Pasture vs. range grass in SWAT

[Justin Goldstein]

Colleagues,
   I am having difficulty deciding whether to assign the grasslands in my watershed the "pasture" cateogry or the "range-grasses" category.  Prevailing grasses are buffalo grass, big and little bluestem, blue grama, and sideoats grama with some johnsongrass.  There is also bermudagrass.  I checked the SWAT input-output designation and saw that pasture is simulated as bermudagrass while range grasses are simulated as little bluestem with an LAI of 2.5.  The watershed's fairly large size (>2500 km2) complicates quantifying how much of the grass actually is used by cows and horses, even on just a subbasin level, although I saw some amount when driving around.  I asimply can't ascertain if much of the grassland is part of larger ranches having cattle, although my gut instinct tells me it is.  The  NASS Crop Dataset Layer considers most of this to be "range grasses" but I'm sure its definition differs for this differs from that of SWAT.  I therefore simply assigned all of the grasslands the "pasture" category; does this make sense?  I ran the simulations with both and while assigning all the grasslands "range" increases the nash-sutcliffe post-calibration, assigning all the grasslands "pasture" greatly improves the overall percent bias.  My watershed is situated between latitudes 35 and 37 north so I don't have to assign it "fescue" (SWAT input-appendix A).  Either way, the calibration is really good (N-S 0.8, PBIAS 5-10%), but the flow results do differ due to varying curve numbers, etc., so I would like to have it as accurate as possible regardless of the fit of the calibration.  Which of the two land use classes would you recommend?

   Thanks,
  Justin