MUONITH v1.0 — A Fast Simulation Tool for Multi-directional Muography Observation Design

Seigo Miyamoto , Shogo Nagahara

Multi-directional muography is a non-destructive technique for imaging the three-dimensional density structure of large objects such as volcanoes. Recent developments in detector technology—miniaturization, low-power operation, and weight reduction—are rapidly expanding the range of feasible observation sites, and detector deployment by unmanned aerial vehicles (UAVs) is also being studied; in the near future, the number of observation-resource allocation patterns to be considered is expected to increase explosively. Yet for targets representable as a digital elevation model (DEM), no existing tool estimates—rapidly, quantitatively, and independently of detector type—how accurately the target structure can be resolved for each of these many candidate resource allocations.

We present MUONITH (MUOgraphic Numerical Inversion and Tomography Harness), a high-speed simulation platform that estimates the performance of diverse detector configurations for DEM-based targets. The characteristic outputs of v1.0 are the following two: (i) a resolution metric visualizing single-direction detectability on the depth–anomaly-size plane, and (ii) three-dimensional density reconstruction for multi-directional observations via regularized linear inversion with maximum a posteriori (MAP) estimation. Its speed comes from the default flux model based on a continuous slowing down approximation (CSDA) incorporating the mean radiative losses (bremsstrahlung and pair production), which reduces flux evaluation to a two-dimensional lookup table indexed by density-length and zenith angle, together with a C++20 implementation combining Digital Differential Analyzer ray tracing, OpenMP parallelization, and OpenBLAS-accelerated linear algebra.

The culmination of MUONITH v1.0 is that a three-dimensional reconstruction that previously required tens of hours on dedicated computers can now be performed in minutes on a consumer laptop. It achieves a two-to-three-order-of-magnitude speed-up at one-quarter the memory of the earlier GNU Octave implementation, enabling individual researchers without dedicated infrastructure to evaluate observation designs. The tool is also useful for narrowing down a few promising parameter sets before running a high-accuracy Monte Carlo simulator.