A framework for assessing the adequacy of Water Quality Index - quantifying parameter sensitivity and uncertainties in missing values distribution

• Main Authors: Pak, Hui Ying, Chuah, C Joon, Tan, Mou Leong, Yong, Ee Ling, Snyder, Shane A
• Other Authors: Nanyang Environment and Water Research Institute
• Format: Article
• Language: English
• Published: 2021
• Subjects: Engineering::Environmental engineering; Water Quality Index; Multivariate Linear Regression
• Online Access: https://hdl.handle.net/10356/154614
• Institution: Nanyang Technological University

Water quality monitoring is a pillar in water resource management, but it can be resource intensive, especially for developing countries with limited resources. As such, Water Quality Indices (WQI) are developed to summarise general water quality, but efforts to assess the utility, flexibility, and practicality of WQI have been limited. In this study, we introduced an additional step to the traditional WQI development framework by introducing an adjusted form of WQI (WQIADJUSTED) to handle missing values, and capitalise on the remaining available information for the development of a WQI. A Sub-WQI was also developed to address local water quality conditions. WQI results (weighted and non-weighted) developed using different parameter optimisation methods, namely Multivariate Linear Regression and Principal Component Analysis were compared. To build upon the current framework, a new procedure was developed to assess the adequacy of WQI based on the sensitivity analysis of parameters and uncertainties associated with each parameter's missing values distribution. The number of observations needed for the development of a robust WQI was optimised with respect to user-defined acceptable change in WQI, based on Monte Carlo probabilistic simulation. The Johor River Basin (JRB), Malaysia is used as a case-study for the application of this new framework. The JRB serves as an important resource for Johor, one of the most populous state in Malaysia, and Singapore, a country south of Johor. WQIMLR performed better in explaining the general water quality than WQIPCA for weighted water quality parameters. Optimisation of sampling frequency revealed that around 130 samples will be required if a 2% change in WQI can be tolerated. The results (specific to the JRB) also revealed that total coliform is the most sensitivity parameter to missing values, and the distribution of sensitive parameters are similar for both WQINON-ADJUSTED and WQIADJUSTED.