Application of Machine Learning Algorithms in Studying Water Quality in the Red River System

Abstract

The Red River system plays an important role in the socio-economic development of the Northern of Vietnam. Therefore, regular monitoring and evaluation of water quality parameters in the Red River system are important in water resources management and protection. However, current monitoring methods are often quite expensive and time-consuming. To predict the downstream water quality, this study uses multiple machine learning algorithms to understand the correlation between environmental parameters measured at upstream and downstream stations of the Red River system. The environmental parameters that are chosen for this study include suspended sediment concentration (SSC), inorganic nitrogen content (total N), phosphorus content (total P), and dissolved silicon (DSi). The results show that machine learning algorithms can estimate the downstream DSi and sediment concentrations based on combining values of three upstream stations with relatively high efficiency (R2 equals 0.75 and 0.66, respectively). Meanwhile, these algorithms have limited performance in estimating total N and P content, due to the influence of many exogenous factors. The study introduces a new direction for applying machine learning algorithms in water quality research in the Red River system with the potential application in other river systems in Vietnam.