We examine the facts behind the growth rate of atmospheric CO2 in 2002 and 2003. Observations show consecutive increases of greater than 2 parts per million (ppmv) per year for the first time on the Mauna Loa record (which extends back to 1958). We use a statistical regression to show that increasing anthropogenic emissions and El Ni¿o activity are able to account for variations in the CO2 growth rate at Mauna Loa, aside from the anomalously low growth rates of 1992 and 1993 following the Pinatubo volcanic eruption, and the anomalously high growth rate of 2003. Increased forest fires in the northern hemisphere, consistent with remote-sensing and carbon monoxide measurements, seem likely to have contributed significantly to the 2003 anomaly. We hypothesise that the hot and dry Eurasian summer of 2003 led to an increase in forest fire emissions from Siberia, and may also have directly suppressed land-carbon uptake. Insofar as the 2003 Eurasian summer may have been a symptom of anthropogenic climate change, it is possible that the 2003 CO2 growth rate anomaly is the first evidence of a positive climate-carbon cycle feedback.