Simultaneous soil moisture and properties estimation for a drip irrigated field by assimilating cosmic-ray neutron intensity

Abstract

Neutron intensity measured by the aboveground cosmic-ray neutron intensity probe (CRP) allows estimating soil moisture content at the field scale. In this work, synthetic neutron intensities were used to remove the bias of simulated soil moisture content or update soil hydraulic properties (together with soil moisture) in the Community Land Model (CLM) using the Local Ensemble Transform Kalman Filter. The cosmic-ray forward model COSMIC was used as the non-linear measurement operator which maps between neutron intensity and soil moisture. The novel aspect of this work is that synthetically measured neutron intensity was used for real time updating of soil states and soil properties (or soil moisture bias) and posterior use for the real time scheduling of irrigation (data assimilation based real-time control approach). Uncertainty of model forcing and soil properties (sand fraction, clay fraction and organic matter density) were considered in the ensemble predictions of the soil moisture profiles. Horizontal and vertical weighting of soil moisture was introduced in the data assimilation in order to handle the scale mismatch between the cosmic-ray footprint and the CLM grid cell.

The approach was illustrated in a synthetic study with the real-time irrigation scheduling of fields of citrus trees. After adjusting soil moisture content by assimilating neutron intensity, the irrigation requirements were calculated based on the water deficit method. Model bias was introduced by using coarser soil texture in the data assimilation experiments than in reality. A series of experiments was done with different combinations of state, parameter and bias estimation in combination with irrigation scheduling.

Assimilation of CRP neutron intensity improved soil moisture characterization. Irrigation requirement was overestimated if biased soil properties were used. The soil moisture bias was reduced by 35% after data assimilation. The scenario of joint state-parameter estimation resulted in the best soil moisture characterization (50% decrease in root mean square error compared to open loop simulations), and the best estimate of needed irrigation amount (86% decrease in Hausdorff distance compared to open loop). The coarse scale synthetic CRP observation was proven to be useful for the fine scale soil moisture and soil properties estimation for the objective of irrigation scheduling