Premise of the study: Selective breeding over thousands of years has prioritized above-ground yield, with little regard for changes happening below-ground. Despite their central role in plant success and resilience, our knowledge of roots lags behind above-ground structures. Accurately phenotyping root traits is labor-intensive, expensive, and destructive. In order to advance understanding of the fundamental biology underlying root systems, and to integrate hard-to-measure root traits into breeding programs, high-throughput non-destructive methods are required. Methods: This study uses American licorice (Glycyrrhiza lepidota Pursh.), a perennial legume with a rich ethnobotanical history, as a model to investigate root system phenotypes. We assess root traits across multiple populations, analyze relationships between above- and below-ground phenotypes, and test the use of multidimensional leaf traits, including spectral reflectance, in predicting root traits. Key results: American licorice displays significant variation in root traits across source populations and strong correlations between above- and below-ground traits. Leaf spectral reflectance and elemental composition show promise in modeling below-ground traits, though the isometric relationship between plant size and root traits complicates model accuracy. Conclusions: These findings demonstrate the use of high-dimensional leaf traits as a proxy for root traits, with potential applications for understanding foundational questions in plant biology and in breeding programs targeting the below-ground structures of perennial herbaceous species. Further optimization and larger studies are needed to improve predictive models.This repository contains a .zip archive of the raw images used for analysis. Scripts and other associated data can be found at https://doi.org/10.6084/m9.figshare.28742870.