as non-crop.

A general limitation in assessing the accuracy of land cover mapping is the availability of ground truth data. At sites where ground truth is not available, potentially inaccurate proxy datasets are used for sub-field-scale resolution investigations at large spatial scales, i.e., in the Contiguous United States. The USDA/NASS Cropland Data Layer (CDL) is a popular agricultural land cover dataset due to its high accuracy (>80%), resolution (30 m), and inclusions of many land cover and crop types. However, because the CDL is derived from satellite imagery and has resulting uncertainties, comparisons to available in situ data are necessary for verifying classification performance. This study compares the cropland mapping accuracies (crop/non-crop) of an optical approach (CDL) and the radar-based crop area (CA) approach used for the upcoming NASA-ISRO Synthetic Aperture Radar (NISAR) L- and S-band mission but using Sentinel-1 C-band data. CDL and CA performance are compared to ground truth data that includes 54 agricultural production and research fields located at USDA’s Beltsville Agricultural Research Center (BARC) in Maryland, USA. We also evaluate non-crop mapping accuracy using twenty-six built-up and thirteen forest sites at BARC. The results show that the CDL and CA have a good pixel-wise agreement with one another (87%). However, the CA is notably more accurate compared to ground truth data than the CDL. The 2017–2021 mean accuracies for the CDL and CA, respectively, are 77% and 96% for crop, 100% and 94% for built-up, and 100% and 100% for forest, yielding an overall accuracy of 86% for the CDL and 96% for CA. This difference mainly stems from the CDL under-detecting crop cover at BARC, especially in 2017 and 2018. We also note that annual accuracy levels varied less for the CA (91–98%) than for the CDL (79–93%). This study demonstrates that a computationally inexpensive radar-based cropland mapping approach can also give accurate results over complex landscapes with accuracies similar to or better than optical approaches.

Planting date (PD) is one of the critical component in crop modeling that has a large-scale impact on seasonal yield estimation. Here, we aimed to estimate the various paddy/rice attributes (particularly, acreage and PD) using the Sentinel-1A observations, and assessing the impact of such key parameters on end-of-season yields. The method comprises: (1) a multi-temporal time-series analysis algorithm of Sentinel-1A observations for rice mapping, (2) a PD retrieval algorithm, and (3) rice yield estimation using the Synthetic Aperture Radar (SAR)-derived PD in a modeling system. Our study was implemented over the rainfed paddy fields in Cambodia between 2017 thru 2020. We demonstrated that the backscatter ratio (σVH0σVV0) from Sentinel-1A can improve the accuracy of rice mapping, while σVH0 backscatter can provide PD estimations. Our results showcased reasonable/good performance in mapping the paddy fields and estimating PD in the study area. PD found in March and April were indicative of irrigation application in some parts of Cambodia, specifically in the southeastern part (also part of the Mekong delta) where irrigation in paddy fields is a common practice. Differences between the crop-calendar based PD (June 1) and the SAR-derived PD can range as much as 75 days. In the rice-dominated provinces such as Banteay-Meanchey and Svay-Rieng of Cambodia, when SAR-derived PD are used, rice yield normalized bias improved compared to the fixed PD data by a mean of 7–12 % and 30–48 % respectively within the study period. In addition, the PD derived from SAR also significantly reduced the uncertainty in yield estimation. Overall, this study demonstrated the potential of using the SAR-derived PD in improving the rice yield estimation over rainfed paddy fields.

Physical understanding and modeling of Ka-band ocean surface backscatter is challenging due to a lack of measurements. In the framework of the NASA Earth Ventures Suborbital-3 Submesoscale Ocean Dynamics Experiment (S-MODE) mission, a Ka-Band Ocean continuous wave Doppler Scatterometer (KaBODS) built by the University of Massachusetts, Amherst (UMass) was installed on the Woods Hole Oceanographic Institution (WHOI) Air-Sea Interaction Tower. Together with ASIT anemometers, a new data set of Ka-band ocean surface backscatter measurements along with surface wind/wave and weather parameters was collected. In this work, we present the KaBODS instrument and an empirical Ka-band wind Geophysical Model Function (GMF), the so-called ASIT GMF, based on the KaBODS data collected over a period of three months, from October 2019 to January 2020, for incidence angles ranging between 40° and 68°. The ASIT GMF results are compared with an existing Ka-band wind GMF developed from data collected during a tower experiment conducted over the Black Sea. The two GMFs show differences in terms of wind speed and wind direction sensitivity. However, they are consistent in the values of the standard deviation of the model residuals. This suggests an intrinsic geophysical variability characterizing the Ka-band surface backscatter. The observed variability does not significantly change when filtering out swell-dominated data, indicating that the long-wave induced backscatter modulation is not the primary source of the KaBODS backscatter variability. We observe evidence of wave breaking events, which increase the skewness of the backscatter distribution in linear space, consistent with previous studies. Interestingly, a better agreement is seen between the GMFs and the actual data at an incidence angle of 60° for both GMFs, and the statistical analysis of the model residuals shows a reduced backscatter variability at this incidence angle. This study shows that the ASIT data set is a valuable reference for studies of Ka-band backscatter. Further investigations are on-going to fully characterize the observed variability and its implication in the wind GMF development.

