The primary constraint seems to be infrastructure to evacuate and shelter people. Reliable forecast of storm surge location and magnitude would save lives by allocating those resources to the correct places. Doing this well involves improved simulation methods, but also designing observational strategies, such as where to make more accurate measurements of bathymetry (one of the leading sources of uncertainty in storm surge and tsunami prediction). Disclaimer: I work in a related field and have colleagues that develop methods for storm surge forecast.
The three biggest needs to prevent a large death-toll from something like this:
1. Education regarding effects of disasters (e.g. what a storm surge is, what areas are likely to be affected by landslides, etc.)
2. Basic first aid medication (having seen my fair share of misadventure in the Philippines, I can confidently say that a subsidized boy scout-style training program would work wonders there)
3. Instructions for putting together 72-hour crash bags, which should be kept in convenient locations (this is a good idea generally -- no matter where one lives a calamity could occur, and having copies of important documents/candles/matches/clean water can make a huge difference).