Data-driven scaling methods for soil moisture cosmic ray neutron sensors

Abstract

Cosmic ray neutron sensors (CRNSs) are state-of-the-art tools for field-scale soil moisture measurements, yet uncertainties persist due to traditional methods for estimating scaling parameters that lack the capacity to account for site-specific and sensor-specific characteristics. This study introduces a novel, data-driven approach to estimate key scaling parameters (beta, psi, and omega) by directly calculating scaling parameters from measurement data, emphasizing local environmental factors and sensor attributes. The method demonstrates reliability and robustness, with strong correlations between estimated scaling parameters and environmental factors such as cutoff rigidity, latitude, and elevation, as well as consistency with semi-analytical traditional methods, e.g. for beta an R2 of 0.46. The study also reveals systematically higher variability in calibration parameters than previously assumed, underscoring the importance of this new method, of data quality, and of the careful selection of Neutron Monitor Database (NMDB) reference sites. The new method reduces RMSE by up to 25 %, with differences in soil moisture estimates between traditional and data-driven methods reaching 0.04 m3m-3 and up to 0.12 m3m-3 under certain conditions. Sensitivity analysis shows that soil moisture estimation is most influenced by scaling parameters in the wet end of the soil moisture spectrum. By improving the accuracy of CRNS data, this approach enhances soil moisture estimation and supports better decisions in agriculture, hydrology, and climate monitoring. Future research should focus on refining these scaling methods and enhancing data quality to further improve CRNS measurement accuracy.