Learning Dependence Relationships of Air Pollution and Meteorological Conditions in Yogyakarta Using Bayesian Network
Teny Handhayani, Jeanny Pragantha · 2024
Air pollution and climate change are global problems for the community. Historical data of meteorology is beneficial to study climate change. In Indonesia, studies related to the relationships between air pollution and meteorological conditions are insufficient. The paper aims to investigate the dependence relationships between air pollutants and meteorological conditions, a case study in Yogyakarta. It uses daily time series data from 19 August 2019 to 31 December 2023. The observed variables comprise particulate matter 10, carbon monoxide, sulfur dioxide, nitrogen dioxide, ground-level ozone, temperature, wind speed, humidity, rainfall, and sunshine duration. It runs the PC algorithm and Greedy Equivalence Search to generate two graphs represented by Completed Partially Directed Acyclic Graphs. An agreement graph is obtained from the learned graphs using the Intersection-Validation method to create a consensus graph. The consensus graph named agreement graph is used to approximate the true graph. The agreement graph shows dependent relationships between air pollution and meteorological conditions that occur between nitrogen dioxide and wind speed, nitrogen dioxide and humidity, and carbon monoxide and temperatures. The result is beneficial to guide further research on air pollution and meteorological conditions concerning climate change.