Assessment of missing value imputation methods in timeseries of air pollutants in the metropolitan area of Sao Paulo

• Grant number: 24/03813-7
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: May 01, 2024
• End date: April 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Meteorology
• Principal Investigator: Luciana Varanda Rizzo
• Grantee: Daniel Morais Trojan
• Host Institution: Instituto de Física (IF). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Air pollution is one of the main environmental problems in urban areas, with negative impacts to human health and quality of life. Air quality monitoring is essential to quantify impacts and to develop mitigation measures. The metropolitan area of São Paulo (MASP) is relatively well served of governmental air quality monitoring stations. Even tough, the data coverage has flaws, like: missing values due to operational procedures and instrument faults; incomplete set of monitored variables; spatial distribution of stations. These flaws affect the air quality characterization and the development of predictive, epidemiologic and valoration models. This undergrad research project aims to assess different missing value imputation methods in time series of air pollutant concentrations in the MASP. Conventional statistical methods, machine-learning methods, uni and multivariate methods will be analyzed. Daily data from 10 air quality stations in the MASP, from 2000 to 2023, will be used in ths project. The main focus is on inhalable particle matter (PM10), although other pollutants will be considered in the analysis. Long term trends, seasonal cycles and correlations will be assessed, supporting the application of missing value imputation methods. Artificial missing values will be inserted in the timeseries to evaluate the performance of the different imputation methods, under a supervised learning framework. Pros and cons of each method will be assessed, aiming to propose an imputation strategy suitable to the MASP conditions. This analysis is unprecedent for the MASP and may contribute to the development of a complete and homogeneous air quality data base.