Through further investigation, we found that the SVVI is highly correlated
to the normalized difference infrared index (NDII) utilizing
ETM+ bands 5 and 4 (Yilmaz et al., 2008). We analyzed a humid-area
subset and a more arid-area subset within a single Landsat image and
found correlation values between SVVI and NDII to be 94% and 83% for
the humid and arid areas, respectively. As with NDII, SVVI generally increase
along a moisture gradient, which in our study area is from dense
forest, to disturbed/secondary forest, to shrub, to non-photosynthetic
vegetation (e.g., dry savanna vegetation), to bare soil. However, compared
to the NDII, the SVVI provides much greater differentiation of agriculture
fields, as well as greater differentiation between agricultural
fields and adjacent natural vegetation (particularly in drier environments).
In the wet-dry savanna portions of the study area, agricultural
plots and fields not apparent on the NDII product are readily apparent
on the SVVI product (Fig. 4b and c). In addition to the benefits of the
SVVI relative to the NDII described above, we also found that radiometric
differences in NDII products yielded artifacts during the multi-temporal
compositing process, while the SVVI composites did not (Fig. 4b
and c).