> I'm not sure it is possible to convince someone
It is possible, but first you have to open a real dialog. That isn't going to happen without first learning to speak their language, because this is "More than just Resistance to Science"[1].
> Only new facts will do the trick
Arguments based on facts are useful when trying to persuade people fluent in the language of science and logic. Unfortunately I suspect only a minority of the population has sufficient understanding of data-focused language. Most people instead see language as phatic expression[2][3] first.
A good example of this from the other side of the language divide that still retains mutual intelligibility are Trump's repeated - and often self-contradictory - promises that he would "build the wall". People that primarily use language to convey facts often claims his rhetoric sounds ignorant, useless, or xenophobic. Trump's intended audience, on the other hand, hears the language of emotion that promises protection. Trump isn't trying to convey facts, so self-contradiction and other ignorant claims are not important.
To reach people on the other side of this divide, it is probably necessary to first learn their language of social hierarchy, emotion, etc. Skipping that step and continuing to throw facts at them is about as useful as sending them a textbook written in ancient Egyptian hieroglyphs.
[1] http://scienceblogs.com/clock/2007/05/31/more-than-just-resi...
[2] https://en.wikipedia.org/wiki/Phatic_expression
[3] While I'm discussing this as a language problem, some people may prefer George Lakoff's explanation[4] about "framing", which is a different interpretation of essentially the same ideas.
It's not about information.
> Like, a tornado destroys their house or such.
That's more like it. In other words, until something is personal people have a very hard time relating to it, no matter how irrefutably true it is.