In arid regions primary pollutants contribute to the increase of ozone levels, which cause negative effects on biotic health. This study investigates the use of abductive networks based on the group method data handling (GMDH) for ozone prediction. Abductive network models are automatically synthesized from a database of inputs and outputs. The models are developed for a location in the Empty Quarter, Saudi Arabia, first using only the meteorological data and derived meteorological data. In the subsequent efforts, NO and NO 2 concentrations and their transformations were incorporated as additional inputs. Another model forecasted ozone level after 1 h using mainly meteorological data, NO, and NO 2 concentrations. Models built for specific period of day are simpler compared to the generic models. Finally, ensemble modeling approach was also investigated. A time specific model produced mean absolute percentage error (MAPE) of 3.82%. The proposed models are self-organizing in nature and require less intervention from the users, and can be implemented easily by the interested practitioners.
Bibliographical noteFunding Information:
The authors would like to gratefully acknowledge the support of King Fahd University of Petroleum & Minerals in conducting this research. They are also grateful to the anonymous reviewers for their constructive reviews which enhanced the quality of the paper.
- Abductive networks
- Empty Quarter
- Group method data handling (GMDH)
- Machine learning
- Ozone modeling
ASJC Scopus subject areas
- Environmental Science (all)
- Atmospheric Science