Soil structures and pedotransfer functions

Soil structural features implemented in new pedotransfer functions

Most soil functions are dependent on maintenance of favorable soil structure; the fragile assembly of biopores and aggregates glued by organic matter (see some examples in Figure 1). The creation of favorable soil structure for crops and soil ecological functioning motivate tillage operations, often repeated annually as a central agronomic activity that defines arable soils (10% of all terrestrial surfaces). From a hydrologic point of view, the presence of large pores and biopore networks influence the partitioning of rainfall to infiltration and runoff and, hence, potential recharge relative to soils lacking structure. It is reasonable to expect that the storage and mean water content in structured soils would be different compared to unstructured soils, thus affecting energy partitioning over such land surfaces.

[Fig 1]

Figure 1: Examples of soil structures at different scales that will determine soil hydraulic properties but cannot be determined based on usual pedotransfer functions.

The representation of land-surface properties in hydrologic and climatic models relies heavily on derivation of hydraulic parameter values (e.g., infiltration rates) or soil hydraulic properties (SHP) from auxiliary and easy-to-measure soil properties such as soil type or soil texture using statistically-based pedotransfer functions (PTF). PTF’s correlate soil texture, bulk density, organic matter and alike to soil hydraulic parameters used as inputs into hydrologic and climate models. Such simplified information “transfer” involves many assumptions, but omits the ubiquitous soil structure in most PTFs and reliance on soil texture. Hence, the general goal of this proposal is to develop new methodologies for generating PTFs of hydraulic soil parameters that include soil structural properties and satisfy physical constraints.

The primary objective of the project is to formulate new pedotransfer functions that include soil structural features (SPTF) using surrogate variables Gross Primary Productivity (GPP). We will evaluate several statistical approaches (linear regression, additive models, neural networks, tree-based ensemble methods such as random forest, Bayesian and boosted trees) to build the SPTF, linking comprehensive sets of covariates with data on soil hydraulic properties.

Team: Dani Or, Peter Lehmann, Andeas Papritz, Sara Bonetti
Funding: ETH Research grant
Contact: