Great question – this really is the area where communication is all important! I have only been in this position once, so far – but each time in the future I would do the same. I was working on some data from Montserrat (West Indies), in 1997. The volcano went into a dramatic eruption phase where it started to have large explosions every 10 – 15 hours or so. I used the timing of these events to make a best guess, using statistics, of how long we would have to wait after the last explosion, in order to be sure that the explosions had finished: a bit like how long do you wait at a bus stop until you give up! My best guess was about 30 hours. I sent this forecast over to the scientists in the observatory and, sure enough, two days later the explosions finished – and my forecast had worked. I have not yet had any success with warnings before eruptions – but again, my first point of contact would be through the observatory scientists – they would in turn be the people to notify the local politicians, disaster managers and so on. The last thing you want in a crisis is public disagreements between scientists!