Testing the accuracy of imaging software for measuring tree root volumes

Abstract

Manually measuring tree root characteristics can be inefficient and limiting. To test the application of a new digital technology in tree root architecture research, root systems from 29 green ash (Fraxinus pennsylvanica 'Patmore') trees were unearthed, cleaned, and photographed to create 3D models using structure from motion (SfM) photogrammetry. Three root segments from each root system were selected, marked, and removed after being photographed. The volumes of these segments (derived from the 3D models) were compared against volumes measured using water displacement. In addition to the root segments, model and water displacement volumes were compared for three complete root systems. Regression analysis showed a strong linear relation between the two volumes measurements (adjusted R2=0.97 for the root segment data). The RMSE for the root segment volume estimates was 40.37cm3 (12.3%), with a bias of 17.2cm3 (5.3%). This error rate was similar to previous published work and suggests the technology used may allow researchers to improve efficiency in data capture, add new measurements (i.e., surface area) to their modeling efforts, and digitally preserve tree root systems for future study.