Banca de QUALIFICAÇÃO: RUBENS OLIVEIRA DA CUNHA JUNIOR

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
DISCENTE : RUBENS OLIVEIRA DA CUNHA JUNIOR
DATA : 17/12/2021
HORA: 14:00
LOCAL: PLATAFORMA GOOGLE MEET
TÍTULO:

EVALUATION OF MODELING TECHNIQUES, IMPUTATION AND TIME SERIES FORECAST OF GROUNDWATER LEVELS

PALAVRAS-CHAVES:

Combination of predictors. Artificial neural networks. Support Vector Regression. Araripe Sedimentary Basin. Sustainable development.

PÁGINAS: 47
GRANDE ÁREA: Outra
ÁREA: Ciências Ambientais
RESUMO:

Groundwater is one of the main sources of water in several countries, mainly in arid and semi-arid regions. The rational use of this resource is essential for sustainable regional development. In this context, the modeling and prediction of groundwater dynamics are essential for the assessment of water availability. However, groundwater data are limited in many regions, and periods without available information are common, that is, with failures in data from monitoring processes. To solve the problem, one can resort to techniques to estimate such missing values, known as fault filling or missing values imputation methods. Attribution is an important step that precedes modelling. Formalisms of time series modeling and prediction are a robust alternative for inferring groundwater levels, especially with the advent of machine learning techniques. The objective of this work is to investigate the application of modeling, imputation and time series forecasting techniques to assess the groundwater level. The study area is the Araripe Sedimentary Basin, south of the State of Ceará, Brazil, the largest underground water reserve in Ceará and the main source of water supply in the Cariri Region. The dataset consists of historical series of monitoring well levels installed over the Araripe basin aquifers, which are part of the Integrated Groundwater Monitoring Network (RIMAS/CPRM). The data will be treated for missing values, considering state-of-the-art methods of imputation in time series, and proposing imputation methods, such as imputation via copulas. Individual and combined modeling and forecasting approaches will be used. Among the individual, stand out Autoregressive Integrated Moving Averages (ARIMA) and Exponential Smoothing (ETS) linear stochastic models, and models based on machine learning, such as Artificial Neural Networks (ANN) and Support Vector Regression (SVR). The combination will be through techniques such as Simple Mean (SA), Simple Median and Minimum Variance (MV), in addition to the combination via RNA and SVR. The performance performance of the models will be compared according to error metrics well known in the literature.

MEMBROS DA BANCA:
Interno - 1643899 - CARLOS WAGNER OLIVEIRA
Interno - 1549937 - CELME TORRES FERREIRA DA COSTA
Presidente - 1751179 - PAULO RENATO ALVES FIRMINO
Externo à Instituição - PAULO SALGADO GOMES DE MATTOS NETO - UFPE