Steller, R.M., N.A. Jelinski, and C.J. Kucharik.  Developing models to predict soil bulk density in southern Wisconsin using soil chemical properties.  Invited submission to the Electronic Journal of Integrative Biosciences.  Jan. 2008.

Abstract
There is an emerging need to estimate and verify soil carbon credits attributed to conservation tillage and prairie restoration in the Midwestern U.S. for the developing global carbon market.  However, current soil sampling strategies may need to be augmented by empirical modeling to minimize costs while covering larger regions.  We constructed models relating soil bulk density (BD) to soil organic carbon (%SOC) and total nitrogen (%TN) concentrations using 146 sites in southern Wisconsin under varied land-use to determine whether empirical models could reliably predict BD – in an effort to support estimates of SOC sequestration for future carbon crediting programs.  As expected, a significant exponential relationship was found between %SOC and BD (R2 = 0.90; P < 0.0001) across all sites.  We then constructed exponential models after categorizing data into undisturbed ecosystems, restored prairie, and croplands, and found that although expected errors due to modeling BD based on %SOC were similarly low among all land-use groupings (~ 0.1 g cm-3, or 7-17%), the correlation between observed and predicted BD values, along with model parameters, were quite different. Predicted values were closest to observed values for undisturbed sites (R2 = 0.90), weaker with restored prairie (R2 = 0.49), and weakest with cropland (R2 = 0.28).  This suggests that highly intensified crop management practices influence BD in a way that might make using %SOC or %TN as single predictor variables unreliable.   We suggest that models relating BD and soil chemical properties should consider the varied effects of land-use management over many different soil types, particularly for the determination of carbon credits on agricultural land in temperate climate regions.