| Summary: | Nerus River is one of the major rivers in Terengganu, Malaysia. The river catchment area has undergone various degrees of land use changes during the last decade, including illegal logging, increased agriculture, ecotourism, and forestry. The activities have significantly affected the ecological and biological functions, soil erosion and runoff, in which they are believed to become one of the main causes that has increased the suspended solid concentration in this river. To date, a lack of comprehensive environmental studies has been conducted in investigating the chemical, physical, and microbiological composition in this river. Therefore, this study aims to develop a water quality prediction model ba ed on environmental degradation levels in Nerus River. Water quality parameters (physicochemical and biological) were measured using the standard methods from the Department of Environment (DOE), International Water
Quality Standards (NWQS), and Water Quality Index (WQI). Water samples were collected from eight sampling stations along Nerus River (every two months, during dry and wet seasons). Statistical Analysis techniques were employed namely, Descriptive Analysis, ANOVA, Correlation. Respectively, Multivariate statistical techniques namely (1) Cluster Analysis (CA), (2) Discriminant Analysis (DA), (3) Principal Component Analysis (PCA), (4) Multiple Linear Regression (MLR), and (5) Artificial Neural N tw rk (ANN) were employed with Geographic Information System (GIS) modelling f r land-use classification on Nerus River catclunent to explain the effects of complex data set variations on water quality variations, and to develop the water quality prediction model in the study area. A new equation model for water quality at Nerus River wa also developed ba ed on the most significant parameters in this river. The results of the water quality parameters clearly showed that Nerus River was categorized based on the pollution levels ranging from low (upstream), moderate (midstream), and polluted (downstream). Furthermore, WQI obviously indicated that the poor water quality had already been measured at the downstream stations, where the water quality fell under Class III (polluted) due to accumulated pollutants from the previous stations, di charge of water treatment plants, domestic wastewater from the city, houses in the rural areas housing surrounding the river, agricultural orchards, and run-offs. In contrast the data showed better prediction performance by using leave-oneout method, and the most effective is Model ANN-B since it gives the higher R2 value (0.971 and RMSE 1.092) compared to ANN-A and MLR where R2 value (0.996,0.778) and RMSE (0.642, 4.437), respectively. This study proved that ANN is undoubtedly
capable to be an alternative method to predict WQI rather than using conventional method (WQI equation), which is currently being used by the Department of Environment (DOE). The application of ANN-B model is successful in predicting WQI according to the significant parameters from DA in Nerus River basin, twenty-seven parameters and six parameters proposed by DOE-WQI. Accordingly, this study accentuates that the ANN constitutes an effective tool for the new equation model and comparative importance of inputs and outputs of the WQI, in comparison with MLR and other traditional WQI calculation methods, simplifying the computation of the WQI that may save substantial efforts and time by optimizing the calculation based on ANN.
|
|---|