The Effect of Spatial Scale on the Prediction of Tropical Forest Attributes from Image Texture

Table 2

Best-supported models describing each forest attribute as a function of one or more image metrics using different multispectral bands and window sizes.

Forest attribute	Model
5-pixel window
AGB	4.482 + 0.622⋅poly (EVI variances) 1 − 3.432⋅poly (EVI variances) 2
BA	3.22 − 0.20⋅EVI entropyt + 0.22 R meant
D	7.347 − 0.70 poly (EVI variances) 1 − 1.26 poly (EVI variances) 2
H	1.84 − 0.05 R means + 0.08 NIR entropyt
HSD	0.6 + 1.96 poly (EVI entropyt) 1 − 1.784⋅poly (EVI entropyt) 2
9-pixel window
AGB	4.59 + 0.12 EVI contrast
BA	3.37 + 0.12 NDVI entropyt
D	7.41 − 0.15 NIR entropyt + 0.08 NDVI data.ranges
H	1.79 + 0.07 EVI meant − 0.05 EVI homogeneityt
HSD	0.71 + 0.09 NIR correlationt − 0.10 EVI homogeneityt
21-pixel window
AGB	4.64 + 0.40 poly (R means) 1 + 0.65 poly (R means) 2
BA	3.40 − 0.11 NIR skewnesss − 0.20 NIR contrastt − 0.28 NIR ASMt
D	7.42 + 0.61 poly (NIR homogeneityt) 1 − 0.36 poly (NIR homogeneityt) 2
H	1.78 − 0.36 poly (NIR homogeneityt) 1 + 0.06 poly (NIR homogeneityt) 2
HSD	0.78 + 0.29 poly (R data.ranges) 1 − 0.64 poly (R data.ranges) 2

Forest attributes: basal area (BA), density (D), aboveground biomass (AGB), standard deviation of height (HSD), and mean height (H). Bands: red (R), near-infrared (NIR), normalised difference vegetation index (NDVI), and enhanced vegetation index (EVI). poly()1 and poly()2 are the first and second terms, respectively, of a quadratic orthogonal polynomial as produced by the poly() function in R. Image metric abbreviations are described in Table S1.