Abstract

Derivative indices may be successfully applied to estimate concentrations of optical water constitutes from remotely sensed data. Reflectance, spectral derivative and spectral ratio technologies are used to evaluate the potentiality of developing spectral models for the prediction of three water quality parameters (pH, NH4-N and DO). Results indicate that derivative spectra (483nm; 697 nm and 647nm) and spectral ratio (665nm/632nm, 572nm/574nm and 494nm/528nm) are more effective in simulating pH (RMSE < 0.63; R2>0.51), NH4-N concentrations (RMSE < 0.28 mg/L; R2>0.64) and DO (RMSE < 3.2 mg/L; R2>0.43) than original reflectance. Multiple bands combination models (430 nm, 434 nm, 680nm and 697nm) of first-derivative spectra and second-derivative spectra (516nm, 684nm and 545nm) can greatly improve NH4-N simulation accuracies (RMSE < 0.17 mg/L; R2>0.94). Comparison with derivative spectra, spectral ratio models are much simpler methods to derive pH (RMSE = 0.23; R2=0.72), NH4-N (RMSE < 0.28 mg/L; R2=0.77) and DO (RMSE < 0.28 mg/L; R2=0.58) while not losing of the simulation accuracies.

© 2012 Optical Society of America

PDF Article