Converting Cosmic Ray Neutron Sensing count rates into average topsoil moisture contents when vertical soil moisture gradients occur in the topsoil

Cosmic Ray Neutron Sensing (CRNS) is becoming a well-established technique to measure topsoil soil moisture content at the multihectare scale. There are prospects that cheaper devices can be developed so the technique could also be used outside of the research community, e.g., for providing soil moisture information in real-time for irrigation scheduling purposes. A limitation for such applications is that the soil depth over which the CRNS technique measures, depends on the soil moisture content (deeper in drier soil) and that it produces a depth-weighted average moisture content, with the weights rapidly decreasing with depth.

For practical applications like irrigation scheduling, it will be necessary to translate the CRNS count in an average soil water content of the topsoil (for, say, the top 20 cm or 30 cm), with an associated confidence interval that accounts for the fact that depth-weighting leads to uncertainty. That is because different average soil moisture contents in the topsoil can produce the same CRNS count: a wetter topsoil layer with a drying front at the top can produce the same CRNS count rate as a much dryer topsoil layer in which an infiltration front is developing at the top. At present, soil moisture contents are always derived from CRNS count rates using the well-established relationship that is strictly speaking only valid when the soil moisture content does not vary with depth.  No procedure is available to account for soil moisture gradients that develop all the time in topsoil under natural conditions. These gradients lead to non-uniqueness of the relationship between the CRNS count and the average soil moisture content of a topsoil layer.

We investigated this non-uniqueness by running simulations with the Hydrus1D-COSMIC model for different soil textures with multi-year time series of daily weather data and sequences of cropped and bare soil conditions similar to the ones occurring in the field. We observed a pronounced clockwise hysteretic relationship when plotting the average (simulated) 20-cm or 30-cm topsoil moisture content versus the CRNS count rate (N/N0) derived with COSMIC from the simulated vertical soil moisture content distribution. Strategies were explored to deal with this hysteretic behaviour and to translate the CRNS count into an average soil water content of the topsoil with an associated confidence interval that accounts for the non-uniqueness.