Which Vegetation Index? Benchmarking Multispectral Metrics to Hyperspectral Mixture Models in Diverse Cropland

The monitoring of agronomic parameters like biomass, water stress, and plant health can benefit from synergistic use all available remotely sensed information. Multispectral imagery has been used for this purpose decades, largely with vegetation indices (VIs). Many multispectral VIs exist, typically relying on a single feature—the spectral red edge—for Where hyperspectral is available, mixture models the full VSWIR spectrum to yield further insight, simultaneously estimating area fractions multiple materials within mixed pixels. Here we investigate relationships between by comparing endmember six common in California’s diverse crops soils. In so doing, isolate effects sensor- acquisition-specific variability associated atmosphere, illumination, view geometry. Specifically, compare: (1) fractional photosynthetic (Fv) 64,000,000 3–5 m resolution AVIRIS-ng reflectance spectra; (2) popular (NDVI, NIRv, EVI, EVI2, SR, DVI) computed simulated Planet SuperDove spectra derived spectra. Hyperspectral Fv are compared using both parametric (Pearson correlation, ρ) nonparametric (Mutual Information, MI) metrics. Four (NIRv, DVI, EVI2) showed strong linear (ρ > 0.94; MI 1.2). NIRv DVI interrelation 0.99, 2.4), but deviated 1:1 correspondence Fv. EVI EVI2 were strongly interrelated 2.3) more closely approximated relationship contrast, NDVI SR weaker, nonlinear, heteroskedastic relation < 0.84, = 0.69). exhibited especially severe sensitivity unvegetated background (–0.05 +0.6) saturation (0.2 0.8 0.7). self-consistent atmospheric correction, radiometry, sun-sensor geometry allows simulation approach be applied indices, sensors, landscapes worldwide.