The NASA/JPL Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) instrument has performed tomographic SAR experiments over a number of study areas, including Rabi Forest in Gabon in 2016 and Sierra National Forest in California, USA in 2021. Tomographic SAR, or TomoSAR, is a technique enabling 3-D radar imaging with diverse applications including mapping of vegetation structure. Convolutional neural networks (CNNs) have shown widespread potential for many image processing and computer vision tasks such as image segmentation, classification, and object recognition. By using 3-D CNNs rather than 2-D CNNs, the filters can be applied to all three dimensions of a forest volume imaged by TomoSAR. We have trained 3-D CNN-based deep learning models to estimate canopy height and canopy cover from fully polarimetric UAVSAR TomoSAR images using lidar data as training and validation. When applied to canopy height estimation in the Rabi Forest study area, a trained network had root mean square error (RMSE) of 3.6 m (11%) compared to the validation dataset. For canopy cover estimation in the Sierra National Forest study area, the RMSE was 12%. Further work can be done to optimize the network architecture, improve the output spatial resolution, and to check if these methods can be applied to other study areas or to other vegetation structure parameters such as above-ground biomass. The results show the strong potential of 3-D CNNs for mapping wall-to-wall vegetation structure from tomographic SAR imagery using lidar training data.

AirSWOT is an airborne Ka-band synthetic aperture radar, capable of mapping water surface elevation (WSE) and water surface slope (WSS) using single-pass interferometry. AirSWOT participated in the NASA EVS-3 Delta-X campaign in 2021, which combined remote sensing from multiple instruments with an extensive coincident field data collection in the Mississippi River Delta, Louisiana, USA. As part of Delta-X, AirSWOT flew a greater number of flight lines than in previous AirSWOT campaigns, collecting a significant volume of data which can provide insight into the dynamics and quantity of water in the Atchafalaya and Terrebonne basins of the Mississippi River Delta. AirSWOT data has been processed into publicly available data products at a number of processing levels, depending on user needs and application, including a new Level-3 water surface product developed specifically for Delta-X. The Level-3 water surface product uses water masking and spatial averaging to produce a science-ready point data product, using the Level-2 GeoTIFF raster products as input. The Level-3 data allows profiles of WSE and WSS within designated channels to be easily calculated. AirSWOT estimates of WSE from Delta-X have been compared to in situ water level data with root mean square error (RMSE) of 9 cm, excluding data from two flights in September, 2021 which were adversely affected by poor weather conditions that affected the instrument hardware. Including all data, the RMSE increases to 12 cm. We have also used AirSWOT to help estimate the vertical datum for water level gauges without accurate vertical reference information. AirSWOT is capable of mapping WSE and WSS at high resolution in spatially complex coastal environments, making it a valuable instrument for studying these regions.

The GLISTIN-A instrument was flown on the NASA/AFRC C-20 (Gulfstream III) in December 2022 to observe the Mauna Loa eruption event in Hawaii, USA. As the volcano was actively erupting, several of the swaths were repeated on both the same and successive days to observe changes in the lava flow thicknesses and lava fronts. These repeated swaths provided a unique opportunity to re-address our understanding of the calibration of GLISTIN-A and its ability to precisely and accurately compute the topography of significantly sloped terrains.After calibration with localized troposphere estimates from nearby GNSS sites, adjustments to the roll from post-processed Applanix data, and applying an empirical temperature model, the average slope difference is 4.3 millidegrees and the average RMS difference is 1.7 meters between overlapping swaths. The slope and RMS differences were computed for look angles from 11 to 52 degrees.

In this work, we demonstrate how harmonized optical and SAR satellite imagery can be utilized for robust identification of open water surfaces at a global scale. We train an image segmentation architecture based convolutional neural network (CNN) to extract the most salient features from the input data and generate a per-pixel water/not-water classification. We find that combining optical and radar imagery helps reduce false positive and false negative inferences, illustrating the effectiveness of this harmonization. The resulting model is able to classify water surfaces at the resolution of the SAR sensor (12.5 meters) with a validation set precision and recall of 0.74 and 0.81 respectively. We also demonstrate that the trained model is capable of generating inferences beyond the geographic bounds of the training data.

We have mapped flooded areas in data collected by the NASA/JPL Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) using two convolutional neural network (CNN) image classifier architectures: U-Net and SegNet. Our study area was a region around Houston, TX, USA affected by widespread flooding in 2017 due to Hurricane Harvey. To train and test the classifiers, we manually labelled over 10000 image segments in two flight lines. Both U-Net and SegNet yielded higher accuracy than a previous non-machine learning classifier we used as a baseline. U-Net had slightly higher accuracy than SegNet. The classifiers performed better in areas with more homogeneous land cover. To independently validate the classifier accuracy we used NOAA aerial imagery, with overall accuracy around 80%. Future work includes assessing the classifier robustness in other study areas, assessing the classifier dependence on UAVSAR incidence angle, particularly for open water and bare ground, and collecting more training data, particularly in urban areas. This study demonstrates the potential of CNN image classifiers for mapping flooded areas in airborne polarimetric SAR imagery, and for land cover classification of polarimetric SAR imagery more generally.

We map mangrove extents in Pongara National Park, Gabon using the Freeman-Durden Decomposition and InSAR Coherence derived from ALOS-2 imagery. Specifically, we obtain a land cover map derived from both this polarimetric decomposition and a 14-day repeat-pass coherence. Our classification model and results are highly interpretable based on a depth 2 decision tree. We further illustrate the correlation between InSAR coherence and height obtaining rough mangrove height estimates from TanDEM-X data. From our results, we observe that repeat-pass interferometric coherence provides invaluable information about mangrove extents and coastal forests. The clear identification of mangrove extents presents a significant opportunity for NISAR, which will provide 12-day repeat pass images over coastal areas globally.

