Calibrating pedotransfer functions to estimate soil hydro limits using limited data

Abstract

The water retained in the soil is determined by factors such as soil texture, structure, organic matter content, clay content and its type. However, laboratory or in-situ determination of water retention curve or hydro limits is an exhaustive process with time, manpower and capital requirement. Researchers, therefore, prefer indirect estimation of soil moisture retention using pedotransfer functions (PTF). PTFs can be defined as predictive functions of certain soil properties from other easily, routinely, or cheaply-measured properties (Minasny and McBratney 2002). Regression tools are often used for establishing such relationships. Regression analysis requires prior knowledge of relationship or at least expected relationship and assumptions to be made regarding probability distribution of the errors. Statistical tests are made on the basis of these assumptions. New methods like artificial neural networks (ANN) do not require prior knowledge or assumptions about error distribution. An ANN is configured for a specific application such as pattern recognition or data classification, through a learning process. It then mimics the relationship between related variables learnt from historical data.