Prediction Ground-Level Ozone Concentrations Using Statistical and Machine Learning Models: Izmir, Türkiye

Öz

Ozone (O₃) is a greenhouse gas that poses significant risks to public health, particularly in large cities, and contributes to global warming. This study aims to identify ozone trends and forecast concentration values based on daily concentration data measured in Izmir, in western Turkey, between 2024 and 2025. The analysis is based on predicting the next day’s ozone concentrations using meteorological parameters (temperature, wind speed, relative humidity, precipitation, cloud cover, sunshine duration, and solar radiation) and nitrogen oxide concentrations. For the prediction, Multiple Linear Regression (MLR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM) models were used. To improve forecasting performance, lagged models were developed and evaluated. Model performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE) metrics. The study revealed that model prediction performance was low when only meteorological parameters and NOx concentrations were used as input variables. To address this, a two-stage design was applied to the models, consisting of a 3-day and 7-day lookback followed by a 1-day lookahead. The highest accuracy in ozone forecasting was achieved with the MLR model, yielding R² = 0.82, RMSE = 6.50, and MAE = 4.73. The feature importance analysis derived from the models identified both O3_lag3 and O3_lag7 as the most effective predictors. The use of machine learning and deep learning algorithms is recommended for more complex scenarios and situations with large datasets. Additionally, it is anticipated that incorporating traffic density and volatile organic compound concentrations into O3 prediction models will yield better performance metrics.

Anahtar Kelimeler

• Ozone prediction
• Izmir
• linear regression
• machine learning

İstatistiksel ve Makine Öğrenmesi Modelleri Kullanılarak Yer Seviyesi Ozon Konsantrasyonunun Tahmin Edilmesi: İzmir, Türkiye

Öz

Ozon (O3), özellikle büyük şehirlerde halk sağlığı açısından önemli risklere ve küresel ısınmaya neden olan sera gazıdır. Bu çalışma, 2024 ile 2025 yılları arasında Türkiye'nin batı kısmında İzmir’de ölçülen günlük konsantrasyon verilerinden ozon eğilimlerini belirlemeyi ve konsantrasyon değerlerini tahmin etmeyi amaçlamaktadır. Analiz, meteorolojik parametrelere (sıcaklık, rüzgar hızı, bağıl nem, yağış, bulut örtüsü, güneşlenme süresi ve güneş radyasyonu) ve azot oksit konsantrasyonlarına dayalı olarak bir gün sonraki ozon konsantrasyonlarını tahmin etmeye dayanmaktadır. Tahmin için; Çoklu Doğrusal Regresyon (MLR), Rastgele Orman (RF), Aşırı Gradyan Artırma (XGBoost) ve Uzun Kısa Süreli Bellek (LSTM) modelleri kullanılmıştır. Tahmin performansını artırmak için, gecikmeli modeller geliştirilmiş ve değerlendirilmiştir. Model performansı, determinasyon katsayısı (R²), kök ortalama kare hatası (RMSE) ve ortalama mutlak hata (MAE) metrikleri kullanılarak değerlendirilmiştir. Çalışma; sadece meteorolojik parametreler ve NOx konsantrasyonlarının girdi değişkeni olduğu durumlarda modellerin tahmin performanslarının düşük olduğunu ortaya koymuştur. Bu amaçla modellerde 3 ve 7 gün geriye bakış → 1 gün ileriye bakış olmak üzere iki aşamalı tasarım uygulanmıştır. Ozon tahmininde en yüksek doğruluk R2= 0,82, RMSE=6,50 ve MAE=4,73 değerleriyle MLR modeliyle elde edilmiştir. Modellerden elde edilen özellik önem analizi, hem O3_lag3'ü hem de O3_lag7'yi en etkili tahmin edici olarak belirlemiştir. Makine ve derin öğrenme algoritmalarının daha karmaşık ve veri setinin çok olduğu durumlarda kullanması önerilmektedir. Ayrıca O3 tahmin modellerinde trafik yoğunluğu ve uçucu organik bileşik konsantrasyonlarının modellere dahil edilmesinin daha iyi performans metrikleri sunması öngörülmektedir.

