Most field crop phenotyping research has focused on the above-ground parts of crops, ignoring a “hidden half”: the rooting system and its activity. Here we propose and test a new approach to produce 3D characterizations of crop water use and root activity in large field genotype (G) by environment (E) by management (M) experimentation, using electromagnetic induction (EMI) instrument coupled with a quasi-2D inversion algorithm, and crop canopy sensing technologies. A root activity factor (R) was calculated as a function of crop water use, soil water availability, and an indicator of crop demand. We ask i) can this approach provide accurate 3D characterizations of sorghum water use and root activity?, and (ii) does the approach capture complex GxExM dynamics?. This study was conducted based on an on-farm field experiment consisting of the factorial combination of six commercial sorghum genotypes (G), three times of sowing, two levels of irrigation (E), four plant densities (M), and three replications. Two EMI surveys ten days apart were collected using a DUALEM-21S sensor. An artificial neural network (ANN) model was developed to predict 3D soil moisture (θv) using depth-specific true soil electrical conductivity (σ, mS m−1) estimated by the inversion algorithm. Crop water use between surveys was described as the difference of θv. A multispectral index derived from satellite imagery was used as a proxy for crop demand i.e., size of the crop canopy. Principal components analysis, linear mixed models, and recursive partitioning tree techniques and crop-eco-physiological principles were used to untangle complex GxExM interactions. Results indicate that 3D crop water use could be predicted with high accuracy (LCCC = 0.81) and low prediction error (RMSE = 0.03 cm3 cm−3). The calculated water use and the value of R were significantly affected by depth, crop growth stage, irrigation treatment, plant density, and their interactions. At flowering, roots were most active at 0–1.3 m under irrigation, and deeper (0.5–1.5 m) under dryland treatment. The highest water use was for three genotypes (i.e., C, E and F) grown under irrigation and high plant densities (i.e., 9 and 12 pl m-2). The smallest water use was observed under dryland treatment, particularly for two genotypes (i.e., B and C) and high plant densities. For the crops at vegetative stages, the values of water use and R were highest in the top 0.5 m of soil depth. Larger water use was observed under dryland treatment and high plant densities, while the effects of genotypes were small (not significant). We conclude that the approach provides a rapid, accurate and cost-efficient option to phenotype crop root activity i.e., water use, in large field experimentation. We also argue that the improved understanding of the crop water use dynamics can help inform optimum combinations of genotypes and management options i.e., crop designs, across contrasting environments, and help untangle complex GxExM interactions.