Models to Predict Water Retention in Semiarid Sandy Soils1

Objectives of this research were to (i) develop statistical models for predicting water content at three matric potentials, using three levels of input data; (ii) compare predictive capability of the locally calibrated models with those of three published models; and (iii) illustrate the changes that occur in water retention when texture of sandy soils is altered by deep plowing or wind erosion. At −10 kPa, accuracy of predicting gravimetric water contents increased with increased number of inputs. The R 2 s and the slopes between predicted and measured water contents for 50 samples increased, and the intercepts came closer to zero as the inputs progress from clay only to soil texture plus organic matter and CaCO 3 and on to sand separates plus texture, organic matter, and CaCO 3 . At the −33‐ and −1500‐kPa matric potentials, the clay‐only model predicted soil‐water content equally as well as the two local models with more inputs. A previously published clay‐only model that had been calibrated using semiarid soils performed almost as well as the locally calibrated models. Two other published models performed unsatis‐factorily in predicting water retention at any of the matric potentials. Deep plowing increased clay content of one specific soil from 4 to 14%. Wind erosion subsequently reduced clay to 6% after 5 yr. Predicted water contents were 0.281, 0.341, and 0.293 kg kg −1 , respectively, at −10 kPa, and 0.012, 0.050, and 0.020 kg kg −1 , respectively, at −1500 kPa.