Anahtar Kelimeler

• Ozon tahmini
• İzmir
• doğrusal regresyon
• makine öğrenmesi

Kaynakça

1. [1] Ding, J. Ren, C. Wang, J. Feng, Z. and Cao, S.-J. Spatial and temporal urban air pollution patterns based on limited data of monitoring stations, Journal of Cleaner Production, 434, p. 140359, 2024; doi: 10.1016/j.jclepro.2023.140359.
2. [2] Zhou, Z. Qiu,C. and Zhang,Y. A comparative analysis of linear regression, neural networks and random forest regression for predicting air ozone employing soft sensor models. Scientific Reports, vol. 13(1), 1–23, 2023; doi: 10.1038/s41598-023-49899-0.
3. [3] Latif, S. D., Lai, V., Hahzaman, F. H., Ahmed, A. N., Huang, Y. F., Birima, A. H., & El-Shafie, A. Ozone concentration forecasting utilizing leveraging of regression machine learnings: A case study at Klang Valley, Malaysia. Results in Engineering, vol. 21, p. 101872, 2024; doi: 10.1016/j.rineng.2024.101872.
4. [4] Demir S. and Saral,A. Identification and Apportionment of Sources of Ozone-forming Potential for Proper Reduction Strategies CLEAN – Soil, Air, Water. 41(2), 107–112, 2013; doi: 10.1002/clen.201100704.
5. [5] Faizan, M. and Çelikten, H. Mitigation of Ozone (O₃) Effects and Enhanced Air-Purifying Capacity of Plants by Carbon-Based Nanomaterials. MAS Journal of Applied Sciences. 11(1), 168–180, 2026; doi: 10.5281/zenodo.19127754.
6. [6] He, Z. Wu, Z. Herzog, O. Hei, J. Li, L. and Li, X. Compound health effects and risk assessment of extreme heat and ozone air pollution under climate change: A case study of 731 urban areas in China. Sustainable Cities and Society, vol. 119, p. 106084, 2025; doi: 10.1016/j.scs.2024.106084.
7. [7] Tang, H. Bhatti, U. A. Li, J. Marjan, S. Baryalai, M. Assam, M., ... & Mohamed, H. G. A New Hybrid Forecasting Model Based on Dual Series Decomposition with Long-Term Short-Term Memory. International Journal of Intelligent Systems, 2023(1), 9407104, 2023; doi: 10.1155/2023/9407104.
8. [8] Karmakar, S. Baul Das, A. Gurung, C. and Ghosh,C. Effects of Ozone on Plant Health and Environment: A Mini Review, Res. Jr. Agril. Sci, 13, 612-619. 2022.

