Validation of water sorption-based clay prediction models for calcareous soils

Soil particle size distribution (PSD), particularly the active clay fraction, mediates soil engineering, agronomic and environmental functions. The tedious and costly nature of traditional methods of determining PSD prompted the development of water sorption-based models for determining the clay fraction. The applicability of such models to semi-arid soils with significant amounts of calcium carbonate and/or gypsum is unknown. The objective of this study was to validate three water sorption-based clay prediction models for 30 calcareous soils from Iran and identify the effect of CaCO3 on prediction accuracy. The soils had clay content ranging from 9 to 61% and CaCO3 from 24 to 97%. The three water sorption models considered showed a reasonably fair prediction of the clay content from water sorption at 28% relative humidity (RMSE and ME values ranging from 10.6 to 12.1 and -8.1 to -4.2, respectively). The model that considers hysteresis had better prediction accuracy than the other two that do not. Moreover, the prediction errors of all three models arose from under-prediction of the clay content. The amount of hygroscopic water scaled by clay content decreased with increasing CaCO3 content. The low organic carbon content of the soils and the low fraction of low-activity clay minerals like kaolinite suggested that the clay content under-predictions were due to large CaCO3 contents. Thus, for such water-sorption based models to work accurately for calcareous soils, a correction factor that considers the reduction of water content due to large CaCO3 content should be included.