Water Quality Assessment Along the Yangtze River Mainstream-Nonparametric Statistical Approach.pdf

1、The nonparametric statistical water quality trend analysis method, Seasonal Mann Kendall Analysis (SMKA), was developed in the 1980's and applied on different river systems. The method was reported to be powerful in dete

2、cting monotonic trends even when the statistical characteristics of the water quality data have been against the requirements of conventional parametric methods such as regression analysis. These characteristics include

3、outliers, missing observations, censored observations, seasonality, serial correlation, and covariate effect of flow. The method is robust to outliers, missing observations and censored observations.In this thesis, the p

4、ossibility of bringing those experiences to the river systems of China was assessed by applying the method to water quality data of the Yangtze River. To that end, three water quality stations and five water quality vari

5、ables were selected along the mainstream of the Yangtze River using some selection criteria. Statistically increasing trends in concentrations of the five water quality variables at the three stations were detected over

6、their respective study periods.Screening water quality data (in terms of outliers, missing observations, and censored observations) and characterizing water quality data (in terms of seasonality, serial correlations, and

7、 monotonic behavior) have been given great emphasis as it determines the applicability and the trend detection power of the method. Although the SMKA is quantitatively explained in statistical terms, there are still prob

8、lems in data screening and characterization. In this paper a graphical approach to identify outlying values is introduced.Though there is a statistical method to test the presence of significant nonmonotonic and/or signi

9、ficant nonlinear trend about a change point, so far there is no method to select the 'true' change point when significant non-monotonic trends were detected at multiple change points. Here in this study, however, criteri

10、a were proposed to select the 'optimum' change point among the possible change points.Besides assessing trends in water quality at different stations, comparisons of water quality among stations over a comparison period

11、was done using mean annual flow adjusted concentrations and trend magnitudes. Water quality (in terms of the selected parameters) at the lower reach station was relatively better than the upper reach and middle reach sta

12、tions over the comparison period.An assessment of source of pollution, based on concentration-flow relations, has shown that ground water is the main source of dissolved solids. Point source in the dry season and non-poi

13、nt source in the rainy season were found to be the causes of dissolved oxygen depletion (assuming depletion due to temperature variation is minimal). No apparent source of nitrate was identified at the upper reach statio

14、n while possibility of non point source found at the lower and upper reach stations.The MATrix LABoratory, MATLAB, capability for technical computation was used for this statistical water quality assessment. The statisti

15、cal toolbox and program development environment together with built in statistical functions were major components of MATLAB used in this study. MATLAB is a highperformancelanguage for technicalcomputing.It integrates co

16、mputation,visualization, and programming in an easy-to-use environment where problems and solutions are expressed in familiar mathematical notation.This study reveals that SMKA can be applied to the river system of China

17、 as far as the water quality time series data satisfies the requirements of the method.The paper also illustrates the importance of screening and characterizing water quality data before selecting trend analysis methods