Table 7.
Statistical indicators for the goodness of fit of mature root system prediction via direct inference from seedling screens (cf. Fig. 5) and using root architecture simulation (cf. Figs 7 and 8)
RMSE is root mean square error; correlation gives the Spearman rank correlation coefficient with P-values; slope=1:1 and intercept=0 indicate if slope and intercept parameters of a linear regression are significantly (P-values) different from the 1:1 line and from zero (no intercept) respectively.
| Indicators | Direct comparison | Model scenario 1 | Model scenario 2 | Model scenario 3 |
|---|---|---|---|---|
| Root length | ||||
| RMSE (cm) | 565.6 | 591.8 | 839.2 | 614.8 |
| Mean error (%) | 99.0 | 19.3 | 29.8 | 23.8 |
| Index of agreement (–) | 0.41 | 0.94 | 0.51 | 0.95 |
| Correlation (RSpearman; P-value) | 0.42(P=0.099) | 0.83 (P<0.001) |
0.70 (P=0.002) |
0.85 (P<0.001) |
| Slope=1:1 (P-value) | <0.001 | 0.358 | 0.815 | 0.090 |
| Intercept=0 (P-value) | 0.123 | 0.115 | 0.507 | 0.455 |
| First-order lateral root number | ||||
| RMSE (n) | 5.07 | 18.5 | 24.4 | 9.9 |
| Mean error (%) | 75.2 | 26.2 | 25.4 | 11.2 |
| Index of agreement (–) | 0.34 | 0.79 | 0.64 | 0.94 |
| Correlation (RSpearman; P-value) | 0.18 (P=0.497) |
0.52 (P=0.038) |
0.35 (P=0.178) |
0.85 (P<0.001) |
| Slope=1:1 (P-value) | <0.001 | 0.002 | 0.028 | 0.555 |
| Intercept=0 (P-value) | 0.059 | 0.004 | 0.081 | 0.137 |
| Root depth distribution | ||||
| RMSE (–) | – | 0.18 | 0.18 | 0.10 |
| Mean error (%) | – | 71.2 | 73.8 | 38.5 |
| Index of agreement (–) | – | 0.52 | 0.47 | 0.92 |
| Correlation (RSpearman; P-value) | – | 0.25 (P=0.025) |
0.20 (P=0.084) |
0.74 (P<0.001) |
| Slope=1:1 (P-value) | – | <0.001 | <0.001 | 0.001 |
| Intercept=0 (P-value) | – | 0.005 | 0.002 | 0.233 |