We propose a framework to estimate high above ground biomass (AGB) from L-band SAR imagery leveraging spaceborne lidars such as GEDI or ICESat-2 and repeat-pass coherence. Our results indicate we are able to overcome model saturation typically associated with purely backscatter methodologies. We validate our approach using lidar-derived AGB maps from the AfriSAR datasets at Mondah, Ogooue, and Lope. We apply our framework to UAVSAR and ALOS-2 imagery to obtain 50 meter resolution biomass maps. We obtain <; 60% nRMSE (in some cases much better) with negligible relative bias using a multiscale random forest model. We illustrate that the inclusion of coherence can significantly improve high AGB estimation particularly at the coastal site Mondah.

We present a flexible methodology to identify forest loss in synthetic aperture radar (SAR) L-band ALOS/PALSAR images. Instead of single pixel analysis, we generate spatial segments (i.e., superpixels) based on local image statistics to track homogeneous patches of forest across a time-series of ALOS/PALSAR images. Forest loss detection is performed with Support Vector Machines (SVMs)trained on local radar backscatter features derived within superpixels. This method is applied to time-series of ALOS-1 and ALOS-2 radar images over a boreal forest within the Laurentides Wildlife Reserve in Québec. We evaluate four spatial arrangements including 1) single pixels, 2) square grid cells, 3) superpixels based on segmentation of the radar images, and 4) superixels derived from ancillary optical imagery (e.g. Landsat). Detection of forest loss with superpixels outperform single pixel and regular grid methods, especially when superpixels are generated from ancillary optical imagery. Results are validated with official Québec forestry data and Hansen forest loss products. Our results indicate that this approach may be applied operationally to monitor forests across large study areas with L-band radar instruments such as ALOS/PALSAR.

Kapok is a Python library created to estimate forest height using repeat-pass polarimetric synthetic aperture radar interferometry (PolInSAR). The library can import data collected by NASA's Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) sensor. The library includes functions for data visualization, coherence region plotting, coherence optimization, and inversion of the random volume over ground forest model. The estimated forest height maps or other output products can be exported in various GIS-ready raster formats for validation and analysis. The software is released in the hopes of growing the UAVSAR user community, as well as in the interests of education and outreach. Kapok has been released under the GNU GPL software license, and the full source code is available for download at: github.com/mdenbina/kapok.

We have explored the effects of temporal baseline and terrain slope on forest height estimation using L-band repeat-pass polarimetric synthetic aperture radar interferometry. Data were collected using NASA's Uninhabited Aerial Vehicle Synthetic Aperture Radar instrument over a study area exhibiting high slope topography in the Laurentides Wildlife Reserve of Québec, Canada. We used lidar-derived canopy height and terrain slope maps to quantify the decorrelation effects that distort the observed coherences compared to the random volume over ground forest model. We derived forest height maps for a number of different temporal baselines using both fixed model parameters and model parameters that varied with slope, and compared the results. Use of a look-up table for the terrain slope effects improved the estimated forest heights, but further work is necessary to see if slope corrections derived from lidar data for this study area can be applied to other study areas, or generalized to a theoretical model.

Radiometric correction of radar images is essential to produce accurate estimates of biophysical parameters related to forest structure and biomass. We present a new algorithm to correct radiometry for 1) terrain topography and 2) variations of canopy reflectivity with viewing and tree-terrain geometry. This algorithm is applicable to radar images spanning a wide range of incidence angles over terrain with significant topography and can also take into account aircraft attitude, antenna steering angle, and target geometry. The approach includes elements of both homomorphic and heteromorphic terrain corrections to correct for topographic effects and is followed by an additional radiometric correction to compensate for variations of canopy reflectivity with viewing and tree-terrain geometry. The latter correction is based on lookup tables and enables derivation of biophysical parameters irrespective of viewing geometry and terrain topography. We evaluate the performance of the new algorithm with airborne radar data and show that it performs better than classical homomorphic methods followed by cosine-based corrections.

Iceberg monitoring, as well as maritime surveillance in general, is an important application of synthetic aperture radar (SAR). Compact polarimetry offers the potential for improved detection performance compared to traditional linear dualpol SAR. The Radarsat Constellation, Canada's next Radarsat satellite, will collect compact polarimetric data in a variety of imaging modes. Using Radarsat-2 fine-quad mode data we have simulated Radarsat Constellation data and calculated the detection performance of the compact data using 25 validated iceberg locations in the Labrador Sea. We also propose a method through which the shape and orientation of the received polarization ellipse in the compact polarimetric data can be used to improve the detection performance.

Circular-transmit/linear-receive compact polarimetry synthetic aperture radar systems combine coherent dual polarization with wide-swath imaging. The polarization information in these data may be represented as a Stokes vector, or one can reconstruct several quadpolarized covariance elements. Two reconstruction algorithms have been published in the literature: one by Souyris and a refined algorithm by Nord We investigated the application of these two algorithms for reconstructing ocean clutter for the purpose of detecting targets. We tested the assumptions underlying these algorithms and found that they were not valid for ocean scenes. We present a simple empirical reconstruction model whose reconstruction and target detection performance is superior to the two published algorithms.

Polarimetric InSAR (PolInSAR) using repeat-pass L-Band has generated interest in recent years because of its potential for extraction of forest height and of bare-earth topography beneath the canopy. However temporal de-correlation remains a problem. In previous papers a single-pass system has been demonstrated which removes the temporal issue. In this paper we extend the single-pass PolInSAR work previously described and show results for forests in which tree height maps and corresponding DTMs have been generated and compared to lidar truth.