GROUND PENETRATING RADAR FOR THE ESTIMATION OF BULKED ROOT MASS

Abstract

Plant root phenotyping is a difficult task because of the opaque nature of soil, leading knowledge of root traits to lag behind above ground traits. While a large multiplicity of methods for observing roots have been described, each is limited in some way. This difficulty in observing plant roots causes difficulty for researchers and breeders working with root crops, such as cassava. Cassava is a crop with a long production cycle and low reproduction rates. As a root crop, assessing root properties such as time to bulking and marketable mass is a major goal for breeders. It has been proposed that GPR could be used to non-destructively estimate cassava root mass. Previous publication has demonstrated the potential for such a tool, but further work is required before GPR can be useful to breeders.

In this body of work we explore the realm of root phenotyping and address the issue of rapid and non-destructive estimation of root mass by GPR. We demonstrate the ability to correlate features of GPR data with bulked root mass in controlled field conditions, and demonstrate the effect of soil water on GPR data in regards to root mass. We proceed to expand that work from controlled conditions into an actual cassava field. There we were able to build upon the predictive model derived in the earlier chapter, confirming that GPR data relate to bulked root mass in unoptimized field conditions. Lastly, we evaluate the ability of predictive models to generalize across locations and genotypes. While strong Pearson correlation up to 0.91 between predicted and observed root mass is achieved, the presented models require calibration by location.