It's unfortunate, but I can see a system like this attracting the attention of some pimply-faced sixteen-year old anxious to demonstrate his hacking ability. A touch of sociopathology and the feeling of invulnerability that comes with imagining yourself as an uber-hacker would make the prospect of sending millions of people fleeing from their homes irresistable.
A couple of incidents like that and the system's credibility is compromised.
Agreed, security would be both important and difficult considering the numerous sources of data injection. Particularly considering the "default to alert" philosophy. Did I miss an article here, or do you plan to mimic the pressure sensors on 'sea floor -> bouy -> satellite' concept? What happens if your data sources go offline? Verification processes? Just wondering if this is hashed out somewhere?
Guys -- I share your security concerns (in fact I wrote something to Bob about script kiddies that's isomorphic with Allan's comment.)
I'll have a longer post on my basic architectural thoughts here shortly; the short answer is that I don't want this to necessarily depend on any single source of data, but that I also dont see this as having its own sensor web; instead, it should use other sources.
My thought is that a "conservative" alert system would be capable of being very simple, because the alerts could be based on seismic data alone. I am, however, by no means a geophysicist; there seems to be some question in others' minds, and they're better informed than I am.