Correction efficiency and error characteristics for cosmic-ray soil moisture on mountainous terrain

Abstract

A Cosmic-Ray Neutron Probe (CRNP) is a promising tool for obtaining reliable intermediate scale soil moisture. Although there are sound ways of correcting neutron counts and calibrating soil moisture, it is difficult to obtain all the data required for corrections and calibration in a complex environment. In this study, three major correcting variables (i.e., absolute humidity, air pressure, incoming neutron flux intensity) were evaluated to determine their impacts on the neutron counts in the Seolmacheon experimental site (SMC) located in mountainous terrain. Efficiency of correction factors for atmospheric water vapor (fwv), air pressure (fp,), and incoming neutron flux intensity (fi) was independently assessed through comparisons between the variables and a inversely calculated N0 (N0,inv). Particular attention was given to the expected biases and errors in estimated soil moisture (SMCRNP) with and without each correction factor, especially at the early stage of the calibration period. Air pressure had the greatest impact on neutron count at the SMC site, followed by absolute humidity. The dependency of neutron count on air pressure (absolute humidity) was found from the inverse relationship with N0,inv that is not corrected for air pressure (absolute humidity), with correlation coefficient −0.88 (−0.78). When fp (fwv) was applied, the dependency of neutron count on the corresponding variable was significantly reduced, with a correlation coefficient 0.06 (−0.07). On the other hand, the effect of applying fi was contrary, slightly degrading the result. When neutron count was not corrected for air pressure (absolute humidity), the performance of SMCRNP decreased, with a correlation coefficient of 0.74 (0.90) and Root Mean Square Error (RMSE) of 0.030 m3/m3 (0.026 m3/m3) at best. The best result was obtained when both of them were corrected (r = 0.95, RMSE = 0.014 m3/m3). As data accumulates, error metrics in terms of bias and variability converged at approximately one periodic cycle, implying at least one seasonal cycle is necessary for recalibration. However, in a case where a variable with large intra-seasonal variability (e.g., air pressure in this study) is missing, it is recommended to adopt a multi-point N0 method.The procedures suggested in this study allow independent evaluations of the factors for which neutron count is corrected, thus providing a basis for judging which factors are appropriate and which are not. In addition, N0 can be more efficiently tuned according to the characteristics of the errors in SMCRNP. Based on these analyses, CRNP stations will be able to selectively apply favorable correction factors under various kinds of local characteristics, making it easier to achieve target accuracy and precision.