1. R is slow and clumsy. 2. Matlab skills are mandatory in virtually any engineering field. 3. Matlab is used for far more than "statistics" and "big data". It is particularly useful for dynamic systems modeling and controls design. For example, if I want to draw the root locus of a system with a known open loop transfer function, I can do it in one line in Matlab. I can't do it in R.

Matlab is hardly a "little known" language. It is universally used by engineers of all stripes. Pretty much any EE, ME, or AE graduate will have at least a passing familiarity with Matlab. I've taken quite a few engineering courses where Matlab knowledge was more or less a prerequisite. Knowledge of Matlab and Simulink is mandatory in the controls and structural dynamics fields.

It sounds like these seismologists are relying almost entirely on template matching, which is nothing more than a pattern recognition algorithm. These kinds of algorithms are no substitute for intelligent analysis, especially in the absence of reliable statistics. They can be used and abused like any other engineering tool if they aren't properly understood.

I'm not familiar with Stewart's book, but judging from the first chapter on Amazon it looks really good. Lots of examples and thorough explanations without dumbing down the material. He even presents the epsilon-delta definition of a limit, something which my watered down undergraduate classes never touched. Several people have mentioned Dover books. They are best used as supplementary material, or for solidifying your knowledge after you know a subject fairly well. You'll learn far more about electrodynamics from Griffiths ($141 on Amazon) than from any 50 year old Dover text.