Crop area mapping is important for tracking agricultural production and supporting food security. Spaceborne approaches using synthetic aperture radar (SAR) now allow for mapping crop area at moderate spatial and temporal resolutions. Multi-frequency SAR data is highly useful for crop monitoring because backscatter response from vegetation canopies is wavelength dependent. This study evaluates the utility of C-band Sentinel-1B (Sentinel-1) and L-band ALOS-2 (PALSAR) data, collected during the 2019 growing season, for generating accurate active crop extent (crop vs. non-crop) classifications over an agricultural region in western Canada. Evaluations were performed against the Agriculture and Agri-Food Canada satellite-based Annual Cropland Inventory (ACI), an open data product that maps land cover across the extent of Canada’s agricultural land. Classifications were performed using the temporal coefficient of variation (CV) approach, where an optimal crop/non-crop delineating CV threshold (CVthr) is selected according to Youden’s J-statistic. Results show that crop area mapping agreed better with the ACI when using Sentinel-1 data (83.5%) compared to PALSAR (73.2%). Analysis of performance by crop reveals that PALSAR’s poorer performance can be attributed to soybean, urban, grassland, and pasture ACI classes. This study also compared CV values to in situ wet biomass data for canola and soybeans, showing that crops with lower biomass (soybean) had correspondingly lower CV values.

This paper describes an automatic mosaicking algorithm for creating large-scale mosaic maps of forest height. In contrast to existing mosaicking approaches through using SAR backscatter power and/or InSAR phase, this paper utilizes the forest height estimates that are inverted from spaceborne repeat-pass cross-pol InSAR correlation magnitude. By using repeat-pass InSAR correlation measurements that are dominated by temporal decorrelation, it has been shown that a simplified inversion approach can be utilized to create a height-sensitive measure over the whole interferometric scene, where two scene-wide fitting parameters are able to characterize the mean behavior of the random motion and dielectric changes of the volume scatterers within the scene. In order to combine these single-scene results into a mosaic, a matrix formulation is used with nonlinear least squares and observations in adjacent-scene overlap areas to create a self-consistent estimate of forest height over the larger region. This automated mosaicking method has the benefit of suppressing the global fitting error and, thus, mitigating the “wallpapering” problem in the manual mosaicking process. The algorithm is validated over the U.S. state of Maine by using InSAR correlation magnitude data from ALOS/PALSAR and comparing the inverted forest height with Laser Vegetation Imaging Sensor (LVIS) height and National Biomass and Carbon Dataset (NBCD) basal area weighted (BAW) height. This paper serves as a companion work to previously demonstrated results, the combination of which is meant to be an observational prototype for NASA’s DESDynI-R (now called NISAR) and JAXA’s ALOS-2 satellite missions.

This paper describes a novel, simple and efficient approach to estimate forest height over a wide region utilizing spaceborne repeat-pass InSAR correlation magnitude data at L-band. We start from a semi-empirical modification of the RVoG model that characterizes repeat-pass InSAR correlation with large temporal baselines (e.g., 46 days for ALOS) by taking account of the temporal change effect of dielectric fluctuation and random motion of scatterers. By assuming (1) the temporal change parameters and forest backscatter profile/extinction coefficient follow some mean behavior across each inteferogram; (2) there is minimal ground scattering contribution for HV-polarization; and (3) the vertical wavenumber is small, a simplified inversion approach is developed to link the observed HV-polarized InSAR correlation magnitude to forest height and validated using ALOS/PALSAR repeat-pass observations against LVIS lidar heights over the Howland Research Forest in central Maine, US (with RMSE < 4 m at a resolution of 32 hectares). The model parameters derived from this supervised regression are used as the basis for propagating the estimates of forest height to available interferometric pairs for the entire state of Maine, thus creating a state-mosaic map of forest height. The present approach described here serves as an alternative and complementary tool for other PolInSAR inversion techniques when full-polarization data may not be available. This work is also meant to be an observational prototype for NASA’s DESDynI-R (now called NISAR) and JAXA’s ALOS-2 satellite missions.

Estimates of above ground biomass density in forests are crucial for refining global climate models and understanding climate change. Although data from field studies can be aggregated to estimate carbon stocks on global scales, the sparsity of such field data, temporal heterogeneity and methodological variations introduce large errors. Remote sensing measurements from spaceborne sensors are a realistic alternative for global carbon accounting; however, the uncertainty of such measurements is not well known and remains an active area of research. This article describes an effort to collect field data at the Harvard and Howland Forest sites, set in the temperate forests of the Northeastern United States in an attempt to establish ground truth forest biomass for calibration of remote sensing measurements. We present an assessment of the quality of ground truth biomass estimates derived from three different sets of diameter-based allometric equations over the Harvard and Howland Forests to establish the contribution of errors in ground truth data to the error in biomass estimates from remote sensing measurements.