Scholarlink

<p>Accurate prediction of dissolved oxygen (DO) dynamics in rivers is crucial for effective water quality management but remains challenging due to the complex nonlinear mechanisms governing river systems. This study introduces a chaotic dynamic-based approach for prediction of DO concentrations. Daily DO data observed over a period of 15&#xa0;years (2009–2023) at 100 monitoring stations across the United States are analysed. The prediction approach consists of three primary steps: (i) reconstruction of the univariate DO time series in a multi-dimensional phase space using delay embedding; (ii) identification of ‘nearest neighbors’, based on their proximity in the reconstructed space; and (iii) prediction of DO values by tracking the temporal evolution of these neighbors. The first 12&#xa0;years of data (2009–2020) are used for phase-space reconstruction and the remaining 3&#xa0;years (2021–2023) for checking the prediction accuracy. A sensitivity analysis is conducted by exploring different values of embedding dimensions and nearest neighbors. The prediction performance is evaluated using the Nash–Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), and the Pearson Correlation Coefficient (CC). The results indicate consistently strong predictive capability of the method across the 100 stations, with NSE values in the range 0.532–0.995, RMSE in the range 0.148–1.984&#xa0;mg/L, and CC in the range 0.735–0.997. Notably, high predictability is achieved with relatively low embedding dimensions, indicating that DO dynamics are governed by only a few dominant governing mechanisms. The present results are encouraging regarding the utility of chaos-based methods for DO prediction in rivers.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of dissolved oxygen dynamics using a nonlinear local approximation approach

  • 期刊论文

  • 发表日期：

    2026年5月2日

查看全文
  • Sakshi Dnyaneshwar Dhumale,
  • Bellie Sivakumar

摘要

Accurate prediction of dissolved oxygen (DO) dynamics in rivers is crucial for effective water quality management but remains challenging due to the complex nonlinear mechanisms governing river systems. This study introduces a chaotic dynamic-based approach for prediction of DO concentrations. Daily DO data observed over a period of 15 years (2009–2023) at 100 monitoring stations across the United States are analysed. The prediction approach consists of three primary steps: (i) reconstruction of the univariate DO time series in a multi-dimensional phase space using delay embedding; (ii) identification of ‘nearest neighbors’, based on their proximity in the reconstructed space; and (iii) prediction of DO values by tracking the temporal evolution of these neighbors. The first 12 years of data (2009–2020) are used for phase-space reconstruction and the remaining 3 years (2021–2023) for checking the prediction accuracy. A sensitivity analysis is conducted by exploring different values of embedding dimensions and nearest neighbors. The prediction performance is evaluated using the Nash–Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), and the Pearson Correlation Coefficient (CC). The results indicate consistently strong predictive capability of the method across the 100 stations, with NSE values in the range 0.532–0.995, RMSE in the range 0.148–1.984 mg/L, and CC in the range 0.735–0.997. Notably, high predictability is achieved with relatively low embedding dimensions, indicating that DO dynamics are governed by only a few dominant governing mechanisms. The present results are encouraging regarding the utility of chaos-based methods for DO prediction in rivers.