Maestría en Analítica Aplicada · 2026

Análisis de variables que influencian la calidad del aire en Bogotá: intervención y medio ambiente

Particulate matter 10 (PM10.) pollution poses a public health risk in cities such as Bogotá, where meteorological factors influence its concentration. This study aims to model and predict PM10 levels using machine learning models and data from eight air quality monitoring stations located throughout the city, to identify the variables that most strongly affect pollution levels. Meteorological variables such as temperature, wind speed and direction, relative hu midity, and precipitation were obtained from the Bogotá Air Quality Monitoring Net work application. Three predictive approaches were evaluated: a basic model based on hourly averages, linear regression, and ExtraTreesRegressor, using metrics such as RMSE and MAE. To understand the influence of these variables on PM10 concen tration, interpretability methods such as SHAP and PDP were applied. The results showed that lagged variables—particularly wind speed and wind direction with a one hour delay—are associated with changes in pollution levels, allowing the model to capture relevant patterns that explain variations in pollutant concentration. Likewise, temporal variables such as hour, month, and year help capture information that me teorological variables alone cannot. Additionally, these seasonal behaviors are often influenced by regional dynamics, in cluding the transport of pollution from areas with a high occurrence of wildfires and other external sources that intensify during certain periods of the year, which helps explain the variations observed in PM10. levels within the city.