Detection of periodicity, aperiodicity, and corresponding driving factors of river dissolved oxygen based on high-frequency measurements

Abstract

Understanding dissolved oxygen (DO) dynamics and corresponding driving factors are essential to improve aquatic environments and protect aquatic organisms. Although several studies have been conducted to investigate DO dynamics on low-frequency scales, high-frequency spectral characteristics of DO and the corresponding connections with hydro-biogeochemical drivers were little known. Accordingly, this study proposed a comprehensive framework to explore the periodic and aperiodic characteristics of DO, and the corresponding hydrological (water temperature) and biogeochemical (pH, ammonia nitrogen (NH3-N), and total phosphorus (TP)) controls in the Pearl River Basin of China based on high-frequency measurements and spectral analysis. The results showed that DO, water temperature, pH, NH3-N and TP all exhibited temporal fractal phenomena, i.e., 1/f fluctuations, in low-frequency domains. The scaling component β of water temperature was larger, implying that the self-similarity of water temperature was stronger. In high-frequency domains, DO exhibited obvious daily and half-day periods. Of the other four elements, water temperature and pH exhibited high-frequency periods, while NH3-N and TP did not. Based on the energy distribution derived from the continuous wavelet transform, three DO transience types, swing, abrupt, and mixed swing and abrupt, were characterized. The first type was characterized by continuous fluctuations, while the second type was characterized by a sudden change with a peak value. The last type had traits of both swing transience and abrupt transience. Through an analysis of maximal information coefficient and correlation coefficient, we found that linear relationships existed between water temperature and DO with correlation coefficient ranging from −0.46 to −0.83. In contrast, pH, NH3-N, and TP generally exhibited nonlinear relationships with DO, and the nonlinear coefficients were above 0.2. On the whole, the driving elements were different for each DO transience type. Hydrological elements mainly corresponded to DO swing transience. In contrast, biogeochemical elements were mainly related to DO abrupt transience. In general, this study provided a new perspective for revealing the high-frequency fluctuations of DO and corresponding drivers. The framework and results summarized in this study broadened our understanding on DO dynamics, and were conducive to river management, such as water quality anomaly detection, water quality prediction and regulation in advance.