[01] Ningbo Jiang, New South Wales Office of Environment and Heritage, Sydney, Australia. [02] Matthew L. Riley, New South Wales Office of Environment and Heritage, Sydney, Australia.

This paper explores the utility of an ensemble decision-tree method called random forest, in comparison with the classic classification and regression trees (CART) algorithm, for forecasting ground-level ozone pollution in the Sydney metropolitan region. Statistical forecasting models are developed to provide daily ozone forecasts in November-March for three subregions, i.e., Sydney east, Sydney south-west and Sydney north-west. The random forest models are evaluated in reference to the single decision-tree models developed from the classic CART algorithm. The results show that the random forest models outperform the CART models for forecasting high ozone pollution in Sydney south-west and Sydney north-west, the areas where the highest ozone pollution are observed. The random forest models also show a lift in forecasting skills in Sydney south-west if compared to the existing forecasting practice for the basin as a whole. These results suggest that random forest is a promising method for air quality forecasting in Sydney. This study promotes the application of a statistical ensemble approach to air quality forecasting.

Air Quality Forecast, Ozone Pollution, Decision Tree, Random Forest, Bagging, Boosting

[01] ABS. 2011. 2011 Census QuickStats. Australian Bureau of Statistics. Access to www.censusdata.abs.gov.au/census_services/getproduct/census/2011/quickstat/1GSYD?opendocument&navpos=220 on 1^st May, 2014. [02] Anderson JO, Thundiyil JG, Stolbach A. 2012. Clearing the Air: a Review of the Effects of Particulate Matter Air Pollution on Human Health. Journal of Medical Toxicology 8: 166-175. [03] Barnett AG. 2012. Air pollution trends in four Australian cities 1996-2011. Air Quality and Climate Change 46(4): 28-33. [04] Barnett AG, Williams GM, Schwartz J, Best TL, Neller AH, Petroeschevski AL, Simpson RW. 2006. The effects of air pollution on hospitalizations for cardiovascular disease in elderly people in Australia and New Zealand cities. Environmental Health Perspectives 114(7): 1018-1023. [05] Breiman L. 1996. Bagging predictors. Machine Learning 24 (2): 123–140. [06] Breiman L. 2001. Random forests. Machine Learning 45(1): 5–32. [07] Breiman L, Cutler A, Liaw A, Wiener M. 2013. Package “randomForest” version 4.6-7. Access to The Comprehensive R Archive Network http://cran.r-project.org/ on 23^rd June, 2014. [08] Breiman L, Friedman J, Olshen RA, Stone CJ. 1984. Classification and Regression Trees. 1^st Edition. Chapman & Hall: USA. [09] EPA. 2012. New South Wales State of Environment 2012. NSW Environment Protection Authority: Sydney, Australia. [10] EPHC. 2010. Expansion of the Multi-city Mortality and Morbidity Study, final report by University of the Sunshine Coast, University of Queensland, Department of Environmental Protection Western Australia, Environment ACT, Environment Protection Authority Victoria, NSW Health, New Zealand Ministry for the Environment, Queensland Health to the Environment Protection and Heritage Council, Canberra. Environment Protection and Heritage Council (EPHC) [www.scew.gov.au/archive/air/#multi]. [11] Han J, Kamber M. 2006. Data Mining: Concepts and Techniques, 2nd ed. Morgan Kaufmann Publishers: San Francisco, CA. [12] Hart M, De Dear R, Hyde R. 2006. A synoptic climatology of tropospheric ozone episodes in Sydney, Australia. International Journal of Climatology 26: 1635-1649. [13] Hastie T, Tibshirani R, Friedman J. 2009. The elements of statistical learning: data mining, inference, and prediction. Springer: USA. [14] Honore C, Ung A, Corbet L, Malherbe L. 2008. CITEAIR II Common Information to European Air: Good Practice Guide to Urban Air Quality Forecast. European Union (EU) European