Earth SAT Data
Home › Technical Guides › Spectral Analytics
Spectral Analytics

Multispectral Vegetation Indices: NDVI, EVI, NDRE, and NDWI

Translate calibrated Red, NIR, Red-Edge, and SWIR reflectance bands into reliable crop vigor, chlorophyll stress, and moisture metrics.

Why Surface Reflectance Matters Before Computing Indices

Healthy leaves absorb visible red light (~665 nm) for photosynthesis and strongly scatter near-infrared (~842 nm) energy inside mesophyll cells. Spectral indices exploit this contrast to track canopy biomass and vigor.

Raw Digital Numbers or Top-of-Atmosphere (TOA) values shift with aerosol haze and solar elevation. Always compute indices from Bottom-of-Atmosphere Surface Reflectance (Level-2A) scaled to [0, 1] so time series capture plant physiology rather than weather.

  • Use Level-2A surface reflectance for multi-date trend comparisons.
  • Apply QA cloud and shadow masks before computing field statistics.
  • Check band center wavelengths when fusing Sentinel-2, PlanetScope, and Landsat.

Diagnosing NDVI Saturation and Switching to EVI

NDVI ((NIR - Red) / (NIR + Red)) cleanly separates bare soil from emerging crops. Once Leaf Area Index exceeds 3.0 in dense corn or forest canopies, however, Red reflectance bottoms near 0.03 and NDVI plateaus around 0.85–0.92.

The Enhanced Vegetation Index (EVI) adds a Blue band aerosol correction and background coefficients (G = 2.5, C1 = 6.0, C2 = 7.5, L = 1.0), keeping sensitivity alive through peak canopy closure.

  • NDVI > 0.65 indicates dense canopy; < 0.25 indicates bare soil or sparse cover.
  • EVI suppresses atmospheric haze via the Blue (~490 nm) band.
  • SAVI adjusts for bright soil background in early-season row crops.

Red-Edge Chlorophyll (NDRE) and SWIR Moisture (NDMI)

Visible red light saturates in top leaves, hiding early nitrogen stress. Red-Edge bands (705–740 nm) penetrate deeper, enabling NDRE ((NIR - RedEdge) / (NIR + RedEdge)) to flag chlorophyll drops days earlier than NDVI.

Shortwave Infrared (~1610 nm) bands power NDMI ((NIR - SWIR) / (NIR + SWIR)) to detect canopy moisture deficits and drought stress before visible wilting.

Operational Workflow for Field-Scale Analytics

Agronomy pipelines query STAC catalogs weekly for cloud-free scenes, mask QA layers, compute zonal statistics inside field polygons buffered inward by one pixel, and trigger alerts when zones fall below historical baselines.

Key Multispectral Indices, Formulas, and Operational Roles

IndexStandard Formula (Surface Reflectance)Required Spectral BandsPrimary Analytical Strength
NDVI(NIR - Red) / (NIR + Red)Red (~665 nm), NIR (~842 nm)Baseline greenness and vegetation cover
EVI2.5 * (NIR - Red) / (NIR + 6*Red - 7.5*Blue + 1)Blue, Red, NIRResists saturation in dense canopy & haze
NDRE(NIR - RedEdge) / (NIR + RedEdge)Red-Edge (~705 nm), NIRMid-to-late season chlorophyll & nitrogen stress
SAVI((NIR - Red) / (NIR + Red + 0.5)) * 1.5Red, NIREarly growth stages with exposed soil
NDMI(NIR - SWIR1) / (NIR + SWIR1)NIR, SWIR (~1610 nm)Canopy water content and drought monitoring

Formulas assume surface reflectance scaled to [0, 1]; apply sensor scale factors and additive offsets before computing.

Worked Illustration: Computing NDVI and EVI from Surface Reflectance

A mid-season maize parcel reports 10 m Level-2A reflectance values of Blue = 0.030, Red = 0.035, and NIR = 0.480. We calculate NDVI and EVI.

  1. Compute NDVI difference and sum: 0.480 - 0.035 = 0.445; 0.480 + 0.035 = 0.515.
  2. Compute NDVI: 0.445 / 0.515 = 0.864 (approaching dense-canopy saturation).
  3. Compute EVI denominator: 0.480 + (6 x 0.035) - (7.5 x 0.030) + 1 = 1.465.
  4. Compute EVI: 2.5 x (0.480 - 0.035) / 1.465 = 1.1125 / 1.465 = 0.759.

Calculated Result & Takeaway: NDVI sits at 0.864 (near saturation), whereas EVI is 0.759, preserving headroom to track late-season vigor differences.

Spectral Vegetation Pipeline Quality Checklist

  • Verify Level-2A Bottom-of-Atmosphere Surface Reflectance inputs.
  • Apply sensor scale factors and radiometric offsets before band math.
  • Mask clouds, cirrus, and cloud shadows with the scene QA/SCL layer.
  • Switch to EVI or NDRE once NDVI exceeds 0.75 in closed canopies.
  • Harmonize bandpasses before blending Sentinel-2 and Landsat series.
  • Buffer field boundaries inward by 1 pixel to remove road edge noise.

Frequently Asked Questions

Why can NDVI drop even when crops are healthy?

Thin cirrus haze, wet dark soil after irrigation, or off-nadir viewing shifts can depress NDVI unless scenes are cloud-masked and atmospherically corrected.

Do all commercial satellites include a Red-Edge band?

No. Standard 4-band RGB+NIR payloads omit Red-Edge; select 8-band commercial sensors or Sentinel-2 when NDRE is required.

What is a typical NDVI range for healthy mid-season row crops?

Closed-canopy row crops typically measure 0.70 to 0.88 in surface-reflectance NDVI depending on crop structure.

How does Sentinel-2 baseline 04.00+ radiometric offset affect NDVI?

Baseline 04.00+ adds a +0.1 reflectance offset (+1000 DN); failing to subtract 0.1 artificially lowers dense vegetation index values.

Primary References & Standards

  • USGS Landsat Surface Reflectance-Derived Spectral Indices Product Guide.
  • ESA Copernicus Sentinel-2 MSI Level-2A Algorithm Theoretical Basis Document.
  • Huete, A. et al. (2002) Performance of MODIS vegetation indices (EVI formulation).

Related Technical Guides