Dynamics markings in organ music can be a little perplexing. When a pianist's score is marked crescendo, there's only one way to increase the volume, and that's to press the keys harder. But how exactly should crescendo be interpreted in an organ score? Does it might mean open the swell box? Does it might mean add more stops? Or does it mean a combination of both?
The only answer which can be given to this question is: it depends. Mostly it depends on the composer, of course. It may take a certain amount of background knowledge about the composer and their musical world to make an informed decision. In the absence of such knowledge, the only advice which can be given to the beginner is to listen to an expert playing the piece and copy what they do! One factor which complicates the issue is that organ composers differ greatly in the extent to which they specify registration for their music. Bach's is (for the most part) just notes on staves, without any additional instructions whatsoever. Felix Mendelssohn declared himself reluctant to give registration instructions because he had found, through playing various organs on his travels, that the effects of same-named stops differed greatly.
Cesar Franck, on the other hand, gives very detailed instructions regarding registration, perhaps because he travelled less than Mendelssohn, and within his immediate environment the success of the organ builder Cavaille-Coll had imposed a certain standardisation of what organ stops were supposed to sound like. This doesn't necessarily mean that his instructions should be followed slavishly, of course, but we can be fairly sure that when he expects a change of registration he will ask for one explicitly. Conversely, dynamics markings without an explicit change of registration should be accomplished using the swell pedal alone. There's a good example towards the end of the introduction to Grand Piece Symphonique where four bars are marked molto crescendo followed by two bars marked f followed by a further four bars marked dim ending with pp. One could be forgiven for thinking that such a dramatic effect could only be achieved by adding and subtracting stops, but that was almost certainly not the composer's intention. The f should be interpreted as meaning as loud as possible with the stops currently drawn, and the pp as quiet as possible with the stops currently drawn.
It was at this point that I was planning to illustrate my point with a Youtube clip, such as this one of Doug Marshall at the Organpower! 2004 event:
The only problem is that Doug does add stops (at 3:06) and he does subtract them (at 3:13). He rides roughshod over the composer's intentions at a number of other places too! No doubt he would claim it as his right as interpreter to make these changes. I can't deny that I greatly enjoy listening to his spirited performance, but I would still question whether the above registration changes really do represent an improvement. They could be said to disrupt the smooth flow of the musical idea, and also to steal the thunder of the really big crescendo which follows just afterwards.