If you monitor crops from satellite data, you have almost certainly used NDVI, the Normalized Difference Vegetation Index. It is the default: simple, available from nearly every multispectral sensor, and backed by an archive going back to the early 1980s. But NDVI has two well-known weaknesses. It saturates over dense, healthy canopy, so a thriving forest and a peak-season crop can return nearly identical values, and it is sensitive to atmospheric haze and bare soil showing through sparse cover. The Enhanced Vegetation Index, EVI, was designed to fix both problems.

Both indices exploit the same physics. Healthy vegetation absorbs red light for photosynthesis and reflects near-infrared light strongly, so the difference between the two bands is a proxy for how green and vigorous a pixel is. NDVI expresses that difference as a simple ratio: near-infrared minus red, divided by near-infrared plus red, giving values from minus one to one, with healthy crops typically reading 0.6 to 0.9. EVI keeps the same core idea but adds the blue band and three fixed coefficients to correct for atmospheric scattering and soil background, which keeps it responsive where NDVI flattens out.

The formulas and what they mean in practice

The NDVI formula is (NIR minus Red) divided by (NIR plus Red), and it needs only two bands, which is why it works on almost any sensor, including older Landsat data and basic drone cameras. EVI uses the formula 2.5 times (NIR minus Red) divided by (NIR plus 6 times Red minus 7.5 times Blue plus 1), with the coefficients fixed by NASA for the MODIS EVI product.

On a real scene the difference is measurable. In published comparisons on Sentinel-2 imagery, dense forest and peak-season crops both returned NDVI values around 0.87 to 0.88, essentially indistinguishable, while EVI separated them clearly. That is the saturation problem in numbers: once a canopy is dense enough, adding more leaf area barely changes the red and near-infrared reflectance, so NDVI stops responding. EVI's atmospheric correction also matters in hazy conditions or over regions with frequent aerosol cover, where NDVI can drift from one date to the next for reasons that have nothing to do with the crop.

When to use NDVI and when to use EVI

Use NDVI when you need comparability and simplicity. It is the right choice for long time series, because the Landsat archive reaches back to 1982 and NDVI can be calculated consistently across decades and sensors. It is also the right choice for general vegetation mapping, drought monitoring and yield estimation at field scale, where the canopy is not so dense that saturation becomes a problem. If your audience expects NDVI, or your workflow compares against published NDVI thresholds, stay with it.

Use EVI when the canopy is dense or the atmosphere is a problem. Forestry, plantation health, tropical agriculture and any study of high-biomass vegetation are EVI territory, because that is where NDVI saturates. EVI is also the better choice in hazy regions or when comparing images from dates with different atmospheric conditions. The trade-off is that EVI needs a blue band, which some sensors and drone cameras lack, and its values are not directly comparable with the NDVI literature, so you need to establish your own interpretation thresholds, with healthy EVI typically reading 0.2 to 0.8 and values below 0.15 flagging low vigour.

Calculating both in QGIS

Both indices run in the QGIS Raster Calculator on surface reflectance data, and Sentinel-2 bands make the mapping straightforward: B08 is near-infrared, B04 is red, and B02 is blue. For NDVI the expression is (B08 minus B04) divided by (B08 plus B04). For EVI it is 2.5 times (B08 minus B04) divided by (B08 plus 6 times B04 minus 7.5 times B02 plus 1). Use reflectance values scaled between zero and one, not raw digital numbers, or the results will be wrong. The terra-lab guide to these indices gives ready-to-paste Raster Calculator expressions for Sentinel-2 if you want to copy them directly.

A third option is worth knowing about: SAVI, the Soil-Adjusted Vegetation Index, adds a soil brightness correction and is the better choice when vegetation cover is sparse and bare soil dominates the pixel, such as early-season crops or arid land. And if your goal is detecting crop stress early rather than measuring general greenness, NDRE, which swaps the red band for the red edge band, responds to chlorophyll changes before NDVI or EVI move.

Frequently Asked Questions

Q: What is the main difference between NDVI and EVI?

A: NDVI is a simple ratio of near-infrared and red light. EVI adds the blue band and three correction coefficients, which keeps it sensitive in dense canopy where NDVI saturates and reduces errors from haze and soil background.

Q: Can I calculate EVI from Landsat data?

A: Yes. Landsat 8 and 9 carry the blue, red and near-infrared bands EVI needs. The USGS applies the same MODIS coefficients, so EVI from Landsat is comparable with EVI from MODIS and Sentinel-2.

Q: Which index is better for monitoring wheat or maize fields?

A: NDVI is usually enough for open field crops at moderate density, and it is easier to compare with published thresholds. Switch to EVI if the canopy is very dense, the air is often hazy, or NDVI values stop changing as the crop grows.