A Bimodal Four-parameter Lognormal Linear Model Of Soil Water Repellency Persistence

摘要

Soil water repellency may be characterized in terms of the delayed infiltration time of a water droplet resting on the soil surface, which is, water drop penetration time (WDPT), or repellency persistence. Such repellency persistence varies nonlinearly with soil water content (θ_g), although no models have been proposed to reproduce the variation of WDPT with θ_g in soils. Dynamic factor analysis (DFA) is used to identify two common patterns of unexplained variability in a scattered dataset of WDPT versus θ_g measurements. A four-parameter lognormal distribution was fitted to both common patterns obtained by DFA, and these were combined additively in a weighted multiple linear bimodal model. We show how such an empirical model is capable of reproducing a large variety of WDPT versus θ_g curve shapes (N = 80) both within a wide range of measured WDPTs (0-17 000 s) and for samples with organic matter content ranging from 21.7 to 80.6 g (100 g)~(-1).