Overall, the crop accuracies achieved in this study met or exceeded those of all previous crop mapping studies. As noted above 70% of the field data collected at the FSA offices was reserved for accuracy assessment. Since ground truth was only collected for crops there was no way to quantify the accuracies of non-crop categories. Accuracy for Winter wheat was a simple comparison of wheat on the output LULC map to areas known to be wheat from ground truth. Accuracies for Winter wheat were, as expected, high (92.4% in County Group 3 to 95.9% in County Group 1). The summer crop accuracy report was in the form of a confusion matrix: a report on how much of each original training area was actually classified as being in the class that the training was meant to represent also showing confusion between categories.

Summer Crops Accuracy Assessment Confusion Matrix: County Group 3

Average accuracy = 88.28%

Overall accuracy = 90.84%

Kappa Coefficient = 0.87552

Confidence Level:

99% 0.87552 +/- 0.004783
95% 0.87552 +/- 0.003633
90% 0.87552 +/- 0.003051

* Totals do not add up to 100% because confusion with non-crop categories was not calculated.

Accuracy numbers for rice were also above 90% mark in all four county groups. Numbers for the sorghum/corn category varied significantly between county groups ranging from 58.2% in Group 4 to 83.5% in Group 1 (Much of the training data was obscured by clouds in Group 1). Soybean accuracies were better than initially expected, ranging from 89.4% in Group 3 (see above) to 72.3% in Group 1. Cotton numbers, once again, ranged from a high of 88.7% in Group 3 to 74.4% in Group 1. Considerable confusion between soybeans and cotton were still evident in the final reports for all county groups. In Groups 1 and 3, confusion between soybeans and sorghum/corn was also a problem.

Crop accuracy numbers for all 27 counties compare favorably with other crop surveys for 1992 as well as with the NASS crop data.