Show simple item record

dc.contributor.advisorKumarasamy, Muthukrishna Vellaisamy.
dc.creatorBanda, Talent Diotrefe.
dc.date.accessioned2021-11-08T07:26:29Z
dc.date.available2021-11-08T07:26:29Z
dc.date.created2020
dc.date.issued2020
dc.identifier.urihttps://researchspace.ukzn.ac.za/handle/10413/19877
dc.descriptionDoctoral Degree. University of KwaZulu-Natal, Durban.en_US
dc.description.abstractThe assessment of water quality has turned to be an ultimate goal for most water resource and environmental stakeholders, with ever-increasing global consideration. Against this background, various tools and water quality guidelines have been adopted worldwide to govern water quality deterioration and institute the sustainable use of water resources. Water quality impairment is mainly associated with a sudden increase in population and related proceedings, which include urbanisation, industrialisation and agricultural production, among others. Such socio-economic activities accelerate water contamination and cause pollution stress to the aquatic environment. Scientifically based water quality index (WQI) models are then essentially important to measure the degree of contamination and advise whether specific water resources require restoration and to what extent. Such comprehensive evaluations reflect the integrated impact of adverse parameter concentrations and assist in the prioritisation of remedial actions. WQI is a simple, yet intelligible and systematically structured, indexing scale beneficial for communicating water quality data to non-technical individuals, policymakers and, more importantly, water scientists. The index number is typically presented as a relative scale ranging from zero (worst quality) to one hundred (best quality). WQIs simplify and streamline what would otherwise be impractical assignments, thus justifying the efforts of developing water quality indices (WQIs). Generally, WQIs are not designed for broad applications; they are customarily developed for specific watersheds and or regions unless different basins share similar attributes and test a comparable range of water quality parameters. Their design and formation is governed by their intended use together with the degree of accuracy required, and such technicalities ultimately define the application boundaries of WQIs. Such an academic gap is perhaps the most demanding scientific need; that is, to establish universally acceptable water quality indices, which can function in most, if not all the catchments in South Africa. In cognisance of such, the study suggests four water quality models that are not limited to specific application boundaries, and such contribution is significant, not only to the authors but to the entire nation. The first model, namely the universal water quality index (UWQI) developed based on conventional techniques using unequal weight coefficients and weighted arithmetic sum method. Model input parameters, relative weight coefficients and sub-index rating curves are established through an expert opinion by means of participatory based Rand Corporation’s Delphi Technique and extracts from literature. The second developed artificial intelligence (AI) in the form of artificial neural network (ANN) has three neuron layers parallel-distributed to accommodate feedforward sequence and backpropagation. The multi-layer perceptron model architecture includes nineteen highly interconnected neuro-nodes and seventy weighted synapses operating in a feedforward manner, from left to right. The study applied the Broyden-Fletcher-Goldfarb- Shanno (BFGS) algorithm to perform backpropagation training and optimising channel weights. The three-layered feedforward neural network indicated an increased performance registering an overall correlation coefficient of 0.985 and specific performance ratings of 0.987, 0.992 and 0.977 for training, testing and validation, respectively. The AI-based demonstrated an average target to an output error margin of ±0.242. Pointwise sensitivity analysis authenticated the robustness and computational aptitude of the suggested artificial neural network model. Both UWQI and ANN model functions with thirteen explanatory variables which are NH3, Ca, Cl, Chl-a, EC, F, CaCO3, Mg, Mn, NO3, pH, SO4 and turbidity (NTU). The third model entitled surrogate WQI works with four proxy water quality parameters comprising of chlorophyll-a, electrical conductivity, pondus Hydrogenium and turbidity. The proxy linear-based mathematical model is an abridged version of an outright WQI, purposefully stablished to substitute the UWQI and ANN model, thereby providing provisional index scores in the absence of extensive datasets. Water quality indices (WQIs) are customarily associated with massive data input demand, making them more rigorous and bulky. Such burdensome attributes are too taxing, time-consuming, and command a significant amount of resources to implement. Which discourages their application and directly influences water resource monitoring—making it increasingly indispensable to concentrate on developing compatible, more straightforward, and less-demanding WQI tools, but with equally matching computational ability. Surrogate models are the best fitting, conforming to the prescribed features and scope. Consequently, the study proposes an alternative water quality monitoring tool requiring fewer nputs, minimal effort, and marginal resources to function. Multivariate statistical techniques which include principal component analysis (PCA), hierarchical clustering analysis (HCA) and multiple linear regression (MLR) are applied primarily to identify four proxy variables and define relevant regression coefficients. Resulting in Chl-a, EC, pH and turbidity being the final four proxy variables retained. The selected input parameters are conformable with remote monitoring systems; which is a vital feature allowing the surrogate index model to be considered for remote monitoring programs. The fourth and final model suggested is a software-based water quality variability model (WQVM) established by integrating three distinctive water quality indices (WQIs) emerging from this study. The three WQIs are founded on different indexing methods, and they are documented as (a) universal water quality index, (b) artificial neural network, and (c) surrogate water quality index. Usually, most WQIs are presented as arithmetic formulas that are somewhat challenging to comprehend and apply in the real world. Therefore, the study attempts to address such research tendencies and set forth an excel-based hybrid water quality monitoring tool applicable at a national level. The WQVM enables the assessment of multiple water quality parameters, thereby solving practical water science problems. The proposed WQVM is earmarked for improving and promoting water quality monitoring programs, by providing a simple, convenient and userfriendly monitoring toolkit. Indeed, putting forward the WQVM has an increasing impact on water resource evaluation and optimising decision making amongst water scientists and professionals. Suggested models yield one-digit index values rending from zero to one hundred, where zero denotes poor water quality, and one hundred represents excellent water quality. Furthermore, the index scores are classified using a categorisation schema having five classes. Whereby “class one” with a possible maximum score of hundred designate the highest degree of purity and vice versa, “class five” signifies water quality of the lowest degree with index scores nearing or equal to zero. The WQIs and WQVM are developed and tested using water quality data from Umgeni Water Board (UWB) in KwaZulu-Natal Province, South Africa. From the original dataset, the current study retained 638 samples with 7,741 tests measured monthly over four years. The water quality records are from six sampling stations located within four different river basins identified as Umgeni, Umdloti, Nungwane and Umzinto/uMuziwezinto River catchments. The data samples are further curtailed to satisfy statistical requirements of each particular WQI model. All four models are considered robust and scientifically stable, with minor divergence from the ideal values. Better off, the prediction pattern matches the exemplary graph having comparable index scores and identical classification ranks, which signifies their readiness to appraise water quality status across South African watersheds. The established models symbolise a significant milestone with the prospects of promoting water resource monitoring and assisting in capturing spatial and temporal changes within river systems. The study intends to substantiate the methods used and document results achieved.en_US
dc.language.isoenen_US
dc.subject.otherWater resource monitoring.en_US
dc.subject.otherWater conservation.en_US
dc.subject.otherDrinking water quality.en_US
dc.subject.otherSustainable water managment.en_US
dc.subject.otherWater supply.en_US
dc.titleDevelopment of a universal water quality index and water quality variability model for South African river catchments.en_US
dc.typeThesisen_US
dc.description.notesPublications listed on page ii-iii.en_US


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record