Investigating the role of soil structure on hydraulic properties using machine learning models and study of soils from long-term conservation agriculture

We investigate the impact of soil structure on soil hydraulic properties of saturated hydraulic conductivity and water retention.

We built a robust machine learning based pedotransfer function to predict saturated hydraulic conductivity and simulate the evolution of saturated hydraulic conductivity as a result of alterations of soil structural variables (bulk density and organic carbon concentration).

We complement and extend these model predictions, as well as evaluate the benefits of conservation agriculture systems by direct measurement of soil structural and hydraulic properties. We analyzed soils with different soil structural variables that resulted from management. We collected soils from a long-term conservation agriculture study farm that has plots subjected to conservation agriculture (no-till and winter cover cropping) and conventional agriculture for two decades (located at the UC West Side Research and Extension Center in Five Points, CA). The soil type at the study site is Panoche clay loam which is representative of many parts of the Central Valley of California. We conduct laboratory characterization of water retention characteristics and hydraulic conductivity using the HYPROPand KSATinstruments (METER Group AG, Munich, Germany) along with other standard measures of soil physical properties.

Our results suggest that the effect of long-term conservation agriculture on soil structure is substantial in terms of its effect on hydraulic properties. The role of soil structural variables on its hydraulic conductivity was reasonably predicted by our machine learning based pedotransfer function models. We further analyzed significance of the differences in soil hydraulic properties in water efficiency by running numerical simulations of soil water with the HYDRUSsoftware.