Statistical modeling of surface water quality in the Ave River Basin

• Autor(a) principal: Fernandes, António Carlos Pinheiro
• Data de Publicação: 2022
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10348/11224
• topic: Ave River Basin; Surface water quality

Water quality is defined as the set of physical, chemical and biological parameters that allow its use for a specific purpose. The civilization evolution is accompanied by the contamination of water resources due to anthropogenic activities. To mitigate and prevent pollution, it is necessary to carry out systematic research to understand the cause-effect interactions and predict the state of water resources. With the implementation of the Water Framework Directive, there has been more research in the scope of water quality to achieve the directive goals. At the moment, 2027 is the target year to reach the good ecological status of all surface waters in Europe. In Portugal, a large part of the surface waters require improvement to fulfil the objective. In this sense, this dissertation was carried out to study the water quality of the Ave River Basin, given its history of pollution, which is still problematic. With an environmental dataset comprising water quality data, point-source pressures, diffuse pressures and landscape metrics, a sequence of statistical studies were performed, with the main objective of understanding which are the most unsettling pressures. It was chosen statistical models rather than physical because in statistical approaches new discoveries can be found, since models are driven by the dataset. It is highlighted that the created models for the Ave River Basin can be applied to any other basin, it is of national interest that the developed studies become reproduced in other basins/hydrographic regions. The first chapter presents a brief introduction about the main sources of contamination on surface water and the approach to this problem. The second chapter presents a study of cause-effect interactions through structural equation models. It demonstrates the dynamics between point pressures, diffuse pressures, diffuse indicators with surface water contamination and effects on ecological integrity in two river basins. The created models for the Ave River Basin showed better results than for Sabor, since the determination coefficient was higher, and the cause-effect relationships were according to theoretical expectation. Thus, the results revealed that point source pressures and livestock production are decreasing Ave River Basin water quality. The third chapter presents an adapted structural equation model for water quality prediction, according to the future scenarios presented in the 2nd cycle of the River Basin Management Plans. This analysis was performed to identify if, even under an environmentally optimistic scenario, the ecological status of surface waters will improve. For that scenario, it is confirmed that the water quality can improve. Still, IPtIN scores will not achieve the highest categories, indicating that the WFD goal of achieving a good or higher ecological status category for European surface waters until 2027 will not be achieved. In the fourth chapter, land-use metrics are approached in order to verify the relationship between landscape and water quality. In multiple structural equation models, contaminant emissions are compared with the effect of land configuration, in an analysis that comprises different sampling scales, also verifying the differences in these relationships between two seasons. By comparing the weights of measured variables and path coefficients of latent variables, it was observed that the adopted scale is important to trace contamination and landscape effects, such as seasons. It was also noted that the sampling scale, entire drainage area, is the scale that resulted in the highest determination coefficients, so it was decided to use that scale for further studies. Upon discovering that the landscape is a determining factor in the water quality of Ave River Basin, the following question arises: has the landscape always affected this river basin, whose water quality is historically impacted by point source pressures? In Chapter 5, this question is answered, demonstrating which landscape metrics have a dominant effect over about 30 hydrological years, as well as the most unsettling metrics regarding landscape changes. It was concluded that landscape has always been related to Ave River Basin water quality and that interventions and strategies to ensure that landscape does not evolve in a tendency that decreases water quality are measures to adopt. As the landscape has always affected the water quality, it remains to apply the landscape metrics in spatial prediction models of water quality parameters. In chapter 6, linear regression and spatially weighted regression models are demonstrated, in which several regressors were tested, including landscape metrics, point source and diffuse pressures. As a by-product of this task, a Python algorithm was developed to find the best linear regressions within a dataset, regarding the statistical assumptions. By applying the algorithm to the case study, it was possible to predict pH (highest performance), nitrates, total alkalinity, and electrical conductivity (worst performance). For other parameters, namely oxygen demands, it was not obtained regressions regarding statistical assumptions. In chapter 7 it is presented a decision support tool, elaborated through two combined multi-criteria analyses, in which sub-basins are prioritized for environmental intervention, regarding the contamination risk and intervention complexity. The prioritization analysis was adequate for the case study, since by crossing the intervention priority with the chemical and ecological status, the sub-catchments with lower quality were the ones with the highest priority. Finally, in chapter 8, the main conclusions and possible future studies are presented.