Markov-chain simulation of soil textural profiles

Abstract The spatial change of textural layers is an important feature of alluvial soils. It influences soil water transport and solute movement in the field. Data from one or several soil profiles cannot explain processes of field-water and solute movement in a large region because of spatial variability within textural layers. Two stochastic models—the MC (Markov-chain) model and the MC-LN (Markov-chain - Lognormal distribution) model were developed in this study using Markov-chain theory to simulate textural profiles of regional alluvial soils. Soil textural layers were divided into six types, namely, sand, sandy loam, light loam, medium loam, clay loam, and clay. The two models reflect vertical changes of textural layers of soil profiles in the research region. Excluding the data related to clay loam layers which are quite rare, element values of TPM (transition probability matrix) of textural layers simulated by the MC model deviate less than 15% from measured, and those by the MC-LN model less than 27%. The MC model shows a larger deviation in the probability distribution of textural layer thickness because the simulated textural layer thickness is inclined to be exponentially distributed, but the MC-LN model is consistent to the measured data, which are similar to a lognormal distribution for each textural type. Simulated mean thicknesses of textural layers from the MC-LN model deviate less than 21.1% from measured, and those from the MC model less than 7.1%. These two models will be useful for further constructing a regional model of field water transport with a consideration of the spatial variability of soil textural layers.