Table 5.

Statistical results from cross-validation comparing USDA pesticide use estimates to the estimates generated in this study from USGS pesticide use data and USDA crop area data.

Category	Comparison	N (# state-crop-year-AI)	Relative difference (%):				Spearman’s rho	Pearson’s r
			Median	25th percentile	75th percentile	Wilcoxon test P
Fungicides	kg applied	501	4	−51	54	0.59	0.79	0.86
	kg/ha (M1)	501	4	−51	54	0.59	0.71	0.95
	kg/ha (M2)	502	2	−56	56	0.99	0.71	0.94
Herbicides	kg applied	6762	2	−38	40	0.45	0.94	0.97
	kg/ha (M1)	6762	2	−38	40	0.45	0.94	0.95
	kg/ha (M2)	6891	3	−39	42	0.04	0.94	0.94
Insecticides	kg applied	1699	−14	−69	43	<0.01	0.85	0.20
	kg/ha (M1)	1699	−14	−69	43	<0.01	0.85	0.36
	kg/ha (M2)	1712	−10	−65	47	<0.01	0.85	0.38
Insecticides (no outliers)*	kg applied	1669	−12	−66	45	<0.01	0.87	0.84
	kg/ha (M1)	1669	−12	−66	45	<0.01	0.86	0.77
	kg/ha (M2)	1682	−8	−62	48	<0.01	0.87	0.78
All active ingredients	kg applied	8962	−0.1	−45	41	0.01	0.92	0.93
	kg/ha (M1)	8962	−0.1	−45	41	0.01	0.92	0.83
	kg/ha (M2)	9105	1	−45	44	0.30	0.92	0.83

For application rate, Method 1 (M1) compared our estimate to one calculated by dividing the USDA total kg by our estimate of crop area (ha), and Method 2 (M2) compared our estimate to one calculated by multiplying the USDA average application rate on treated hectares by the percent of area treated. Relative difference = [(USGS – USDA)/average of the two values] × 100.

*Recalculated after removing malathion use in cotton.