Estimating Particle Density from Soil Inventory Data in the Lake Erie Lowlands

The objective of this study was to develop a statistically based function for the estimation of particle density of texturally diverse soils sampled from a reasonably large geographic area (∼12,000 km 2 ) to enhance further development of pedotransfer functions with regional application. Available soil physical property data were assembled for soil series mapped during five municipal‐level soil inventory upgrades in southwestern Ontario, Canada. A total of 282 soil horizons from 91 soil profiles were identified that had the requisite measured data for particle density (water pycnometer method), soil organic carbon content (wet oxidation method), and particle‐size distribution (pipette method). Both linear and nonlinear regression procedures were used to relate particle density to soil organic matter content. Plausible estimates of particle density for the mineral component (2.65 Mg m −3 ), and particularly for the humic component (1.23 Mg m −3 ), were obtained ( r 2 = 0.208, RMSE = 0.11 Mg m −3 , P < 0.0001) even though the calibration data set had a limited range of soil organic matter content (<12%). The particle density of different mineral particle‐size fractions (e.g., clay) could also be distinguished statistically. The predictive capability of regression equations originating from soil inventory data sets encompassing large geographic areas are likely to be influenced by the soil taxonomic range sampled and the general data quality.