Salt marshes constitute an important terrestrial-aquatic interface that remains underrepresented in Earth System Models due to constraining biophysical controls and spatially limited land cover. One promising approach to improve representativeness is the application of proximal remote sensing to generate phenological information, yet we lack detailed knowledge on how proximal sensors and indices perform within these ecosystems. We use measurements of net ecosystem productivity (NEP) from eddy covariance (EC) and derive ecologically-relevant phenology parameters (i.e., phenoperiods) to use as carbon phenology benchmarks. These benchmarks are compared against vegetation indices and spectral bands derived from spaceborne (i.e., MODIS) or common proximal sensors (i.e., phenocam and spectral reflectance sensors; SRS).
Phenocam derived indices, which exclude infrared wavelengths (i.e., vegetation contrast index; VCI and greenness chromatic coordinate; GCC), aligned closely with NEP benchmarks and provided best predictions of carbon sink season length (within 1–6 days of benchmark). Although isolating infrared from vegetation (NIRv) offered improvements, other indices utilizing infrared bands (i.e., normalized difference vegetation index; NDVI and enhanced vegetation index; EVI) primarily underestimated season start dates (5–30 days prior to benchmark) while overestimating season end dates (7–47 days after benchmark). These discrepancies are greatest for indices derived from MODIS and SRS sensors, which have narrower full width half maximum spectral bandwidths and sharper orientation angles. The phenocam (VCI and GCC) provides the most accurate phenology parameters while offering near-infrared (NIR) response which can generate additional information on seasonal changes in canopy structure and function.
The distinctive characteristics of the salt marsh environment and vegetation properties including standing dead biomass can introduce interpretation challenges for commonly used vegetation indices (NDVI, EVI). Incorporating information from proximal sensors utilizing only visible wavelengths (VCI, GCC) or isolating the near-infrared reflectance of vegetation (NIRv) offers improvements for studying carbon phenology within salt marshes.