Predicting water balance in a sandy soil: model sensitivity to the variability of measured saturated and near saturated hydraulic properties

Water balance modelling based on Richards' equation requires accurate description of the soils' hydraulic parameters. Unfortunately, these parameters vary spatially and temporally as well as between measurement techniques. For most field modelling exercises, the hydraulic parameters are obtained from a small number of measurements or predicted from soil properties using pedo-transfer functions. The effect of different measurement techniques on the description of soil hydraulic parameters has been the subject of many studies but the effect of the variability of the hydraulic parameters on the predicted water balance has not been widely investigated. In this study we compared the hydraulic parameters obtained solely from laboratory measurements with those obtained from a rapid wet end field measurement technique, augmented by dry end laboratory data. The water balance was modelled using the laboratory and field hydraulic parameter sets and compared to field water contents measured by time domain reflectometry (TDR). In a sandy soil, we found the total profile water content to be well modelled by both hydraulic parameter datasets, but the water content at a specific depth was less well predicted using either of the measured parameter sets. The water content at a specific depth was under-predicted prior to the rainfall event and over-predicted after the rainfall, regardless of whether the hydraulic parameters were obtained from laboratory or field measurements. Generally, the hydraulic parameters that were obtained from the field measurements gave a closer fit to the measured TDR water contents. The sensitivity of the modelled water balance to changes in the hydraulic parameters within the observed range of parameter values was also investigated. Parameter percentage coefficient of variation within measurement techniques ranged from 60% for air entry, he; 19% for residual water content, theta(r); 5% for slope of the water retention curve, n; and 7% for saturated water content, Os. The percentage differences between the parameters obtained from the laboratory and field measurement techniques for the topsoil and subsoil respectively were 47% and 50% for h(e), 100% for theta(r), 28% and 40% for n, and -14.4% and 4.0% for theta(s). Modelling water content changes at a particular depth in the sandy soil was found to be most influenced by variations in theta(s), and n. Predicted water contents were also affected by the Or but less influenced by the saturated hydraulic conductivity, K-s. The h(e) was the least influential parameter but also the most variable. This suggests that measurement of theta(s), related to bulk density changes caused by tillage, wheel compaction, and consolidation, is required for water balance studies. Generally, n had small variability between measurements at a particular depth, which is promising for the use of pedo-transfer functions related to soil texture.