Assessment of long-time series of soil water content through an innovative robust statistical model

The aim of this research is testing a new statistical model able to describe the interaction between soil and atmosphere. The model is based on a robust parametric LTS (Least Trimmed Squares) method and harmonic functions. It has been developed taking into account field measurements of quantities involved in both infiltration and evapo-transpiration phenomena, such as volumetric water content, soil-water potential, air temperature, rainfall amounts and solar radiation. The proposed statistical model allows assessing the volumetric water content at different sites using only time series of daily rainfall amount as input data. The model was applied in different test sites, whose data were assumed by the International Soil Moisture Network (ISMN). In fact, ISMN allows getting free time series of soil and meteorological data from monitoring stations all over the world. This note shows how the proposed model is accurate with respect to field data in estimating the volumetric water content in different soils, climates and depths. Future implications of this research will regard water content predictions, especially in areas where field data are scarce. Since the proposed LTS algorithm is very efficient and the computational workload is rather low, the possibility of coupling it with a slope stability analysis over large areas will be investigated, in order to get a distributed real-time model for shallow landslides susceptibility.