Comparison of Classical and Recent Predictive Models for Soil-Gas Diffusivity

Abstract

Accurate prediction of soil- gas diffusivity (Dp/Do: where Dp and Do are gas diffusion coefficients in soil and free air, respectively) and its variation with air-filled porosity (ε) is important for understanding soil aeration and subsurface greenhouse gas emissions and thereby to characterize essential soil functional services in terrestrial ecosystems. Since measuring Dp/Do is instrumentally challenging and requires maintaining controlled boundary conditions, different predictive models have been developed to estimate Dp/Do from easily measurable soil properties such as air-filled porosity (ε) and soil total porosity (Ф). In this study, a total of 593 gas diffusivity measurements conducted on 150 data from differently characterized undisturbed Danish soils were used to evaluate the performance of five prospective predictive models developed over the period of 1904 -2013. The selected soils represent agricultural soils, forest soils, urban soils, and landfill cover soils and measurements were within a selected range of matric potentials (−10 to −500 cm H2O) typically occurring in subsurface. Results of the model comparison made using two statistical indices (RMSE and Bias) showed that widely used model for repacked soils made a significant overprediction of undisturbed data. This study clearly distinguished the effect of soil structure status on soil gas diffusivity as demonstrated by the best performance of SWLR model over the other predictive models by yielding minimum RMSE and bias.