Both Fortran and Julia have a concise syntax for array operations, and both, with a little care, can produce fast programs. Unless you want to program in C or C++, those are your only options¹ for high-level languages that have already proven themselves in high performance computing.
I wouldn’t advise rewriting a substantial Fortran project in Julia; but for new projects, Julia is the better choice. Fortran is excellent in the scientific/numerical high performance niche, but it’s still awkward to move beyond that; it’s still stuck in the Formula Translator role. So the non-numerical parts of the program, that deal with input/output, for example, are awkward; Julia is more pleasant to program these tasks.
Julia lets you build your program interactively, by trying things out in the REPL. This alone may be sufficient reason to prefer it over Fortran.
Although most scientists won’t become sophisticated as programmers, for those that will, Julia can grow with them. It has extensive support for metaprogramming (for example), including real macros. Its type/dispatch system, and other features, allow you to write far more concise and better organized code than you can in Fortran. Sharing code and composing libraries² is easier in Julia.
Really, Julia is just more fun to program in.
1 HPCWire (2017) ‘Julia Joins Petaflop Club’. Available from: https://cacm.acm.org/news/221003-julia-joins-petaflop-club/f...
2 Phillips, Lee (2020) ‘The Unreasonable Effectiveness of the Julia Programming Language’. Ars Technica. Available from: https://arstechnica.com/science/2020/10/the-unreasonable-eff...