Regional Development Fund Regional Initiative Project. [15] Jiang N, Betts A, Quigley S, 2013. Visualisation of climate and air quality data using self-organising maps. In proceedings of 21^st International Clean Air and Environment Conference (CASANZ), 7-11 September 2013, Sydney. [16] Jiang N, Luo K, Beggs PJ, Cheung K, and Scorgie Y. 2014b. Insights into the implementation of synoptic weather-type classification using self-organizing maps: an Australian case study. International Journal of Climatolology doi: 10.1002/joc.4221 [17] Jiang N, Riley M, Scorgie Y, Betts B, Kirkwood J, Duc H, Trieu T, Salter D, Ji F, Chang L. 2015. Enhancing air quality forecast in New South Wales. In proceedings of 22^nd International Clean Air and Environment Conference (CASANZ), 20-23 September 2015: Melbourne. [18] Jiang N, Salter D, Dutt U, Alan B. 2014a. 2008-2013 air quality forecast performance evaluation. CAS Technical paper CAR-Tech-001. New South Wales Office of Environment and Heritage: Australia. [19] Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D. 1996. The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society 77: 437–471. DOI: 10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2. [20] Katestone Scientific. 1997. Anthropogenic Influences on Australian Urban Airsheds. A report to the Australian Academy if Technological Science and Engineering: Brisbane, Australia. [21] Leighton RM, Spark E. 1997. Relationship between synoptic climatology and pollution events in Sydney. International Journal of Biometeorology 41: 76–89. [22] Liaw A, Wiener M, 2002. Classification and Regression by randomForest. R News 2/3: 18-20. [23] Liaw A, Wiener M, 2013. Package ‘randomForest’ Version 4.6-7. Access to The Comprehensive R Archive Network http://cran.r-project.org/ on 1^st July, 2014. [24] NEPC. 2003. National Environment Protection (Ambient Air Quality) Measure – As varied 7 July 2003. Environment Protection & Heritage Council, Level 5, 81 Flinders Street, Adelaide, SA 5000, Australia. [25] OEH. 2014. Air Quality Trends in Sydney. New South Wales Office of Environment and Heritage: Sydney, Australia. [26] Shapire R, Freund Y, Bartlett P, Lee W. 1998. Boosting the margin: A new explanation for the effectiveness of voting methods. Annals of Statistics 26:1651–1686. [27] Therneau T, Atkinson B, Ripley B. 2014. Rackage “rpart” Version 4.1-8. Access to The Comprehensive R Archive Network http://cran.r-project.org/ on 23^rd June, 2014. [28] US EPA. 2003. Guidelines for developing an air quality (ozone and PM2.5) forecasting system. EPA-456/R-03-002. U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Part, North Carolina. [29] US EPA. 2015. How effective are air quality alerts in reducing adverse effects in the real world. U.S. Environmental Protection Agency website: http://www.epa.gov/o3healthtraining/aqi.html#alerts, accessed on 8 May 2015. [30] White J. 2011. EPA AIRNow Program. Air Quality Index Reporting and Forecasts. Presentation at the 1st Workshop on Satellite Observations for Air Quality Management. U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Part, North Carolina. [31] WHO. 2006. WHO Air Quality Guidelines for Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide. Global Update 2005. Summary of Risk Assessment. World Health Organization, Geneva. [32] WHO. 2014. Ambient (outdoor) air quality and health. World Health Organisation http://www.who.int/mediacentre/factsheets/fs313/en/ accessed on 17 April, 2015. [33] Zhang Y, Bocquet M, Mallet V, Seigneur C, Baklanov A. 2012a. Real-time air quality forecasting, part I: History, techniques, and current status. Atmospheric Environment 60: 632-655. [34] Zhang Y, Bocquet M, Mallet V, Seigneur C, Baklanov A. 2012b. Real-time air quality forecasting, part II: State of the science, current research needs, and future prospects. Atmospheric Environment 60: 656-676.