Extension of cosmic-ray neutron probe measurement depth for improving field scale root-zone soil moisture estimation by coupling with representative in-situ sensors

Abstract

Accurate estimation of field-scale root-zone soil moisture (RZSM) is essential for hydro-meteorological and agricultural management. The cosmic-ray neutron probe (CRNP) is an innovative technique for field-scale soil moisture observation. However, it is limited at deeper depths, and requires suitable methods for scaling the CRNP effective depth to represent the root-zone layer (up to 100 cm depth). A merging framework was developed in this study for improving field-scale RZSM with the CRNP via coupling the representative ancillary RZSM information with cosmic-ray soil moisture. By using ancillary RZSM retrieved at the most time stable location, this approach alleviates the problems associated with lack of independent datasets, while maintaining scale representativeness. A linear autoregressive model was adopted to forecast the errors between two input datasets (the cosmic-ray soil moisture and ancillary RZSM) and a reference product, which was computed by spatially weighting the deepest in-situ soil moisture measurements apart from the most time stable location. The variances of estimated errors were then used to compute a suitable weight for each single product by following the original linear combination of forecasts, and subsequently merging the ancillary RZSM and cosmic-ray soil moisture. Performance of the merged RZSM in comparison to input datasets and exponential filter-based RZSM over three different environments was evaluated against the reference RZSM product. The results indicated that differences in vegetation coverages related to total aboveground and belowground biomass accumulation, root water uptake rate and canopy density are the major factors controlling the temporal variation in merged RZSM, and they can be partly interpreted via the framework procedure. Superior performance achieved by the merging framework demonstrated its robustness in improving field-scale RZSM measurement compared to other products. This study also underlines the strong relationship between input data quality and the performance of the selected merging method, with respect to variations of CRNP effective depths. Overall, the merging framework is simple to apply, enables unrestricted-use in different environments, and is flexible to combine further standalone data sources.