Retrieval of Sediment Filling Factor in a Salt Panne from Multi-View Hyperspectral Imagery

Abstract

This work describes a study using multi-view hyperspectral imagery to retrieve sediment filling factor through inversion of a modified version of the Hapke radiative transfer model. We collected multi-view hyperspectral imagery from a hyperspectral imaging system mounted atop a telescopic mast from multiple locations and viewing angles of a salt panne on a barrier island at the Virginia Coast Reserve Long-Term Ecological Research site. We also collected ground truth data, including sediment bulk density and moisture content, within the common field of view of the collected hyperspectral imagery. For samples below a density threshold for coherent effects, originally predicted by Hapke, the retrieved sediment filling factor correlates well with directly measured sediment bulk density ( R 2 = 0.85 ). The majority of collected samples satisfied this condition. The onset of the threshold occurs at significantly higher filling factors than Hapke’s predictions for dry sediments because the salt panne sediment has significant moisture content. We applied our validated inversion model to successfully map sediment filling factor across the common region of overlap of the multi-view hyperspectral imagery of the salt panne.

Mapping barrier island soil moisture using a radiative transfer model of hyperspectral imagery from an unmanned aerial system

“…(5) is likely correlated with sediment porosity 26 . Other radiative transfer models such as those due to Hapke explicitly parameterize sediment fill factor 36,40 , and we have successfully used a variant of this model to retrieve sediment fill factor from multi-view hyperspectral imagery 33,41 ; however, models such as these have more free parameters that must be optimized. We have focused on the retrieval of the soil moisture content through the inversion of MARMIT from five different soil data sets.…”

“…The four samples represent distinctly different soil types, varying in physical characteristics such as mineralogy, salinity, texture, organic matter content and roughness. The samples were acquired from the Algodones Dunes 51,52 in California, a lakebed region in Northwest Nevada 53 , and Hog Island, VA 33,54 in the Delmarva Peninsula of Virginia. The Algodones sediment sample (ALG) is typically composed of quartz, feldspar, rock fragment and an assortment of heavy minerals 55 .…”

“…The Nevada sample (NEV) is mostly composed of clay. One of the sediment samples from Hog Island is from a saltpanne (HOGP) environment 33 and soil from this region can be considered to be a loam, consisting of mostly clay, silt and sand. Like the ALG desert sediment, the other sample from the island is also a sandy sediment taken from the beach shore on Hog Island (HOGB).…”

“…This study was part of a broader field campaign conducted in summers of 2018 and 2019 to study the beach and intertidal zone sediments 33 as well as to characterize and map the salt marsh ecosystem 69,70 in the southern portion of Hog island. The data collected for this study took place specifically along the shore on the southern tip of the island.…”

“…We collected hyperspectral data along the shore on the southern tip of Hog island, VA, using UAS-based imaging platforms 15 during field campaigns in 2018 and 2019 33 . From the UAS platforms, we collected hyperspectral imagery from two HSI systems, one a Headwall Nano-Hyperspec with 270 spectral bands in the visible and near infrared (0.4-1.0 µm ), and the second a Headwall Micro-Hyperspec with 267 spectral bands in the shortwave infrared (SWIR) (0.9-2.5 µm ) .…”