I write this as Hurricane Irma bears down on South Florida, with the potential to be up there with the worst ever disasters from a hurricane in Florida. I also hear some people saying they have ridden out other hurricanes and so are planning on riding out Irma. This ridiculous notion deserves some consideration ...
In my tornado research, I spent decades becoming familiar with the climatological record of tornado occurrences in the USA. In the process, one can't help but observe that really big, bad tornadoes are but a small minority of the 1000 or so tornadoes that hit the USA every year. In 1998, a tornado rated F3 hit Gainesville, GA in the early morning, killing 12 people in an event that was unusual in that it was not warned-for in advance. In the wake of that event, my colleague Dr. Harold Brooks was talking via the phone to an emergency manager in the Gainesville area and she told him (I'm paraphrasing) that she had no idea things could get that bad in Gainesville! Clearly, she didn't know anything about the last single tornado in the USA to kill more than 200 people - the tornado that struck Gainesville, GA on 06 April 1936 (part of a two-day outbreak including a single tornado that killed 200+ in Tupelo, MS the day before). If a "generation" is roughly 30 years, this means that the institutional memory of that awful day in 1936 had been mostly lost in, even in an agency about being prepared, within roughly two generations! My experience says that's pretty typical. After a big disaster, awareness is high and people are receptive to the call for preparation. But as time passes, people move away, people die, new people move in and the local memory of disaster fades all too quickly. Resources for event preparation are re-allocated to other projects. Complacency grows. All too soon, the disaster is mostly forgotten. But the weather data base doesn't ever forget.
Studying the climatology of hazardous weather gives researchers a mental model of dangerous storms that isn't widely known in the "general public". While I was visiting Australia in 1989, it turned out there was a flash flood event in Melbourne while I was there. It wasn't a major event, being confined mostly to urban flooding. I watched a TV interview the next day with a couple living in the area hit by the flash flood, and they said "We've lived here for 9 years and we've never seen anything like this!" So they apparently believed that living in Melbourne for 9 years was going to representative of all the possible weather in Melbourne for all the rest of eternity! And this relatively modest event was a big deal for them!
It's understandable that non-meteorologists would fail to have an accurate understanding of the occurrence of rare events. I'm not sure how to go about fixing this shortfall in our communication of science, but here, today, with the landfall of Irma in South Florida likely in the next two days, the complacency associated with people's flawed understanding of what is "typical" for their area seems to be influential in the choices some people are making. Ignorance of such things almost never implies a blissful outcome.
Immediately after a major storm disaster, people are likely to want to think of what happened to them as a "freak" weather event: something unprecedented and very unlikely. Being hit by a major storm is a relatively rare occurrence, but calling it a "freak" event is misleading and counter-productive. If you're familiar with the climatology of tornadoes, someplace (and possibly someone) is going to be hit by a violent tornado virtually every year! Violent tornadoes are rare in any one place, but they aren't "freak" events, somehow outside the range of our human experience. Of course, they likely are outside of your own personal experience! You could live in central OK and not be hit by a violent tornado in 1000 years; on the other hand, Moore, OK has been hit by violent tornadoes in 1999, 2003, 2010, and 2013! The distribution of tornadoes has all the signs of being "random": being truly random doesn't mean the events are spread irregularly but more or less uniformly. Instead, random spatial distributions have both clusters and voids. If we had enough data (at least 1000 years worth), we might have a very clear picture of the climatology of violent tornadoes, at least in central OK. But we don't have that much reliable data on tornadoes before, say, 1953. The more rare the event, the more data are required for a meaningful analysis of the danger.
It's also likely that our knowledge of Cat-4/5 tropical cyclones is similarly flawed. A much longer period of record is needed for an accurate picture to emerge. The climatology of major events determines such numbers as the "return period" for these events. The more data one has, the greater the number of major events in the database and the "return period" calculation is less about iffy extrapolation and more about reliable information. The notion of "return period" is widely misunderstood by the public but that (obviously) is off-topic for this blog.
Folks, it's just not helpful for you to have faith that your personal experience with storms includes all that could possibly happen to you. When the storms become more intense, the less useful your experience is. This affects the decisions you make in advance of being hit by a particular storm, and your decisions will determine such issues as whether you live through it or not. Your choice will affect your family and perhaps even your friends. Don't trust your knowledge of the past to be helpful - listen to what the forecasters are saying and take it seriously! Lives hang in the balance!