Linking common multispectral plant indices to models of hyperspectral mixtures

Bivariate distributions of spectral indices versus vegetation fraction. The 6 commonly used multispectral indices (y-axis) are compared with the photosynthetic vegetation fraction (Fv) calculated directly from AVIRIS-ng hyperspectral reflections (X-axis). DVI and NIRv are closely related to each other (0.99) and with Fv (0.95), but not near the 1:1 line (red). Similarly, EVI and EVI2 are closely related to each other (0.99) and with Fv (0.94 or 0.95). NDVI and SR show a significantly lower association with Fv (0.84 and 0.81). Mutual information (MI) generally agrees with these correlations. The MI values ​​(relative to Fv) for DVI, NIRv, EVI, and EVI2 are all 1.35 +/- 0.1. NDVI and SR MI values ​​are lower, each at 0.69. SR* values ​​are scaled as 0.1, spectra with values ​​of 1.2 are excluded.

Astro-girl

For decades, agricultural engineers have used remote sensing to monitor key parameters of crops such as biomass, partial cover and plant health. Vegetation indices (VIs) are popular for this purpose, primarily making use of the spectral red edge in multispectral images.

In contrast, spectroscopic mixture models use the full reflection spectrum to simultaneously estimate region fractions from multilateral materials contained within a mixed pixel.

Here, we characterize the relationships between hyperspectral member fractions and 6 common multispectral fibres in crops and soils in California cultivation. The partial area of ​​green vegetation (Fv) was estimated directly from 564,000,000 nm, 3–5m reflectance spectra assembled from a mosaic of 15 AVIRIS-ng flylines. The simulated Planet SuperDove reflection spectra were then derived from AVIRIS-ng, and used to calculate 6 common VIs (NDVI, NIRv, EVI, EVI2, SR and DVI). Multispectral VIs were compared with hyperspectral Fv using parametric (Pearson correlation, r) and non-parametric (mutual information, MI) similarity measures. The 4 VIs (NIRv, DVI, EVI, EVI2) showed strong linear relationships with Fv (r>0.94; MI>1.2). NIRv & DVI showed a strong correlation (r > 0.99, MI > 2.4), but deviated significantly from 1:1 for Fv. Both EVI and EVI2 were strongly correlated (r>0.99, MI>2.3) and closely followed a 1:1 relationship with Fv.

In contrast, NDVI & SR showed a weaker, nonlinear, and heterogeneous relationship with Fv (r < 0.84, MI = 0.69). NDVI showed a particularly high sensitivity to substrate background reflection (-0.05

Correlating common multispectral vegetation indices with hyperspectral mixture models: results from 5 nm and 3 m airborne imaging spectroscopy in a diverse agricultural landscape.

Daniel Souza and Christopher Small

Comments: 18 pages
Topics: geophysics (physics.geo-ph); Earth and Planetary Astrophysics (astro-ph.EP)
Cited as follows: arXiv: 2208.06480 [physics.geo-ph] (or arXiv: 2208.06480v1 [physics.geo-ph] for this version)
Submission date
WHO: Daniel Souza
[v1] Friday, August 12, 2022, 19:55:48 UTC (5,274 KB)
Whole sheet: https://arxiv.org/abs/2208.06480
astrobiology,