GIS and Artificial Neural Network–Based Water Quality Model for a Stream Network in the Upper Green River Basin, Kentucky, USA

Abstract

The prediction of stream water quality (WQ) is essential to understand and quantitatively describe water quality parameters (which include physical characteristics, inorganic metallic, and nonmetallic concentrations) and their structure, watershed health, biodiversity, and ecology of a basin. The spatial variability and temporal randomness of stream water quality parameters makes the problem a complex modeling task by ordinary statistical regression methods. The determination of water quality parameters and their spatial and temporal description in stream networks is even more complex due to the stochastic nature of water flow, atmospheric conditions, meteorological patterns, and nonlocal effects of precipitation and temperature. In this paper, a statistical, geographic information system (GIS) and a neural network based water quality model is developed to study stream water quality parameter structure in a geographic framework in the United States of America (USA) consisting of stream network, watershed, and a variety of different land-use practices. Also, a novel way of representing land use in the form of land-use factor (LUF) is formulated for modeling purposes.