A new index for detection of submerged aquatic plants under variable quality water: an extension of the soil-line concept

Abstract

Submersed aquatic plant (SAP) communities are important determinants of estuarine and lacustrine food web structure, nutrient cycling, and water quality. Variation in water quality, SAP community composition and cover present challenges when mapping SAP robustly across estuarine ecosystems. We propose three new spectral indices based on the soil-line concept that overcome the confounding influence of varying water quality and SAP cover in shallow inland waters. Spectral variability of water due to water quality differences can be modeled using a “water line” while SAP spectral differences due to cover and/or composition, can be modeled using a “SAP line.” Spectral distance of the pixel from the SAP line in a hypothetical two-band spectral space represents the inverse probability that the pixel contains SAP. This distance from the SAP line represents the Perpendicular SAP Index using an SAP line (PSIS). Correspondingly, the distance of a pixel from the water line represents the Perpendicular SAP Index using a Water line (PSIW). We also tested a compound index, PSIΔ, the difference between PSIS and PSIW, following the reasoning that an SAP pixel is closer to the SAP line compared to a water line and a water pixel is closer to the water line compared to the SAP line. The results of this study indicate that yellow and red-edge band-pairs performed best for calculating PSIS and red and red-edge band-pairs performed best for calculating PSIW. The same band-pairs as PSIW worked best for PSIΔ. The accuracy for PSIΔ was either as good as or better than the accuracy for PSIS or PSIW. The red-edge band required to calculate three PSI indices should be narrow (10–20 nm bandwidth) and centered around 700 nm. Bands ideal for calculating PSI are available on the Sentinel-2 and WorldView-3 satellite sensors. When compared to seven other narrow and broadband spectral indices from the same spectral region, PSIW and PSIΔ performed best for separating water and SAP classes. Thus, this family of indices shows promise for detection of SAP mats in shallow inland waters over a wide range of water quality parameters as observed in our estuarine study system and could prove to be a major step forward in detection of SAP using multispectral data.

Keywords

• Hyperspectral remote sensing
• inland estuary
• mapping
• environmental monitoring
• aquatic vegetation