Fortran has enough critical mass in that field. I.e. good libraries and applications are written in Fortran, so enough people in the field learn Fortran and use Fortran instead of spending time rewriting stuff that works and is well-tested in C++.
It's also that there is no recognition or other incentives for developing or maintaining software in these fields. There was an article a week or so ago about a growing panic in particle physics over an important C program maintained for 40 years by a single prof who will soon retire.
