Estimating soil water retention function from its particle-size distribution

Abstract

The soil water retention function or soil water retention curve (SWRC), which describes how much volumetric water (θ) is held within the soil at a given suction or matric potential (h), is one of the important hydraulic properties for characterizing and modeling water flow and solute transport in soils. SWRC is difficult to measure directly, i.e., the pressure plate extraction method, because the technique is expensive, time-consuming, and laborious process. Instead, SWRC has been predicted from easily obtainable soil properties. From previous studies, however, there were some limitations such as low predicting power and restricted to apply only sandy soil. In this study, SWRC was estimated from particle-size distribution (PSD) based on similarity of distributional shape with SWRC that van Genuchten suggested. Data for conduct study were selected from the UNSODA, which contain more than five data points and saturated water contents (149 datasets were selected

103 datasets for calibration and 46 datasets for validation). ii Verification was conducted additionally with data from previous studies (Ariana silty clay loam and Yolo loam) and experimentally obtained data (Bancheon silty clay, Upyeong silty clay, Chusan clay loam). From calibration dataset, PSD and SWRC were fitted independently and Pearson product moment correlation coefficients (R2) of each model were shown 0.987 and 0.965, respectively, which means estimation was appropriate based on the van Genuchten model. Shape-related parameters, m, for cumulative PSD and SWRC were shown nonlinear relationship each other. In contrast, any relationship between inflection points of each cumulative PSD and SWRC could not be found. Alternatively, particle-size and reciprocal of matric head were partial linearly related at the point of 43 % of each normalized cumulative distribution. Root mean square of residuals (RMSR) of predicted SWRC were 0.091 to entire verification dataset, which was highest in sandy clay (RMSR=0.241) and lowest in silty clay loam (RMSR=0.016). Estimated water contents were relatively smaller than actual contents, because the inflection point was predicted higher than ideal value. Although particle-size and reciprocal of matric head were asymmetrically related in each soil, the relationship was shown very different among the soil survey data. Further researches need to be conducted to solve under-estimation by verifying the relationship between particle- and pore-size that could cover overall soil.