The Synthesis Company of San Francisco Mountain Logo

Abstract

Abstract Mungbean is an important sub‐tropical legume crop grown across Asia, Africa, and Australia. Yield improvement is crucial for expanding production, but phenotyping important traits across diverse environments using current approaches is challenging, limiting the scale and complexity of information captured. High‐throughput phenotyping platforms offer a solution by rapidly screening traits at scale. This study deploys an unmanned aerial vehicle (UAV) platform to determine the potential of phenotyping a range of agronomic and physiological traits within a diverse mungbean population evaluated across three field trials. Three predictive data‐driven modeling approaches were undertaken to evaluate performance accuracy in predicting these traits: linear regression, stepwise regression, and partial least squares regression. Results show that using the geometric trait “coverage” as a proxy is most suitable for screening visual traits like early vigor. For functional traits (i.e., aboveground biomass), predictive data‐driven models demonstrate high accuracy during early‐ and mid‐canopy development stages ( R 2 0.79, root mean square error [RMSE] 4.08 and R 2 0.8, RMSE 26.92, respectively), but accuracy declines in late‐canopy development ( R 2 0.33 and RMSE 43.15). Prediction accuracy can be optimized by using different modeling approaches at different stages during the transition from early‐ to mid‐canopy development as well as canopy closure. Similar findings were observed when examining the prediction models for the physiological trait, stomatal conductance ( R 2 0.69 and RMSE 0.10). These approaches are expected to enable breeders and researchers to incorporate UAV‐based phenotyping systems into mungbean improvement programs. Such approaches might be most efficiently used at scale if applied as part of a “real‐time” calibration approach.

Unmanned aerial vehicle phenotyping of agronomic and physiological traits in mungbean

Citations Over TimeTop 10% of 2025 papers

Abstract

Related Papers