pedotransfer

Pedotransfer, or pedotransfer functions (PTFs), are predictive models that estimate soil hydraulic properties from readily available soil data. They are used to bridge the gap between soil survey information and hydrological or agronomic modeling by providing estimates of properties that are time-consuming or costly to measure directly.

Typical inputs for PTFs include readily measured soil characteristics such as texture (percent sand, silt, and clay), bulk density, organic carbon content, and sometimes soil depth or porosity. Depending on the model, inputs may also include water content at specific tensions or other basic soil properties. Outputs commonly estimated by PTFs include the soil water retention curve at key matric potentials, field capacity, wilting point, available water capacity, and the saturated hydraulic conductivity, as well as parameters describing soil pore size distribution.

PTFs are derived using various methods, ranging from traditional statistical approaches like linear or non-linear regression to machine learning techniques such as neural networks and regression trees. A well-known early example is the Saxton and Rawls series for water retention and plant available water, while more comprehensive approaches include the Rosetta pedotransfer model developed by Schaap, Leij, and van Genuchten that estimates multiple hydraulic properties from texture and related inputs.

Applications span hydrological modeling, irrigation planning, groundwater recharge estimation, soil erosion and contaminant transport modeling, and regional or global soil property databases where measured data are sparse. The accuracy and applicability of PTFs depend on the calibration dataset and may be region- or soil-type specific. Limitations include restricted transferability, sensitivity to input quality, and the potential to propagate measurement errors or misrepresent soil heterogeneity.

---