Inverse Modeling of Beaver Reservoir's Water Spectral Reflectance


Estimation of inherent optical properties (IOP) needed for water quality evaluation by remote sensing models is very complex, primarily due to the large number of model simulations needed to find optimal parameter values. This study presents an approach for optimally parameterizing the IOP values of a physical hyperspectral optical - Monte Carlo (PHO-MC) model. An artificial neural network (ANN) based pseudo simulator combined with the Nondominated Sorted Genetic Algorithm II (NSGA II) was used to efficiently perform a large number of model simulations and to search the optimal parameter values for IOP determination. Concentrations of suspended matter (sm), chlorophyll-a (chl), and total dissolved organic matter (DOM) along with the reflectance data at 16 different wavelengths were measured at 48 sampling stations in the Beaver Reservoir, Arkansas, between 2003 and 2005 and were used to evaluate the IOP values. Measured concentrations and reflectance data from 24 sampling stations were used to optimize IOP parameter values for sm, chl, and DOM. The data collected from the remaining 24 sampling stations were used for the validation of PHO-MC model-predicted reflectance by using optimized IOP values. PHO-MC predicted reflectance values were significantly correlated (r = 0.90, p < 0.01) with the corresponding measured reflectance values, indicating that the pseudo simulator combined with the NSGA II accurately estimated the IOP values. An estimated 10 10 years of calculation time was reduced to less than 3 min by using the pseudo simulator and NSGA II to supplement the PHO-MC model for estimating the IOP values.


ANN, Beaver Reservoir, GA, Inherent optical properties, Inverse modeling, Remote sensing, Water quality

Date of this Version








Link Out to Full Text