Image-based high-throughput field phenotyping of crop roots

Abstract

Current plant phenotyping technologies to characterize agriculturally relevant traits have been primarily developed for use in laboratory and/or greenhouse conditions. In the case of root architectural traits, this limits phenotyping efforts, largely, to young plants grown in specialized containers and growth media. Hence, novel approaches are required to characterize mature root systems of older plants grown under actual soil conditions in the field. Imaging methods able to address the challenges associated with characterizing mature root systems are rare due, in part, to the greater complexity of mature root systems, including the larger size, overlap, and diversity of root components. Our imaging solution combines a field-imaging protocol and algorithmic approach to analyze mature root systems grown in the field. Via two case studies, we demonstrate how image analysis can be utilized to estimate localized root traits that reliably capture heritable architectural diversity as well as environmentally induced architectural variation of both monocot and dicot plants. In the first study, we show that our algorithms and traits (including 13 novel traits inaccessible to manual estimation) can differentiate nine maize (Zea mays) genotypes 8 weeks after planting. The second study focuses on a diversity panel of 188 cowpea (Vigna unguiculata) genotypes to identify which traits are sufficient to differentiate genotypes even when comparing plants whose harvesting date differs up to 14 d. Overall, we find that automatically derived traits can increase both the speed and reproducibility of the trait estimation pipeline under field conditions.

Figure 1.

A, Classic shovelomics scoring board to score the angle of maize roots with the soil tissue. B, An example to score rooting depth and angle in common bean.

Figure 2.

I, Imaging board on the example of a maize root. The experiment tag is used to capture an experiment number, and the scale marker allows the correction of camera tilting and transforming image coordinates into metric units. II, Camera mounted on a tripod placed on top of the imaging board coated with blackboard paint. Note that images were taken with protection against direct sunlight not shown in the image. III, Example of the segmentation of the original image into a binary image and then into a series of image masks that serve as input to estimate traits for monocot and dicot roots. The sample is that of a maize root, 40 d after planting at the URBC. IV, The imaging pipeline for dicot roots and sparse monocot roots: Original image on the imaging board (a), derived distance map where the lighter gray level represents a larger diameter of the imaged object (b), medial axis includes loops (c), and loop RTP with a sample of the root branching structure (d). Colors are randomly assigned to each path. The sample is that of a cowpea root, approximately 30 d after planting at the URBC.

Figure 3.

Trait correlations in cowpea. A, Results from field measurements of taproot diameters at 5- and 10-cm depth show correlations close to r = 0.5 for Spearman and Pearson coefficients. Note that not all plants reached 10-cm depth. B, An example of confirming results with r = 0.5 for Spearman and Pearson coefficients was achieved in correlation between the field-measured 5-cm depth level and the diameters at 75% of the central path length. C, The central path diameters extracted from images are shown with similar correlation coefficients than observed with manual field measurements. D, A surprising correlation: The number of third order basal roots in cowpea is highly correlated with the overall number of RTPs. Both correlation measures together (Spearman r and Pearson r) suggest even linear dependency. The ANOVA analysis confirms that both measurements are taken from different quantities for both correlations. All diameter measurements are in millimeters and data points represent averages per genotype.

Figure 4.

Examples of maize genotypes selection from the Wisconsin diversity panel. The figure shows increasing root system width and shovelomics angle from left to right. The top row shows the complete root crown, and the bottom row shows the root crown with brace roots removed to reveal the crown roots. All six images show a selected representative nodal root at the right of the root crown.

Figure 5.

RPV analysis of the crown root measurements. Traits are more likely to be useful in differentiating genotypes when their RPV is significantly greater than 1 (blue line). Trait definitions are found in Tables I and II. TD, Tip diameter.

Figure 6.

Top, Image-based phenotype differentiation. Normalized mean trait values of traits derived from crown root images. The intergenotype variation for the crown roots of the nine examined maize genotypes is shown. The points represent average normalized values. The connection between points allows the reader to visually identify the three genotypes shown in Figure 4. The error bars indicate the sem. Note that relative traits only differentiate at certain depth levels. Bottom, The number of traits that distinguish a pair of genotypes in the maize study. For each combination, at least eight distinguishing traits were found. DD90max, Maximum diameter at 90% to 100% depth; TD, tip diameter.

Figure 7.

Top, Phenotype differentiation with shovelomics. Normalized mean trait values show the intergenotype variation for the crown roots of the nine examined maize genotypes. The points represent average normalized values. The connection between points allows the reader to visually identify the three genotypes shown in Figure 4. All density and distance measures are researcher scores. Bottom, The number of traits that distinguish a pair of genotypes in the maize study. For each combination, at least one distinguishing trait was found.

Figure 8.

Six examples from the cowpea diversity panel. Different root architectures in mature cowpea with the genotype denoted in each image are illustrated.

Figure 9.

RPVs of shovelomics and image-based traits for the cowpea diversity panel. Traits are more likely to be useful in differentiating genotypes when their RPV is significantly greater than 1. Note that the manual traits counting first order laterals do not include counts in the basal region of the root. All field scores ranged from 0 (low) to 9 (high). The figure is comparable to Supplemental Fig. S30 showing RPVs for GIA Root traits.

Figure 10.

Overall phenotype differentiation of the cowpea diversity panel. Black crosses show the trait values of the whole data set, and the lines are three selected examples with error bars denoting the sem. The three example genotypes distinguish by at least one D and DS value and by the second dominant angle. The D values demonstrate that UCR 779 differentiates from Petite n Green at higher depth levels and from Early Scarlet at deeper depth levels. UCR 779 differentiates additionally in central path diameters from Petite n Green and Early Scarlet. The connection between points allows the reader to visually identify the three of the six genotypes shown in Figure 8. DD90max, Maximum diameter at 90% to 100% depth; dia., diameter; TD, tip diameter; CPD, central path diameter.

Figure 11.

Image masks overlaid with the width profile (red) and the corresponding cumulative width and slope function from which the D and DS values are calculated. Top, Cowpea root. Bottom, Maize root. Note that the width profile does not have to match the outline of the root, because the outline is not always symmetric to a vertical and straight line. It is also visible in the cumulative curve that the D values are robust to single roots sticking out of the crown.

Figure 12.

Principle of the RTP algorithm as pseudocode.

Figure 13.

Difference between the RTA (green) and the STA (red). The RTA does not change if the root is rotated on the board. By contrast, the STA on the rotated root on the right is larger than on the left.