Retrieval of Soil Organic Carbon from Hyperspectral Imagery
Abstract
Even though wetlands are among the more effective ecosystems that store carbon (Chmura, Anisfeld, Cahoon, & Lynch, 2003; Duarte, Middelburg, & Caraco, 2005), it can be challenging to accurately map carbon storage at fine scales due to the inherent variability in these systems. However, the erectophile nature of the salt marsh plant canopy and subsequent visibility of the sediment suggest that remote imaging may be useful for retrieving soil organic carbon (SOC) at the surface. The NASA SBG mission will provide global coverage with high spectral and spatial resolution, allowing the potential for more accurate quantitative estimates of SOC in wetland systems. As a prelude, we evaluated the relationship between SOC and soil reflectance using hyperspectral imagery of coastal wetlands obtained from unmanned aerial systems (UAS) (Kaputa et al., 2019; Eon & Bachmann, 2021) and a mast-mounted hyperspectral imaging system (Bachmann et al., 2019). Previously, we used some of this imagery and ground-truth to develop models for retrieving above-ground biomass using PROSAIL in an earlier study (Eon et al., 2019), as well as models of salt-marsh stress (Goldsmith et al., 2020). The present study considers the sources of variability that make accurate and repeatable retrieval of SOC challenging. Hyperspectral imagery and concomitant in situ ground truth data were collected over a three-year period between 2017-2019 on Hog Island, a barrier island that is part of the Virginia Coast Reserve Long Term Ecological Research site. We also consider companion laboratory analysis of some of these field samples, examining the angular dependence of spectral reflectance derived from a hyperspectral goniometer system (Harms et al., 2017) at various illumination angles and its relationship to more accurate SOC retrieval. We discuss the challenges of transferring laboratory-based models to predict SOC in inaccessible areas. Improved models are essential for more accurately mapping wetland SOC to support natural resource managers.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMGC35A..03N