“…An emerging trend to address the subsurface-data scarcity problem is the use of machine-learning approaches. Such approaches have been used to infer subsurface structure, properties, and functioning, including stream water quality, groundwater chemistry and permeability that can potentially be used to infer CZ thickness (Erickson, Burnett, et al, 2021;Erickson, Elliott, et al, 2021;Ouedraogo et al, 2019;Podgorski et al, 2020Podgorski et al, , 2022Wen et al, 2021). Traditional methods such as random forest or generalized boosted regression models are being used (Bergen et al, 2019;Hare et al, 2021;Li et al, 2021) as well as deep-learning models with multiple layers of neural networks that automate pattern extraction (Zhi et al, 2021;Zhi, Ouyang, et al, 2023;Zhi, Klingler, et al, 2023).…”