I find to Fortran90 to be highly useful. I use it essentially for all my modelling, and I use Matlab pretty much just for data analysis. As Transport pointed out, there are decades worth of legacy Fortran code out there still in use. Further, with MPI and HPF, Fortran90 is very useful for intensive computational work. Both the solar evolution model I use from LANL, and my own solar atmosphere work is Fortran based.
Here's the thing I guess: If you plan on doing major system modelling in a computer, using the computer system as a sort of lab to calculate results from theories that can then be compared to observational data, learn Fortran. If all you want is to data analysis (as opposed to data calculation) Matlab will be better.
See: http://www.lahey.com/PRENTICE.HTM"
The letter is old, but his points are still valid.