How anthropogenic factors influence the dissolved oxygen in surface water over three decades in eastern China?

Abstract

The aquatic environment, linked to the sustainable development of human existence and ecological environment, is influenced comprehensively by anthropogenic and natural activities. In light of the continuously low concentration of dissolved oxygen (DO) in surface water in plain river networks and the phenomenon of delay in the improvement of surface water quality, this research aims to introduce a method that may be utilized in identifying the critical driving forces of DO in surface water and their lagging characteristics, which will contribute to the assessment and adjustment of water quality drivers and/or policies. The research analyzes a typical small watershed in a river network region of the Yangtze River Delta plain as the study area, collecting 35-year (1986–2020) data on several water quality parameters, decades of anthropogenic activities, and two natural factors. The time series methods of vector autoregressive model, Granger causality tests, forecast error variance decompositions, and impulse response functions (hereinafter referred to as VAR+), which are rarely applied in related research, were employed in this study and proved helpful for screening out pivotal drivers and capturing the lagging responses of DO level to driving forces at each lagged time. Results show that there exists a fluctuating drop in DO level in surface water from 1986 to 2008 and a steady climb from 2008 to 2020, with the lowest DO level being present in 2008. The impulsive perturbations of phosphate fertilizer consumption (PFC), motor vessel number, and precipitation minimally increase DO concentration, while the impulsive perturbation of gross domestic product (GDP) causes the sharpest drop in DO level. With these perturbations, the driving force of PFC persists for approximately seven years, and the driving forces of water temperature, permanent population, and GDP persist for only five years. Future research could be conducted with spatial hysteresis, selection of lag order and variable quantity within the model, as well as intermediate variables between drivers and DO level for exploring driving pathways and mechanisms.