if you calculate it with two populations, they become a single population at that point, and if the environments are different, that changes the heritability too
No, populations are defined. And yes, it's hard to measure heritability.
But if you defined your populations as all people with blue eyes and all people with brown eyes you'd have a pretty easy time demonstrating eye color is heritable.
I think this is crossing heritability in the traditional biology sense (passed down via genes) with the way it's used in things like "heritability estimate" (within pop statistical value that says nothing about genetic causation).
But it's also sort of true when you're talking about interval variables and you don't have a distinct definition of the populations that indicates some relevant genetic difference.