Investigating the potentialities of Monte Carlo simulation for assessing soil water content via proximal gamma-ray spectroscopy

Proximal gamma-ray spectroscopy recently emerged as a promising technique for non-stop monitoring of soil water content with possible applications in the field of precision farming. The potentialities of the method are investigated by means of Monte Carlo simulations applied to the reconstruction of gamma-ray spectra collected by a NaI scintillation detector permanently installed at an agricultural experimental site. A two steps simulation strategy based on a geometrical translational invariance is developed. The strengths of this approach are the reduction of computational time with respect to a direct source-detector simulation, the reconstruction of $^{40}K$, $^{232}Th$ and $^{238}U$ fundamental spectra, the customization in relation to different experimental scenarios and the investigation of effects due to individual variables for sensitivity studies. The reliability of the simulation is effectively validated against an experimental measurement with known soil water content and radionuclides abundances. The relation between soil water content and gamma signal is theoretically derived and applied to a Monte Carlo synthetic calibration performed with the specific soil composition of the experimental site. Ready to use general formulae and simulated coefficients for the estimation of soil water content are also provided adopting standard soil compositions. Linear regressions between input and output soil water contents, inferred from simulated $^{40}K$ and $^{208}Tl$ gamma signals, provide excellent results demonstrating the capability of the proposed method in estimating soil water content with an average uncertainty <1%.

💡 Research Summary

The paper presents a comprehensive study on using proximal gamma‑ray spectroscopy for continuous, non‑destructive monitoring of soil water content (SWC) in precision agriculture. The authors develop a two‑step Monte Carlo (MC) simulation framework that exploits geometrical translational invariance to dramatically reduce computational effort compared with a full source‑detector model. In the first step, an infinite planar soil source uniformly containing the natural radionuclides ^40K, ^232Th, and ^238U is simulated to obtain energy‑dependent gamma‑ray transport functions. These functions are then linearly transformed according to the actual detector height and lateral distance, reflecting the invariance of the transport problem under translation. The second step uses the transformed transport functions as inputs to a detailed model of a NaI(Tl) scintillation detector, accounting for crystal size, energy resolution, and background. This approach yields the fundamental spectra of each radionuclide, allowing the authors to isolate the contributions of the 1460 keV ^40K line, the 2614 keV ^208Tl line (from the ^232Th decay chain), and the 1764 keV ^214Bi line (from the ^238U chain).

The simulation framework is validated against field measurements taken at an agricultural experimental site in Tuscany, where a NaI detector was permanently installed 1 m above the soil surface. Soil samples collected concurrently provided ground‑truth SWC (by gravimetric drying) and radionuclide concentrations (by high‑purity germanium spectroscopy). Comparison of simulated and measured spectra shows peak‑area differences below 2 % and an average absolute error in SWC of 0.8 %, confirming the reliability of the MC model.

A theoretical relationship between the gamma signal S and SWC w is derived as S = S₀ exp