Causal direction inference for air pollutants data

Air pollution is a serious problem in many places all over the world. Many efforts have been made to discover the causal relationships between air pollutants. Some of the air pollutants are highly correlated with others and environmental scientists detect the mutual transformations. This paper aims to reproduce the results of environmental scientists by computing the causal directions from the observational data instead of performing chemistry laboratory experiments. A causal direction inference method based on the Gaussian process model and the information geometric causal inference criterion is proposed. Simulations show satisfactory results on air pollutant data collected by automatic monitoring stations.