Evaluation of a 2 GHz Prototype Ground Penetrating Radar System for Estimating Cassava Root Biomass

Accurate assessment and estimation of root biomass of tuber crops like cassava and potato is essential for agricultural management activities and plant breeders who need to evaluate bulking rates associated with genotyping. Traditional biomass measurement methods are labor intensive and destructive, although Ground Penetrating Radar (GPR) systems offer new opportunities to quantitatively assess root architecture and biomass. Most GPR applications, however, focus on detection and do not address the characterization of root architecture and biomass due to difficulties in preprocessing, analysis and information extraction. This research focuses on establishing and evaluating new analytical solutions for tuberous root detection and biomass estimation from multi-channel GPR data. Specifically, we estimate cassava biomass based on frequency and amplitude information extracted from multi-channel GPR radargrams, and validate our approach using field measurements. In order to capture enough details of subsurface structures, cassava plots in Colombia was scanned with a newly designed 2 GHz multi-array cart system. The data were preprocessed using an optimized workflow which includes automated surface detection, gain correction, background correction, bandpass filtering and F-K migration. The optimization was based on a sensitivity test to evaluate how each preprocessing step influenced our ability to characterize biomass variation. Both migrated B-scans and time slices were then submitted to analysis algorithms to extract frequency and amplitude signatures that are indicative of biomass. Preliminary results indicate that cassava can be detected on both B-scans and time slices with proper preprocessing, and there was a significant negative correlation between signal spatial frequency and field-measured root biomass. These results indicate that is possible to assess tuberous root bulking rates using multi-temporal GPR data.