Here you have some references: http://swat.tamu.edu/media/70993/azdez.asp.pdf and http://swat.tamu.edu/media/40029/lu.pdf

Above links are very useful. You need to use inputs to compute SWAT and compare with Observed data (Output of SWAT). You have to play with your inputs until achieve similar pattern of observed pattern. 

SWAT_CUP very useful for SWAT calibration.

http://www.eawag.ch/forschung/siam/software/swat/index

It's important to note, however, that SWAT does not take into account withdrawals/additions via groundwater (i.e. if you have a large groundwater to to surface water or surface water to groundwater flux).

This is based on a recent conference presentation given by Chris Breed from Baylor University (presentation here: http://proceedings.esri.com/library/userconf/proc14/papers/610_704.pdf)