Efficacy of Multiseason Sentinel-2 Imagery for Classifying and Mapping Grassland Condition
Table 4
Top 20 variables with the highest number of confirmations by the Boruta algorithm for scenario D1D2R_BI at 10 and 20 m resolutions for 40 RF models.
D1D2R_BI (10 m)
D1D2R_BI (20 m)
Variable
Mean number of confirmations (0–40)
Mean Boruta importance
Variable
Mean number of confirmations (0–40)
Mean Boruta importance
ndvi.2
40
12.2
evi2.2
24
7.29
evi2.2
40
12.1
ndvi.2
24
7.13
ndvi.6
40
9.3
evi2.5
24
6.79
evi2.6
40
9.3
ndvi.5
24
6.75
ndvi.1
40
9
cire.2
24
6.62
evi2.1
40
9
ndre.2
24
6.6
evi2.5
40
8.5
red_evi2.2
24
6.55
ndvi.5
40
8.5
SWIR11.2
24
6.23
B.3
40
7.1
R.5
24
5.69
B.2
40
6.8
ndwi.2
23
5.59
G.3
40
6
cire.1
24
5.36
R.5
40
5.8
ndre.1
24
5.35
R.3
40
5.5
red_evi2.1
24
5.35
G.4
40
5.3
red_evi2.5
24
5.15
B.1
40
5.1
cire.5
24
5.13
cire.2
40
5
ndre.5
24
5.09
red_evi2.2
40
4.9
SWIR11.5
23
4.87
ndre.2
40
4.9
RE5.5
20
4.81
R.2
40
4.9
cire.6
20
4.33
evi2.3
40
4.8
red_evi2.6
21
4.33
ndvi.3
40
4.8
ndre.6
21
4.32
Number of confirmations range from 0 to 40. Each variable presents a mean boruta importance value (n = 40). Numbers in variables correspond to the image date. 2 Jan = 1, 7 March = 2, 15 Jul = 3, 14 Aug = 4, 17 Nov = 5, and 27 Nov = 6. D1, early dry season; D2, late dry season; R, rainy season; B, spectral bands; and I, vegetation indices.