Tracking Evaporation Evolution in Soil Columns using Dose Reduced Neutron Tomography

Neutrons provide a unique probe of matter with a particularly strong sensitivity to hydrogen making neutron imaging a powerful method to characterize water content in geological samples. Because the brightness of typical neutron sources is low long exposure times are required to achieve high-quality images. This represents a significant temporal limitation to imaging samples that undergo changes on short time scales such as water transport in porous networks. An avenue to circumvent this limitation is to employ neural networks to restore the signal quality that is degraded by accelerating tomogram acquisition. This work demonstrates the use of this technique toward characterizing water transport in a soil column over the course of seven dose-reduced scans. A high-resolution seed dataset was collected to train the network that was subsequently used to de-noise the rapidly collected tomograms. Tomogram acquisition time was cut from an hour down to 8 minutes using this technique, which will allow greater temporal fidelity on the progression of evaporation.