Assessing the Reliability of Pedotransfer Functions for Predicting Water Retention Variability in Tropical Soils: Implications for Environmental Process Modeling

With the scarcity of observed data, the use of pedotransfer functions (PTFs) to calculate soil hydrological properties has gained popularity. A survey of the literature reveals that most of the PTFs for soil volumetric moisture content (VMC) are developed in subtropical environments. Observed soil moisture measurements at 33 and 1500 KPa levels (VMC33 and VMC1500) for 334 and 332 profiles from Sri Lanka were used to study 15 published PTFs for tropical soils. Luvisols, as the prominent soil type in dry and intermediate zones of the country, covers most of the study area. Since the soil moisture varies with soil texture, the capabilities of PTFs to estimate VMC at the textural class level were also examined and the best PTF for the individual textural class were identified. Based on the global sensitivity analysis, clay content was identified as the most sensitive input parameter in PTFs. Although the sparsity of original observations impedes the creation of valid maps, the analysis of each PTF to re-produce patterns of spatial variability of VMC33 and VMC1500 showed that models that make use of structural soil information can predict moisture levels at unsampled locations. This method will greatly improve the prediction of environmental process modelling.