Data Imputation for Spatial and Time Series

• Grant number: 25/20963-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: November 01, 2025
• End date: October 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Statistics
• Principal Investigator: Aluísio de Souza Pinheiro
• Grantee: Júlia Pedroso Leal
• Host Institution: Instituto de Matemática, Estatística e Computação Científica (IMECC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated research grant: 23/02538-0 - Time series, wavelets, high dimensional data and applications, AP.TEM

Abstract

Remote sensing is one area in which time series have both time and spatial dynamics. This kind of complex stochastics is basal for several applications in environmental sciences. For instance such data is used in change detection, natural disaster monitoring and sustainable agriculture [Krichen et al.(2024)]. Notwithstanding its applicability, there are diverse technical issues in data acquisition, and regular spatial and temporal sampling is rarely attained. The current major models for time series and/or spatial time series are based on regular data sampling (monthly, weekly, daily, hourly etc). This means that data imputation is usually performed for the ubiquitous missing data points. Moreover, these procedures are done marginally in space or time [Hamdi et al.(2022)]. We seek performing data imputation on both time and space simultaneously. For this artificial intelligence methods may be a numerically costly but statistically efficient solution [Gond et al.(2021)]. The deep learning options which will be studied in this project are: convolutional neural networkds (CNN) and recurrent neural networks (RNN), as can be seen in [Goodfellow et al.(2016)]. The project will focus on the performance of environmental statistical methods for remote sensing based on deep learning spatial-temporal data imputations.