I did my dissertation in astrophysics a few years ago using simulations I programmed in Fortran. As others have said, its efficient for number crunching, there's a lot of legacy code in scientific fields and most older Profs are familiar with it.

Also, as a language, there isn't much to it. If you're already familiar with the basics of programming, you can learn most of what there is to know about Fortran 77 in an afternoon, and pick up later additions to the language as you go. In physics, where the focus is getting grad-students to a point where they can do research quickly, having a language they don't have to spend a month figuring out is an advantage.