Table 4.

Random forest confusion matrix for date 2 of data collection (August 1)

	Reference data
	1 (%)	2 (%)	3 (%)	4 (%)	5 (%)
Predicted data
1	85.1	14.9	0	0	0
2	10.4	79.6	9.3	0.5	0.2
3	0.2	13	74	12	0.8
4	0	0.6	9.9	73.6	15.9
5	0	0	3.1	18.4	78.5
Overall accuracy (%) = 77

The % green, % yellow, and % brown pixels from each of the research plots were used as features in a random forest model. This confusion matrix shows how well the random forest model predicted the iron deficiency chlorosis (IDC) score from ground-based reference data where each plot was rated on a one through five scale. The overall accuracy was 77%

The diagonal elements are italicized to highlight the percentage of correctly classified field plots in terms of IDC score