A key message from scientific communities and industrial partners is that future transformative impacts in earth, energy, and environmental modeling may be realized using machine/deep learning (ML) to improve linkages between data, prediction, and decision making. We seek case studies illustrating how ML workflows promote assimilation of information in multimodal field measurements, discover knowledge about subsurface processes including linkages to natural and human pressures, provide quantitative information over field-relevant spatial scales, and provide important model input that improves predictions. Important model input could include one or more of the following: (1) identifying influential model parameters, (2) identifying calibration measurements with mutual information about parameters, (3) providing knowledge or estimation of boundary conditions, (4) minimizing multimodal objective function topologies, (5) reducing computational resources required for field-scale simulations, and (6) quantifying prediction uncertainty. Successful ML applications will render informed policy and sustainable earth, energy, and environmental resource management.