Skip to contents

Compute a vegetation index from a multi-band raster

Source: R/compute_vi.R
compute_vi.Rd

Generic engine that computes the main vegetation indices used in nitrogen diagnosis from a multi-band raster::RasterStack or RasterBrick. NFert supports six indices with configurable band mapping, so the function works with Sentinel-2 L2A, UAV multispectral sensors (MicaSense, Parrot Sequoia, Pix4D) or any other multispectral product.

Usage

compute_vi(
  stack,
  index,
  bands = list(red = "B04", red_edge = "B05", nir = "B08", green = "B03"),
  scale_factor = 1,
  clamp = TRUE
)

Arguments

stack

A raster::RasterStack or RasterBrick with the required spectral bands. Band names (or layer indices) are mapped through bands.

index

One of "NDVI", "NDRE", "GNDVI", "CIred", "MCARI", "MSAVI2" (case-insensitive).

bands

Named list or character vector mapping the index inputs (red, red_edge, nir, green) to layer names in stack. Defaults to Sentinel-2 L2A naming (red="B04", red_edge="B05", nir="B08", green="B03"). Only the bands required by the chosen index are looked up.

scale_factor

Numeric. If the input reflectance is stored as integer DN (e.g. Sentinel-2 L2A 0–10000), divide by this factor before computing the index. Default 1 (reflectance already in 0–1).

clamp

Logical. If TRUE (default for bounded indices NDVI, NDRE, GNDVI, MSAVI2), clamp output to the range -1 to 1. Ignored for CIred and MCARI (unbounded).

Value

A raster::RasterLayer with the computed index. Layer name is set to the index name.

Details

Index formulas (references in the References section):

NDVI

(NIR minus Red) divided by (NIR plus Red). Rouse 1974.

NDRE

(NIR minus RedEdge) divided by (NIR plus RedEdge). Gitelson and Merzlyak 1994.

GNDVI

(NIR minus Green) divided by (NIR plus Green). Gitelson 1996.

CIred

NIR divided by RedEdge, minus one. Gitelson 2003.

MCARI

The quantity RedEdge minus Red minus 0.2 times the quantity RedEdge minus Green, all multiplied by the ratio RedEdge over Red. Daughtry 2000.

MSAVI2

2 NIR + 1 minus the square root of (2 NIR + 1)^2 minus 8 times NIR minus Red, all divided by two. Qi 1994.

NDVI saturates at LAI > 3–4 in closed canopies; in mid- to late-vegetative stages (GS30+ for cereals, V8+ for maize) the red-edge indices (NDRE, CIred) are considerably more sensitive to canopy nitrogen status (Clarke 2001; Li 2014; Cao 2015). NFert variable-rate functions (variable_rate_N(), estimate_N_rate_from_calibration_curve(), estimate_N_rate_from_holland_schepers()) all accept any normalised VI on a 0–1 scale, so the user can substitute NDVI with NDRE or GNDVI transparently.

References

Rouse, J.W. et al. (1974). Monitoring vegetation systems in the Great Plains with ERTS. NASA SP-351.

Gitelson, A.A. & Merzlyak, M.N. (1994). Quantitative estimation of chlorophyll-a using reflectance spectra. J. Photochem. Photobiol. B 22.

Gitelson, A.A. et al. (1996). Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ. 58, 289-298.

Gitelson, A.A. et al. (2003). Relationships between leaf chlorophyll content and spectral reflectance. J. Plant Physiol. 160, 271-282.

Daughtry, C.S.T. et al. (2000). Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance. Remote Sens. Environ. 74.

Qi, J. et al. (1994). A modified soil adjusted vegetation index. Remote Sens. Environ. 48, 119-126.

Clarke, T.R. et al. (2001). Remote sensing of nitrogen status in wheat. Proc. Beltwide Cotton Conf.

Li, F. et al. (2014). Improving estimation of summer maize nitrogen status with red edge-based spectral vegetation indices. F. Crops Res. 157.

Cao, Q. et al. (2015). Active canopy sensing of winter wheat nitrogen status: An evaluation of two sensor systems. Comput. Electron. Agric. 112.

Examples

if (FALSE) { # \dontrun{
library(raster)
library(NFert)

# Sentinel-2 L2A stack (integer DN, 0-10000)
s2 <- raster::stack("S2_Cremonesi_20260415.tif")
names(s2) <- c("B03", "B04", "B05", "B08")

ndvi  <- compute_vi(s2, "NDVI",  scale_factor = 10000)
ndre  <- compute_vi(s2, "NDRE",  scale_factor = 10000)
gndvi <- compute_vi(s2, "GNDVI", scale_factor = 10000)

# Use any VI as input to the VRT pipeline
vr <- variable_rate_N(ndre, n_dose = 142, method = "holland",
                      minN = 60, maxN = 180)
} # }