Identification and quantification of aquatic vegetation with hyperspectral remote sensing in western Nevada rivers, USA

Abstract

This study used a portable spectrometer to assess the feasibility of using airborne hyperspectral imagery to map the dominant types and amounts of aquatic vegetation in the Carson and Truckee Rivers of western Nevada. Spectral reflectance data were acquired for a number of periphyton and macrophyte types, and corresponding vegetation samples were processed in the laboratory to quantify chlorophyll a (chla) and ash-free dry mass. The dominant periphyton and macrophyte communities encountered in the field could be identified with an overall accuracy of greater than 95%. The lowest individual class accuracy was 82% for one community type, primarily green filamentous algae (GF), which was brown in colour and mixed with diatoms and sediments. Separate stepwise regression models were developed for chla and biomass of each type of vegetation. Regression models had r ²s greater than 0.92, except for the aforementioned brown-coloured community of mixed algae that had r ²s of just over 0.5 for both laboratory measurements. This study suggests good prospects for airborne hyperspectral surveys of aquatic vegetation for water quality studies, assuming a sensor with a high signal-to-noise ratio, high spatial resolution and good environmental conditions at the time of image acquisition.