During several experiments we have monitored the eruptions at Stromboli with infrared sensors. Infrared emission during the explosions were recorded in parallel to the infrasonic and seismic signals by the same multichannel acquisition system. We detected each strombolian explosion as a sudden spike in the thermal signal seen by an infrared radiometer pointed at the vent. This occurs as soon as the hot gas and entrained ejecta reaches the visible top of the conduit, which is at height h above the level of the free surface of the magma within the conduit. If the gas rises at a speed U then this occurs at a time that is later than the bursting of the gas slug by h/U. It will then be detected effectively instantaneously by the radiometer. On the other hand, the onset of the infrasonic pulse generated by the arrival of the gas slug at the free surface will not be observed until the pulse has had time to travel up the conduit and then through the air to the detector. We assumed that the infrared onset is representative for the explosion onset at the surface. Time delays between infrared and infrasound changes with the explosions in a range from -0.5 to 3.5 seconds. This indicates that infrasonic waves are not produced outside the vent when the gas-fragments mixture hits the atmosphere, but inside the volcanic conduit and indicates that infrasound inside the conduit propagates faster than gas jet, which lead us to conclude that gas jet is subsonic also inside the conduit. Independently from the model we use to explain the origin of infrasound, time delay with respect to the infrared emission, will depend i) on depth where gas-fragments mixture separates from the magma column and accelerates towards the surface and ii) on the gas jet velocity changing as the system moves from one degassing state to another. Time differences between seismic, infrasound and infrared onsets provide strong constraint for modeling the explosive dynamics of open-vent volcanoes.
