I think most of what keeps FORTRAN going is inertia. Tons of legacy code out there in the form of libraries and applications which have been written over the decades. FORTRAN is to scientific computing what COBOL is to the banking and insurance industries.
The last time I had any sort of dealings with FORTRAN was probably about 5 years ago on these forums, when someone was asking for help getting a FORTRAN program to compile. Before that, I had some minor contact with it at Fermilab back in the '90s (most of the stuff I worked on there was C/C++ or assembly though); and when I worked in the financial sector back in the late '80s we had some financial modeling code that was coded in FORTRAN (but I eventually ended up porting all of the legacy FORTRAN code to C since that's what we were calling it from anyway).
Nostalgia isn't what it used to be.