Simplified evaporation method for determining soil hydraulic properties

Evaporation experiments are commonly used to derive hydraulic properties of soils. In the simplified evaporation method, as proposed by Schindler [Schindler, U., 1980. Ein Schnellverfahren zur Messung der Wasserleitfähigkeit im teilgesättigten Boden an Stechzylinderproben. Arch. Acker- u. Pflanzenbau u. Bodenkd. Berlin 24, 1–7], the weight of a soil sample and pressure heads at two height levels are recorded at consecutive times. The evaluation of these measurements relies on linearization assumptions with respect to time, space and the water content–pressure head relationship. In this article, we investigate the errors that result from the linearization assumptions, and show how systematic and stochastic measurement errors affect the calculation of water retention and hydraulic conductivity data and the resulting fits of soil hydraulic functions. We find that linearization errors with respect to time are negligible if cubic Hermite splines are used for data interpolation. Linearizations in space lead to minor errors, even in the late stage of evaporation where strongly non-linear pressure head profiles emerge. A bias in the estimated retention function results from the negligence of a non-linear water content distribution in the sample at the begin of the evaporation process, and affects primarily coarse sands or soils with structured pore systems. This error can be avoided if an integral evaluation of the measurements is used. We introduce an applicable rejection criterion for unreliable hydraulic conductivity data near saturation, based on the error in the hydraulic gradient. Calibration errors of tensiometers lead to biased estimates of hydraulic properties in the wet range, whereas errors in tensiometer installation positions yield biases in the dry range. Random errors in data cause no significant bias, and parametric hydraulic functions can be estimated with small uncertainties, if water retention and conductivity functions are coupled and the underlying model structure is correct.