I get a bit tired of people assuming that the fact that they have a preference on a matter is justification enough to draw strong conclusions. I like this better, therefore it is better for everyone is not a valid conclustion. That's why I'm sharing my response to this article, rather than the article itself, which I find completely void of content. Arguably, I should have just ignored the article, but I wanted to take the opportunity and try, just try, to help the author not make the same mistake again.
I'm sorry, but this article lacked any sort of justification for its conclusions. The author seems to have predetermined that his preference is Python, therefore Python must be the winner. Both are high-level languages that have advantages and disadvantages. Fortran started life essentially as a DSL for math, much like Matlab is today. It's the FORmula TRANslator and is designed with the necessary rules built into the language to make it useful for that task and easy to compile into fast execution. Python is an interpreted language that brings along many modern conveniences and conventions, making it well-suited for the high-level plumbing of a scientific application, but requires calling into libraries like numpy to obtain high performance on numerical kernels because the rules of the language are less friendly to building high performance machine code. It's also very well-suited for building DSLs, but again, those are generally translated down to libraries that are written in lower-level, compiled languages (sometimes Fortran!).
Arguing that one language is "easier to learn" than another is a bogus argument. Easier to whom? I work with a lot of developers who have a mathematics or domain science background first and to them the FORmula TRANslator is generally much easier to learn. To someone leaving a CS program now, where they may have only been taught dynamic, interpreted languages such as python, maybe python is easier to learn. I have a CS background, but learned static typed, compiled languages first, so to me Fortran was dramatically simpler to learn than python. "Easy" is in the eye of the beholder, so one can't make blanket arguments like "A is easier to learn than B."
I think a far more useful discussion than "this language is better than that" is a discussion on how to choose a language that fits the needs of your project. Every programming language has tradeoffs. I do think that Python is a viable language for HPC and scientific computing because it's high level, interfaces well to other languages, makes writing DSLs fairly straightforward, and even takes some inspirations from Fortran. It's downsides are scalability and raw speed, so it's best suited for prototyping and plumbing. Fortran has native support for arrays, unlike C/C++, and many mathematical built-ins that make it very straightforward to use and obtain good performance on math-heavy operations. It's primary failings are that compilers often lack recent language features and it's taught less frequently now in favor of Matlab and sometimes python. Given those pieces of information I can look at the needs of my project and determine whether either language meets my requirements and which meets them better.