Think of a watershed as having a number of compartments where water can reside: the soil, the shallow aquifer, ponds, wetlands, lakes. SWAT tracks the volume of water in each of these compartments in its mass balance calculations. While the user can set the initial volume of some of these compartments, some (soil moisture, perhaps) are not set, and they start the model run dry. It can take several years of a model run to fill these compartments up to a point where their input and output are approximately the same. During these first few years, the water yield from the model is generally less than expected, because the model is filling up these compartments to their "equilibrium" volume rather than passing water through the system.
While most of us think of a warm-up period as being needed for system hydrology, it should also be realized that the same concept could be applied to nutrients, as well. Are soils absorbing, or releasing, too many nutrients over time, i.e., are they building up or being depleted? Is that what is really happening in the watershed, or are the soils at approximate steady state with respect to nutrient content?
-- Jim