Sites included in the study extended 930 km north to south and 230 km east to west within the Great Plains states of Nebraska, South Dakota, and North Dakota (Fig. 1). Major Land Resource Areas (MLRA) representative of study sites encompassed approximately 30 Mha and included 53B and C (Central and Southern Dark Brown Glaciated Plains), 55A, B, and C (Northern, Central, and Southern Black Glaciated Plains), 65 (Nebraska Sand Hills), 75 (Central Loess Plains), 102C (Loess Uplands), and 106 (Nebraska and Kansas Loess Drift Hills) . Climate within the region is generally classified as semiarid to subhumid continental, with cold and dry winters, warm to hot summers, and erratic precipitation . Mean annual precipitation for the study sites ranges from 432 to 777 mm increasing from west to east, while mean annual temperature ranges from 4.7°C in the north to 10.6°C in the south. Soils at the sites possessed high inherent fertility, with Ustolls and Udolls as the dominant taxonomic suborders  (Table 1).
Initial soil conditions at the ten sites were not limiting to switchgrass establishment and growth [see supplementary online material]. Medium textured soils were prevalent across the sites, with the exception of the soil at Atkinson, which was classified as sand. Values for soil bulk density across sites were below critical threshold values for restriction of root growth . Soil pH across sites varied from strongly acid to moderately alkaline  (data not shown), and fell within a range for successful switchgrass germination .
Sites seeded to switchgrass were fields on working farms previously used for annual crop production. Field characteristics were such that they would have qualified for enrollment in the Conservation Reserve Program (CRP) . Field size across sites averaged 6.7 ha (Range = 3.0 to 9.5 ha). Sites at Douglas, Lawrence, and Crofton were seeded to switchgrass in 2000, while all other sites were seeded in 2001. Cultivars used for seeding included Cave-in-Rock (Douglas, Lawrence), Trailblazer (Douglas, Lawrence, Crofton, Atkinson, Huron, Highmore, Bristol), Shawnee (Lawrence, Crofton, Ethan), and Sunburst (Streeter, Munich). With the exception of Sunburst, all cultivars originated south of 43°N latitude . Switchgrass was seeded at a rate of 322 pure live seeds (PLS) m−2, or approximately 10 kg ha−1 using minimum or no-till management practices. Application of N varied in amount and type across sites based on biomass yield expectations and soil moisture conditions. Over the 5 year period of the switchgrass stands, site averages of applied N ranged from 31 to 104 kg N ha−1 year−1 (Mean = 74 kg N ha−1). With the exception of the establishment year, aboveground biomass was harvested annually and baled. Most cooperating farmers harvested at emerged inflorescence to post-anthesis (early to mid-August) in post-establishment years. In contrast, farmers at Bristol, SD and Munich, ND harvested after a killing frost . Biomass samples from switchgrass bales were used to determine dry matter and C concentration. Harvested C was determined by multiplying the biomass yield by the biomass C concentration. To verify machine harvested yields, aboveground biomass was hand-clipped, dried, and weighed from 1.1 m2 quadrants at 16 locations within each field . Additional details on site establishment, management, and biomass harvest and analysis are outlined elsewhere [31, 32].
To evaluate change in SOC under switchgrass over time, soil samples were collected from each site on a 5 year time-step. Sites in Nebraska were sampled in 2000 and 2005, while all other sites were sampled in 2001 and 2006. Samples were collected in the spring once soils were no longer frozen and surface conditions were dry enough to permit vehicular traffic. In 2000 and 2001, samples were collected immediately prior to switchgrass planting.
At each site, soil samples were collected from two transects with three sampling locations each (located approximately 30 m apart), resulting in a total of six sampling locations per site. Coordinates for each sampling location were recorded using a handheld GPS devise with an accuracy of <3 m (Garmin International, Inc., Olathe, KS). Sampling locations were treated as pseudoreplicates as reviewed by Gomez . Due to difficulty during initial sample collection at Atkinson, only two locations per transect were sampled.
Soil samples were collected using a truck-mounted Giddings hydraulic probe (Giddings Machine Company, Windsor, CO, USA) with an inner tip diameter of 4.2 or 4.4 cm, depending on soil conditions at the time of sampling. Soil depths sampled were 0–5, 5–10, 10–20, 20–30, 30–60, 60–90, and 90–120 cm for three Nebraska sites (Crofton, Douglas, Lawrence) and 0–5, 5–10, 10–20, and 20–30 cm for the remaining sites. To ensure adequate sample mass for laboratory analyses, seven soil cores were composited at each sampling location (approximately 1 m2) for the 0–5 and 5–10 cm depths, five soil cores for the 10–20 and 20–30 cm depths, and two soil cores for the 30–60, 60–90, and 90–120 cm depths. Following collection, each sample was saved in a double-lined plastic bag, stored in coolers while in transit to the laboratory, and then placed in cold storage at 5°C until processing.
Prior to analyses, whole soil samples were dried at 35°C for 3 to 4 days and then ground by hand to pass a 2.0 mm sieve. Identifiable plant material (>2.0 mm in diameter, >10 mm in length) was removed during sieving. Total soil C was determined by dry combustion on soil ground to pass a 0.106 mm sieve using a Carlo Erba NA 1500 CN analyzer (Thermo Scientific, Waltham, MA, USA). Using the same fine-ground soil, inorganic C was measured on soils with a pH ≥ 7.2 by quantifying the amount of CO2 produced using a volumetric calcimeter after application of dilute HCl stabilized with FeCl2 . Soil organic C was calculated as the difference between total C and inorganic C. Gravimetric data were converted to a volumetric basis for each sampling depth using field measured soil bulk density, which was determined using the oven-dry weight and known volume of the composited samples . All data were expressed on an oven-dry basis.
Biomass samples were ground to pass a 1 mm screen prior to C determination by near infrared spectrophotometry (NIRS) . A switchgrass NIRS prediction equation was based on total C analyses of a set of 108 switchgrass samples using combustion analysis . The NIRS standard error of calibration and prediction for biomass C were 2.13 and 3.95 g kg−1, respectively.
Changes in SOC between sampling times were calculated by subtracting initial values from values after 5 years within a sampling location. Calculated changes were then evaluated within a site by depth using a paired t-test in PROC MIXED . Changes in SOC over time for cumulative sampling depths (0–30 and 0–120 cm) within and across sites were evaluated similarly.