Thanks everyone for their feedback. Nothing like getting slash-dotted for finding out what the world thinks.

Now that its off the front page, I'm going to to take the opportunity to get the last word.

Few more points:
1) In the initial description it does say "... this application is a general guide and not a true prediction" and so yes I agree flood fill isn't the most accurate algorithm for predicting tsunamis. More accurate is GeoClaw which has a lot of fortran and phython code and some very detailed wave models - but obviously its not easy to incorporate into a simple Applet like mine.

2) IMHO, flood fill is better than simply testing the elevation of all points against the wave cause it prevents the water moving through mountain ranges. The logic for flood fill is very simple (maybe 100 lines of code) and its a small extension to basic elevation testing we can handle.

3) Keep watching the web site, cause I do have some improvements planned. Namely handling more than a 10km by 10km area and taking what we do map down to lower level of detail. But for the algorithm I'll probably be sticking with flood fill at least in the short term.

Finally, if someone is bored I am open to collaboration - in particular to with regard to point 1 above, I'm sure someone with the right knowledge could make a much better algorithm. Check the site for email address.

Thanks for the feedback. A couple of points:
1) Yes - you do need to set the direction before clicking on the map. Otherwise the wave will go in the wrong direction and the results will be bad.
Next release (give me 24 hours) will force the user to enter a direction and also put a little arrow showing the direction chosen.

2) Yes - the flooded area can only be 10km by 10km. I chose this based on some basic research on tsunami reach and needing a reasonable response time from google's elevation service.

3) Yes - it only uses flood fill algorithm to determine if an error is blue or not. If the elevation of the location is less than the wave and it has a neighboring sector which is blue, then its blue.
If someone can point me to the maths that describes how to calculate this based on the wave's energy, velocity etc then I'll have a try. But there may be a limit to what can be done in 300 lines of JavaScript :-)