There is a growing need to quantify large-scale below ground carbon sequestration rates in coastal wetlands to understand marsh resilience to sea level rise and define eligibility for carbon offset credits. Remote sensing-based estimation of biomass and plant productivity over large spatial extents will help address these needs. Working on field plots in the Sacramento-San Joaquin River Delta, California, every 2 to 3 weeks during the growing seasons of 2011 and 2012, we collected above ground biomass samples, ƒAPAR (fraction of absorbed photosynthetically active radiation) data, and reflectance data from hardstem bulrush and cattail using a field spectrometer (350-2,500 nm).
The plot-level reflectance data were analyzed to develop hyperspectral and multispectral vegetation indices that predict biomass and ƒAPAR, and account for seasonal and background effects of water inundation and litter. Biomass estimation was most successful (R2 = 0.78) for the mid-summer months when water levels were less than 5 cm when using a Normalized Difference Vegetation Index (NDVI)-type two band vegetation index (TBVI) based on wavelengths 1,790 nm and 1,164 nm (figure below). A TBVI based on wavelengths 1,565 nm and 865 nm (the mid-points of the shortwave (SWIR)-1 and near-infrared (NIR) Landsat 8 bands) also successfully estimated biomass, with R² = 0.75, indicating the suitability of this sensor to estimate wetland vegetation biomass. Imagery from the Hyperion satellite, Landsat and DigitalGlobe WorldView-2 has been acquired for the study sites. ƒAPAR values were combined with gross primary productivity (GPP) estimates from eddy correlation flux measurements to develop a light-use efficiency model of plant production. APAR, calculated as daily PAR*average ƒAPAR, increased with GPP at the USGS experimental site, with R² = 0.99 for a regression of GPP and APAR. Based upon our field study, the best indices for each sensor will be used to scale up biomass measurements to the site level.
Contour Plot of all Simulated Hyperion TBVI-Biomass Regressions