NDVI vs EVI for a drought index insurance anomaly: which one to lock in

Pick an index before you pick a season. That's the practical rule once you've sat through a basis dispute where the grower's agronomist is looking at one vegetation index and your pricing model was built on another. NDVI and EVI both describe how green a region is relative to bare soil, but they get there differently enough that swapping one for the other mid-contract will shift your anomaly numbers in ways that are hard to explain to a cedant.

What NDVI and EVI actually measure

NDVI (Normalized Difference Vegetation Index) is the ratio of near-infrared to red reflectance. It's been the default since the Landsat era, which means most of the multi-decade archives you'd want for a baseline are built on it. Its weakness is well known in the trade: it saturates in dense canopy. Once a region's biomass gets thick, say a fully grown maize stand or closed-canopy forest, NDVI flattens out and stops distinguishing "healthy" from "very healthy." For drought work that mostly matters at the shoulders, early season stress and late-season recovery, where saturation isn't usually the problem.

EVI (Enhanced Vegetation Index) adds a blue band and a correction term to reduce canopy-background noise and residual atmospheric scatter that NDVI doesn't fully cancel out. It holds up better in high-biomass regions and over soil types with strong background reflectance, which is why agronomic researchers often prefer it for row crops like corn and soy at peak growth. The tradeoff is that EVI needs a clean blue band, and blue is the most atmosphere-sensitive channel there is. In hazy or smoke-affected periods, an EVI series can carry more noise than an NDVI series over the same pixels.

Why the choice matters more for pricing than for monitoring

If you're just eyeballing greenness for a field visit, either index tells a similar story most months. Pricing a trigger is different. You're computing an anomaly: this month's value against a baseline distribution built from years of history for that same region. If your baseline was built on NDVI and your live feed switches to EVI, or blends the two, the anomaly isn't measuring drought anymore. It's measuring the difference between two indices with different saturation curves and different sensitivity to atmospheric correction. That shows up as a trigger that fires when reality didn't warrant it, or stays quiet when it should have fired, and no one can tell which effect is the index switch and which is the weather.

The practical answer most desks land on: NDVI for anomaly work unless the insured crop or biome has a documented saturation problem at the growth stages that matter for the trigger window. Tall-canopy, high-biomass row crops near peak growth are the clearest case for EVI. Grassland, rangeland, and most dryland cereal exposures don't hit NDVI's saturation ceiling often enough to justify giving up NDVI's longer, cleaner archive. Whichever you pick, the baseline and the live series need to be computed with the same index, the same compositing method, and ideally the same sensor lineage, or the anomaly number isn't comparable to itself across years.

Keeping the index fixed once you've filed it

Once an index is written into the contract methodology, treat it as a constant, not a parameter you revisit each renewal. Changing it mid-program means rebuilding the baseline distribution and re-stating historical anomalies, which is a conversation reinsurers and regulators will want documented well before claims season. If you're still deciding, run both series over a few past drought years for the specific region and growth window in the contract and see where they diverge. The years they agree on tell you the index choice doesn't matter much there. The years they disagree on tell you exactly what you're trading off.

Drought Monitoring runs the anomaly against a fixed, documented baseline so the index choice stays consistent from one month's reading to the next. If you're scoping a new trigger and want to see how a region's anomaly reads under that baseline, get in touch.