Information on temporal and spatial variation in switchgrass (Panicum virgatum L.) biomass composition as it affects ethanol recovery (L Mg^-1) at a biorefinery and ethanol production (L ha^-1) at the field scale has previously not been available.  Switchgrass biomass samples were collected from a regional, on-farm trial and biomass composition and ethanol recovery was quantified using newly developed near-infrared reflectance spectroscopy (NIRS) prediction equations.  Total hexose (cell wall carbohydrates and soluble sugars) concentration means ranged from 342 g kg^-1 to 398 g kg^-1 while pentose (arabinose and xylose) concentration means ranged from 216 g kg^-1 to 245 g kg^-1 across locations. Theoretical ethanol recovery rates varied significantly by year and location, with 5 yr means ranging from 381 L Mg^-1 to 430 L Mg^-1.   Total theoretical ethanol production ranged from 1749 L ha^-1 to 3691 L ha^-1 across locations.  Variability (coefficient of variation) ranged from 1% to 4% for theoretical ethanol yield recovery (L Mg^-1) and 14% to 38% for theoretical ethanol yield production (L ha^-1) within fields for established switchgrass.   Spatial persistence across harvest years for theoretical ethanol recovery rates or total theoretical ethanol production values within fields was not found to be highly correlated. Switchgrass biomass composition from farmer fields can be expected to have significant yearly and field to field variation in a production region and this variation will significantly affect ethanol or other liquid fuel yields per ton or hectare.  Biorefineries will need to assay biomass for quality to optimize conversion efficiency.  Research to develop improved biomass cultivars and management practices that stabilize and increase biofuel recovery and production will be essential for an emerging cellulosic biofuel industry.