Pattern recognition approach for spatial and temporal variation analysis of surface water quality of Klang River Basin, Malaysia
In Malaysia, the hydrology and surface water quality of a river are often discussed among the city dwellers especially in the urban area as a consequence from the flash flood scenario during the rain event seasons. This study was conducted in Klang River in attempt to interpret the relationship betw...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/67214/1/FPAS%202013%2020%20IR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In Malaysia, the hydrology and surface water quality of a river are often discussed among the city dwellers especially in the urban area as a consequence from the flash flood scenario during the rain event seasons. This study was conducted in Klang River in attempt to interpret the relationship between hydrological and surface water quality parameters, to estimate the pollution loading in the river with and without the utilization of the hydrological data and to ascertain the input and output parameters of the water quality data using the artificial neural network (ANN). The data was collected from the Department of Environment (DOE) and Department of Irrigation and Drainage (DID), Malaysia. Approximately 30 water quality parameters, rainfall, discharge and direct runoff data were considered for data analysis. This study integrates statistical tools such as non-parametric analysis, multivariate methods and ANN modeling in order to describe the spatial and temporal variation of water quality at Klang River Basin. This study revealed that among the 24 rainfall stations, Rs11 and Rs14 indicate the highest variations of rainfall with Rs02 exhibiting a downward trend with rainfall while the other stations indicate no trend. Among all the variables, temperature (TEMP) and mercury (Hg) were indicate the highest correlation value at p<0.05 (R=-0.245 and R=0.295) with rainfall while other stations have no significant relationship with rainfall. Despite that, river discharge illustrate Ds03 and Ds04 having an upward trend with only dissolved oxygen (DO), ammoniacal nitrogen (NH3N), TEMP, conductivity (COND), salinity (SAL), total solid (TS), chloride (Cl-), phosphate (PO43-), magnesium (Mg) and methylene blue active substances (MBAS) significantly (p<0.05) correlated with discharge and direct runoff. Source identification using principal component analysis (PCA) confirmed that an anthropogenic activity does influence the river water quality. The clustering of 30 monitoring stations exhibit 4 clusters categorizes as Clean (C), Moderately Polluted (MP), Polluted (P) and Highly Polluted (HP) as the spatial factor meanwhile seasonal Wet, Dry and Transitional (Trans) and water level Low (L), Normal (N) and High (H) as temporal factors. The spatial (92%), seasonal (93%) and water level (99%) are correctly assigned using standard mode, forward stepwise and backward stepwise mode of discriminant analysis (DA). PCA recognized 22 parameters in the C groups. Meanwhile, all the parameters were identified in MP and P except for arsenic (As), zinc (Zn), iron (Fe), oil and grease (OG)) and (NH3N, PO43- and MBAS). These pollutants are mainly comes from soil erosion, anthropogenic, household, domestic and industrial wastewater and sullage. Based on the source identification and absolute principal component score-ANN (APCS-ANN) model, industrial and domestic wastewaters effluent are categories as pollution source 1 (PS1) and notified as the highest contributor (33%) while flood mitigation activities and seasonal effect (PS5) as the lowest contributor (1%). The ANN-DA models revealed better prediction performance using ANN-DA Seasonal (ADSe) model due to strong correlation (R=0.9871) with low root mean square error (RMSE) and sum of squares error (SSE) values compared to the other 2 models. This study caters an integrated picture for government agencies to solve the large complex datasets into a more comprehensive and systematic approach. This multivariate statistical technique provides substantial information and knowledge on the identification of pollution sources for improvement and maintenance of the watershed management practices in Malaysia. Hence, assist the Klang River Basin management to meet the criteria required by the authorities, especially DOE and DID. |
|---|