Influence of the spatial structure of the geosystems on the water quality parameters regionalization: a case study in the upper Piracicaba-Jaguari river watershed, southeastern of Brazil

Author(s)

Danilo Francisco Trovo Garofalo

• Document type: Doctoral Thesis
• Press: Campinas, SP.
• Institution: Universidade Estadual de Campinas (UNICAMP). Instituto de Geociências
• Defense date: 2017-12-07

Examining board members

Marcos César Ferreira; Sueli Yoshinaga Pereira; Veraldo Liesenberg; Archimedes Perez Filho; Andréia Medinilha Pancher

• Advisor: Marcos César Ferreira

Abstract

The physical, chemical and biological composition of water from a certain river system is driven by its geomorphological, geological, pedological, climate, natural vegetation as well as by the land use and land cover patterns of the selected watershed. The water quality can be checked by analyzing different water quality parameters (PQA). However, due to the high cost involved, frequent analyzes of the different water quality parameters are performed only for a given number of Brazilian river systems. Therefore, there is a current need for both techniques and alternative methods for analyzing the water quality parameters. This doctoral dissertation presents a methodological procedure to forecast, to spatialize and to regionalize water quality parameters in small watersheds without in-situ measurements. This procedure was performed by quantifying biophysiographical characteristics of watersheds (4st order basins) by using spatial analysis and remote sensing techniques, integrating water quality parameter data from regression methods. The selected study area is the high Piracicaba-Jaguari river watershed, located in the southern part of the Minas Gerais State. Belonging to the Cantareira System, this particular watershed encompasses the main spring sources of the Piracicaba River - of great importance for the water supply of several cities of the São Paulo State and the metropolitan region of the São Paulo city itself. The suggested prediction was based on the survey of 27 biophysiographical variables (explanatory variables) in 44 small watersheds and 12 water quality parameters (response variables) validated in 12 watersheds, in triplicate, in four seasons during the year, accounting therefore for 144 analyzes for each water quality parameter. A total of 120 samples were selected for training (83.33%) and 24 for validation (16.67%). The non-parametric regression algorithms such as the Support Vector Regression (SVR) and K-Nearest Neighbors Regression (K-NNR) were used as predictor models to estimate PQA values for the 32 small watersheds. The performance of the SVR and k-NN models was evaluated using the mean absolute error, the mean absolute percentage error, and the coefficient of determination (R2). The results show that the K-NNR model presented the best performance. The results also show that the regression models highlighted the ability to identify water quality parameters from biophysiographical variables. After, water quality regions from similar small watershed groups were mapped based on the set of characteristics of the water quality parameters and the biophysiographical variables of the 44 small watersheds. For the water quality region, correlational similarity and distances were used. The assumptions of the Ward's hierarchical method were used for the analysis of the homogeneous groups of watershed (AU)

• FAPESP's process: 13/22185-2 - INFLUENCE OF GEOSYSTEMS SPATIAL STRUCTURE ON THE REGIONALIZATION OF THE WATER QUALITY PARAMETERS IN PIRACICABA- JAGUARI UPPER BASIN
• Grantee: Danilo Francisco Trovo Garofalo
• Support Opportunities: Scholarships in Brazil - Doctorate