A water quality prediction model for large-scale rivers based on projection pursuit regression in the Yangtze River

Abstract

In recent decades, the Yangtze River Basin, which carries hundreds of millions of people and a substantial economic scale, has been plagued by water quality deterioration, threatening considerably sustainable development. In this paper, a sample set is established based on the water quality indexes of chemical oxygen demand and dissolved oxygen obtained by week-by-week monitoring on the main stream of the Yangtze River in Panzhihua, Yueyang, Jiujiang, and Nanjing from 2006 to 2018. The twelve characteristic variables are selected by random forest technique, and the week-by-week dynamic prediction models of chemical oxygen demand and dissolved oxygen at each section of main stream are established by the projection pursuit regression, which can effectively predict the water quality dynamics of the Yangtze River main stream.