Spatially representative estimates of saturated hydraulic conductivity, Ks, are needed for simulating surface runoff and infiltration at different scales. This quantity is highly variable in space with values that can vary even more than two orders of magnitude at small catchment scale as well as at the plot or hillslope scale. The spatial variation of Ks is linked with both the basic soil properties and soil use and in addition with the presence of macropores and other preferential flow paths that are difficult to quantify. Therefore, the saturated hydraulic conductivity is considered a random variable depending also on random factors. From a theoretical point of view, the most recent areal infiltration approaches available in the literature and incorporated as components of many hydrological models require the assessment of the first two moments of the Ks probability density function. In principle, the estimate of these two parameters would imply the realization of a great number of local Ks measurements. On the other hand, many devices commonly used rely on the attainment of a steady-state flow rate, so that the time taken for each measurement is usually very large. In this context, it is important to determine in a given area the appropriate number of Ks measurements required for estimating the two moments. We previously performed an analysis for the first moment (the Ks mean) that is extended in this study to the Ks coefficient of variation linked with the second moment (the variance). On the basis of a dataset of 69 Ks observations on three grassy plots of a small Austrian catchment an analysis of uncertainty based on the non-parametric bootstrap method has been performed. It has allowed to estimate the 95% confidence interval around the coefficient of variation of Ks for different numbers of observations and different plot sizes. The outcomes have shown that the width of the normalized confidence interval obtained with a specific number of measurements is almost invariant with plot size while decreases with increasing the number of measurements in a specific area. The presented approach defines a methodology useful for determining the minimum number of measurements to be carried out in an area of specific dimensions to have a fixed level of uncertainty in the estimate of Ks coefficient of variation.
Statistical Characterization of the Saturated Hydraulic Conductivity at the Plot Scale in Natural Grassy Soils
Flammini A.
;Morbidelli R.;Corradini C.;Saltalippi C.;Dari J.
2020
Abstract
Spatially representative estimates of saturated hydraulic conductivity, Ks, are needed for simulating surface runoff and infiltration at different scales. This quantity is highly variable in space with values that can vary even more than two orders of magnitude at small catchment scale as well as at the plot or hillslope scale. The spatial variation of Ks is linked with both the basic soil properties and soil use and in addition with the presence of macropores and other preferential flow paths that are difficult to quantify. Therefore, the saturated hydraulic conductivity is considered a random variable depending also on random factors. From a theoretical point of view, the most recent areal infiltration approaches available in the literature and incorporated as components of many hydrological models require the assessment of the first two moments of the Ks probability density function. In principle, the estimate of these two parameters would imply the realization of a great number of local Ks measurements. On the other hand, many devices commonly used rely on the attainment of a steady-state flow rate, so that the time taken for each measurement is usually very large. In this context, it is important to determine in a given area the appropriate number of Ks measurements required for estimating the two moments. We previously performed an analysis for the first moment (the Ks mean) that is extended in this study to the Ks coefficient of variation linked with the second moment (the variance). On the basis of a dataset of 69 Ks observations on three grassy plots of a small Austrian catchment an analysis of uncertainty based on the non-parametric bootstrap method has been performed. It has allowed to estimate the 95% confidence interval around the coefficient of variation of Ks for different numbers of observations and different plot sizes. The outcomes have shown that the width of the normalized confidence interval obtained with a specific number of measurements is almost invariant with plot size while decreases with increasing the number of measurements in a specific area. The presented approach defines a methodology useful for determining the minimum number of measurements to be carried out in an area of specific dimensions to have a fixed level of uncertainty in the estimate of Ks coefficient of variation.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
