From December 2002 to July 2003, Stromboli volcano was characterized by a new effusive stage of eruption after a period of extraordinary strombolian activity. Signals recorded in two continuous monitoring stations during the eruption, which have already been presented in very recent papers, evidenced anomalies in the CO2 flux just before the onset of the eruption. A more detailed analysis carried out on the data subset acquired during the eruption, integrated by daily field observations of the scientific personnel working at the volcanological observatory in Stromboli, showed that CO2 flux and soil temperature are strictly related to volcanic events. Furthermore, the relative minima and maxima of the two parameters showed a strong correlation with wind speed and direction. This fact was especially true at the summit station, whereas at the coastal sites seasonal and meteorological effects masked the volcanic signal. The analysis of the wind data, particularly the relationships between wind speed and direction, air and soil temperature, and local circulation of atmospheric air masses revealed that during the eruption, in the summit area of Stromboli air movements were not only related to atmospheric circulation but were also significantly affected, and in certain cases caused, by volcanic activity. This conclusion was reached by observing several anomalies, such as the discrepancies in the wind direction between the two stations, higher air temperatures at the summit site, and inversion of direction for wind before and after the reopening of the conduit in a major explosion on 5 April 2003. The relationships found between volcanic activity, soil temperatures, CO2 fluxes, and wind speed and direction indicate that soil temperature measurements, in an open conduit volcano such as in this case, could be used to monitor the level of volcanic activity, along with CO2 flux. Furthermore, the possible volcanic origin of a peculiar type of air circulation identified in the summit area of Stromboli suggests that the separation between volcanic and atmospheric signals might not be obvious, requiring monitoring over a wide area, rather than a single location.