Abstract. 2003 was a record year for wildfires worldwide. Severe forest fires killed four people, displaced 20 500 others and burnt 260 000 ha in South-East Australia in January 2003. The uncontrolled fires ignited in early January 2003 as a result of a prolonged El Niño drought in South-East Australia. Severe weather conditions resulted in a fast spread of the fires and poor air quality in a region where 70% of the population of Australia lives. We use state-of-art global chemistry and transport model GEOS-Chem in conjunction with ground- and space-based observations to study the ozone (O₃) and aerosol enhancement due to fires. Firstly, the monthly mean surface O₃ and Aerosol Optical Depth (AOD) in January 2003 are compared to January 2004 and, secondly, from sensitivity model simulations, four episodes are isolated and an attempt is made to quantify the contribution of the fires to air quality in south and South-East Australia.

In January 2003 the observed monthly mean afternoon surface O₃ in Victoria (VIC) and South Australia (SA) reached 27.5 ppb, which is 6.5 ppb (i.e. 30%) higher than in 2004. The simulated O₃ is 29.5 ppb, which is 10 ppb higher than in 2004. While the model tends to overestimate the observed peak O₃, it exhibits very good skill in reproducing the O₃ temporal variability in January 2003 with a correlation of 0.83. In VIC, the air quality 4-h ozone (O₃) standard exceedences are reported on 17, 24 and 25 January. On 12, 17, 24–25 and 29 January 2003, the observed O₃ peaks above 40 ppb and the simulated fire contribution is higher than 10 ppb. During these 4 episodes, the range of observed O₃ enhancement due to fires is 20–35 ppb, which is a factor of 3 to 5 higher than the monthly mean. The simulated fire O₃ enhancement is in the range 15–50 ppb with a factor of 1.5 to 5 higher than the monthly mean. During two episodes, a well-formed surface wind channel stretches across the Tasman Sea facilitating the long range transport to New Zealand contributing to a 10% increase of surface O₃.

During the four episodes in January 2003, the observed AOD was up to a factor of five higher that the monthly mean AOD. The simulated and observed AODs agree on the spatial structure. Despite the model tendency to underestimate the AOD, it proves a useful tool in reconstructing the mostly patchy observations.