Concentration estimation of dissolved oxygen in Pearl River Basin using input variable selection and machine learning techniques.

Dissolved oxygen (DO) concentration is an essential index for water environment assessment. Here, we present a modeling approach to estimate DO concentrations using input variable selection and data-driven models. Specifically, the input variable selection technique, the maximal information coefficient (MIC), was used to identify and screen the primary environmental factors driving variation in DO. The data-driven model, support vector regression (SVR), was then used to construct a robust model to estimate DO concentration. The approach was illustrated through a case study of the Pearl River Basin in China. We show that the MIC technique can effectively screen major local environmental factors affecting DO concentrations. MIC value tended to stabilize when the sample size >3000 and EC had the highest score with an MIC >0.3 at both of the stations. The variable-reduced datasets improved the performance of the SVR model by a reduction of 28.65% in RMSE, and increase of 22.16%, 56.27% in R2, NSE, respectively, relative to complete candidate sets. The MIC-SVR model constructed at the tidal river network performed better than nontidal river network by a reduction of approximately 63.01% in RMSE, an increase of 62.36% in NSE, and R2 >0.9. Overall, the proposed technique was able to handle nonlinearity among environmental factors and accurately estimate DO concentrations in tidal river network regions.

The change of representative elementary volume of DNAPL influenced by surface active agents during long-term remediation period in heterogeneous porous media.

Representative elementary volume (REV) is important to characterize dense nonaqueous phase liquids (DNAPLs) during surfactant-Enhanced aquifer remediation (SEAR) period. To investigate the REVs of DNAPL in remediation, a perchloroethylene (PCE) SEAR experiment is conducted in a two dimensional (2D) heterogeneous translucent porous media. Light transmission techniques are used to quantify PCE saturation (Soil) and PCE-water interfacial area (AOW). Afterward, corresponding REVs are estimated using a criterion of relative gradient error (εgi) to reveal the change of REVs of DNAPL over the entire remediation period. Results from this work suggest the presence of surface active agents strongly affect the REVs of DNAPL. At the beginning of the SEAR experiment, the frequency of minimum Soil-REV size closely follows a Gaussian distribution in 0.0mm-11.0mm. Simultaneously, the frequency of minimum AOW-REV size is close to a Gaussian distribution in 2.0mm-9.0mm and appears a peak value in 13.0mm-14.0mm. As SEAR experiment proceeds, both the shapes of frequency and cumulative frequency of REV sizes are changed. At the end of SEAR experiment, the frequency of minimum Soil-REV and minimum AOW-REV size tend to Gaussian distributions in 0.0mm-6.0mm and 0.0mm-9.0mm, respectively, which suggest both minimum Soil-REV size and minimum AOW-REV size show decreasing tendency. Continuous quantification of the REVs of DNAPL is realized in this study to reveal the change of REVs influenced by surface active agent. The finding has important significance on improving our understanding of the characteristics of DNAPL in SEAR process, simulating DNAPL remediation and designing appropriate remediation scheme with high-resolution.