9. [9] He W. and Chen, H. More extreme-heat occurrences related to humidity in China. Atmospheric and Oceanic Science Letters, 16(5), 100391, 2023; doi: 10.1016/j.aosl.2023.100391.
10. [10] Ajayi,F. Adewale, A. and Osho, M. Integrating Engineering Innovations to Enhance Environmental Resilience: Evaluating the Impact of Greenhouse Gas Emissions, Ozone Depletion, And Aquatic Ecosystem Degradation on Public Health, International Journal of Computer Applications Technology and Research, 13, 16–36, 2024; doi: 10.7753/IJCATR1310.1002.
11. [11] Nuvolone,D. Petri, D. and Voller, F. The effects of ozone on human health. Environ Science and Pollution Research, 25(9), 8074–8088, 2018; doi: 10.1007/s11356-017-9239-3.
12. [12] Aljafen, B. N. Shaikh, N. AlKhalifah, J. M. & Meo, S. A. Correction: Effect of environmental pollutants particulate matter (PM2.5, PM10), nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon monoxide (CO) and ground level Ozone (O3) on epilepsy.,” BMC Neurology, 25(1), 1–1, 2025; doi: 10.1186/s12883-025-04142-3.
13. [13] Eslami, E. Choi,Y. Lops, Y. and Sayeed, A. A real-time hourly ozone prediction system using deep convolutional neural network. Neural Computing & Applications, 32(13), 8783–8797, 2020; doi: 10.1007/s00521-019-04282-x.
14. [14] Katipoğlu, O. M. Elshaboury, N. Kartal, V. Ertugay, N. Kilinc, H. C. Şenocak, S. & Pande, C. B. Application of Time Series Methodologies for Robust Forecasting of Atmospheric Pollutant Concentrations: PM10, SO2, NO2, NOx, and O3 in an Urban Environment. Environmental Quality Management, 35(2), 70206, 2025; doi: 10.1002/tqem.70206.
15. [15] Çelikten, H. A Multi-Output Deep Learning Framework for Simultaneous Forecasting of PM10 and Air Quality Index in High-Altitude Basins: A Case Study of Igdir, Türkiye. Sustainability, 18(8), 2026; doi: 10.3390/su18083883.
16. [16] Yılmaz, A. Ozone Level Prediction with Machine Learning Algorithms. Journal of Aeronautics & Space Technologies / Havacilik ve Uzay Teknolojileri Dergisi, 14(2), 177–183, 2021.
17. [17] Carbo-Bustinza, N. Iftikhar, H. Belmonte, M. Cabello-Torres, R. J. De La Cruz, A. R. H. and López-Gonzales, J. L. Short-Term Forecasting of Ozone Concentration in Metropolitan Lima Using Hybrid Combinations of Time Series Models. Applied Sciences, 13(18), 2023; doi: 10.3390/app131810514.
18. [18] Birim, N. G. Turhan, C. Atalay, A. S. and Gokcen Akkurt G. The Influence of Meteorological Parameters on PM10: A Statistical Analysis of an Urban and Rural Environment in Izmir/Türkiye. Atmosphere, 14(3), 421, 2023; doi: 10.3390/atmos14030421.
19. [19] Toz, A. C. Buber, M. Koseoglu, B. and Sakar, C. An estimation of shipping emissions to analysing air pollution density in the Izmir Bay. Air Quality, Atmosphere & Health, 14(1), 69–81, 2021; doi: 10.1007/s11869-020-00914-7.
20. [20] Gümüş, M. G. Investigation of Seasonal Relationships Between Air Pollution and Plant Health With Satellite Imagery: Turkey-Izmir Province Case (2020-2024). vol. 12, pp. 7–23, 2025.
21. [21] Elbir, T. (2002). Application of an ISCST3 model for predicting urban air pollution in the Izmir metropolitan area,” International Journal of Environment and Pollution, 18(5), 498-507.
22. [22] Çelikten, H. Evaluating machine learning (RF, XGBoost) and statistical model (MLR) for PM10 and air quality prediction: A case from Kars, Türkiye. Atmospheric Pollution Research, 17(6), 102975, 2026; doi: 10.1016/j.apr.2026.102975.
23. [23] Stafoggia, M., Johansson, C., Glantz, P., Renzi, M., Shtein, A., de Hoogh, K., ... & Bellander, T. A Random Forest Approach to Estimate Daily Particulate Matter, Nitrogen Dioxide, and Ozone at Fine Spatial Resolution in Sweden. Atmosphere, 11(3), 239, 2020; doi: 10.3390/atmos11030239.
24. [24] Liu, R. Ma, Z. Liu, Y. Shao, Y. Zhao, W. and Bi, J. Spatiotemporal distributions of surface ozone levels in China from 2005 to 2017: A machine learning approach. Environment International, vol. 142, p. 105823, 2020; doi: 10.1016/j.envint.2020.105823.
25. [25] Comrie, A. C. Comparing Neural Networks and Regression Models for Ozone Forecasting. Journal of the Air & Waste Management Association, 47(6), 653–663, 1997; doi: 10.1080/10473289.1997.10463925.
26. [26] Hashim, N. M. Noor, N. M. Ul-Saufie, A. Z. Sandu, A. V. Vizureanu, P., Deák, G., & Kheimi, M. Forecasting Daytime Ground-Level Ozone Concentration in Urbanized Areas of Malaysia Using Predictive Models. Sustainability, 14(13), 7936, 2022; doi: 10.3390/su14137936.
27. [27] Çelik, G. Sülün, Ş. Keçeci, S. and Ergenekon, P. Ozon Konsantrasyonunun Marmara Bölgesi İstasyonlarındaki Hava Kirliliği ve Meteorolojik Ölçümler Kullanılarak Makine Öğrenmesi Modelleriyle Tahmini, Sura Akademi. 9, 9–23, 2025.
28. [28] Ay, Ş. and Eki̇nci̇, E. Comparison of Machine Learning and Deep Learning Methods for Modeling Ozone Concentrations. jista, 5(2), 106–118, Sep. 2022; doi: 10.38016/jista.1054331.