Author(s): Jiaming Liu; Yanjun Zhang; Xingyuan Song; Yang Kuang; Di Yuan

Linked Author(s):

Keywords: Rtificial neural networks (ANNs); Multiple regression; IKONOS multispectral data; High resolution satellite remote sensing; Water quality; Lake Cihu

Abstract: This paper presents different methodologies to estimate water quality parameters which are chemical oxygen demand (COD), ammonia nitrogen (NH3-H), total nitrogen (TN) and total phosphorus (TP) concentration in Lake Cihu, Huangshi, from high resolution satellite remote sensing data, based on multiple regression and artificial neural networks (ANNs). Image procedure including radiometric calibration and atmospheric correction converts digital numbers into surface reflectance. Then multiple regression and ANNs are applied to the visible and near-infrared bands of IKONOS in order to determine a relationship between the surface reflectance of the lake and the water quality parameters obtained by in situ measurements. Statistical analysis using determination coefficients and error estimation is employed, aiming to evaluate the most accurate methodology. The results show that the estimated accuracy of water quality parameters using ANNs is higher than the accuracy using multivariate regression approaches, and the measured and estimated values for water quality parameters are in good consistency (R2 > 0. 9). The spatial distribution maps of water quality parameters generated by ANNs model present apparent spatial variations and inform the decision makers of water quality variations in Lake Cihu.

DOI:

Year: 2013