Cokriging Spatial Interpolation of Soil Water Dependent Repellency Parameters Determined with Two Different Tests

Abstract

The nonlinear behavior of water repellency (WR) with respect to soil water content (θg) may be described by curve parameters derived from two widespread methodologies such as the water drop penetration time (WDPT) and molarity of an ethanol droplet (MED) tests. While the former measures WR persistence in terms of the infiltration time of a water drop sitting on the soil surface, the latter quantifies the repellency degree as contact angle or 90° surface tension, γ90° From a hydrologic point of view, the WDPT test may be considered more closely related to the natural infiltration process of the soil, but the infiltration times become prohibitively large (on the order of hours) in highly repellent soils; thus, the MED test is more convenient. This study was conducted to determine whether the spatial interpolation of time‐consuming WDPT measurements may be complemented with less time‐demanding WR parameters derived from the MED test. Measurements from saturation to oven dryness using both MED and WDPT were performed on 140 topsoil samples from a humid evergreen forest watershed of the Garajonay National Park (Canary Islands). We found useful links between curve parameters derived from the MED and WDPT tests. The area below the WR–θg curve measured with the WDPT test (SWDPT) was found to be correlated with that measured with the MED test (SMED) and with the θg at which repellency is triggered (θg‐min). By contrast, previously published relations that established a logarithmic relationship between 1/WDPT and γ90° were valid only in the dry soil state, but must be ruled out if soil water content is taken into account. We took advantage of the existence of such parameter correlations to show how cokriging geostatistical tools may be applied to improve the spatial interpolation of SWDPT, with SMED or θg‐min as subsidiary variables. In general, cokriging was superior to kriging in terms of mean variance minimization with little extra labor cost. A minimal‐effort spatial sampling strategy may be designed taking into account these considerations.