Analyzing dual porosity in soil hydraulic properties using soil databases for pedotransfer function development.

Current databases of soil hydraulic properties (SHPs) have typically been used to develop pedotransfer functions (PTFs) to estimate water retention [θ(h)] assuming a unimodal pore‐size distribution. However, natural soils often show the presence of bimodal to multimodal pore‐size distributions. Here, we used three widely spread databases for PTF development: UNsaturated SOil hydraulic DAtabase (UNSODA) 2.0, Vereecken, and European hydropedological data inventory (EU‐HYDI), to analyze the presence of structural effects in both θ(h) and hydraulic conductivity [K(h)]. Only undisturbed samples were included in the analysis that contained enough datapoints for both θ(h) and K(h) properties, especially in the wet range. One‐hundred ninety‐two samples were suitable for our analysis, which is only 1% of the total samples in the three databases. Results showed that 65% of the samples exhibited a bimodal pore‐size distribution, and bimodality was not limited to fine‐textured but also coarser‐textured soils. The Mualem–van Genuchten (MvG) expression for both unimodal and bimodal soils was not able to predict the observed unsaturated K. Only a joint fitting of measured θ(h) and K(h) functions provided parameter estimates that were able to describe unsaturated K for uni‐ and bimodal soils. In addition, we observed a negative relationship between α and n in the case of low sand content (<52%) for both unimodal and bimodal matrix domain properties, contradicting the classical notion. The ratio of α for the macropore and matrix domain was positively correlated with the fraction of macropores and sand content. We anticipate that the results will contribute to deriving PTF for structured soils and avoid unrealistic combinations of MvG parameters. Core Ideas: Only 1% of the samples in the three databases was suitable for analyzing the presence of bimodality.Bimodality was seen in 65% of the samples, both in fine‐textured and coarser‐textured soils.Joint fitting of θ(h) and K(h) data outperforms the fitting of only θ(h) and using MvG function to estimate K(h).Contradicting the classical notion, a negative relationship between the shape parameters α and n was observed for both uni‐ and bimodal properties.α for macropore and matrix domains was positively correlated with the fraction of macropores and sand content. [ABSTRACT FROM AUTHOR]