How we measure crops

Every visible crop pixel reflects light differently when it's healthy than when it's stressed. Four indices we compute every week pull that read apart:

• NDVI — overall greenness. The classic "is it growing?" number. High NDVI through July is what every July reads are looking for. Falls fast under drought, hail, or hard freeze.
• EVI — canopy closure. Like NDVI but more sensitive once the field is fully covered. Better than NDVI for separating "great" from "OK" in mid-summer when both look green from the road.
• NDWI — leaf water. Drops during drought before NDVI does — an early warning for stress that's about to bite yield.
• NDMI — soil and canopy moisture. Pairs with NDWI to separate "the plant is thirsty" from "the soil is dry."

Default views show NDVI because it's the most widely understood. Pro+ users can switch indices on the dashboard.

A practical note on NDVI: it saturates once the canopy fully closes (around LAI 3, typically corn V12–VT). Past that point, NDVI plateaus near 0.85 even as the canopy keeps building. EVI doesn't saturate as quickly, so it stays informative through the corn VT–R3 window when management decisions matter most. NDWI usually drops a few days before NDVI under drought stress — early warning for pollination and pod-set windows. For a stage-by-stage read of how the four indices map to corn and soy phenology, see corn and soybean growth stages.

Sentinel-2 satellites pass over every spot in the Corn Belt about every five days. A single pass can be cloudy or partial. To get a clean read, we aggregate every usable pixel inside a calendar week, mask to the actual crop using USDA's Cropland Data Layer, and publish one number per county per week.

New data lands every Monday morning. If clouds were heavy across a state for an entire week, that state's read for that week may be slim — we surface a confidence score so you can tell.

When a county misses a clean read, we don't leave a gap. A weather-conditioned model interpolates the missing weekly value from surrounding clear-sky observations combined with that county's temperature, precipitation, and humidity for the same week. The interpolated rows are flagged as such, and the gap-fill model is benchmarked against held-out clear-sky observations every season. The short version: you get a continuous curve, but you can always tell which weeks were observed and which were filled.

If you need real-time, we're not the right tool. We're the right tool for the weekly check-in that has to happen anyway.

Every number we publish traces back to a public agency. We don't repackage somebody else's product — the pipeline reads originals.

Copernicus Sentinel-2 (ESA)

The optical imagery the indices come from. NDVI, EVI, and NDWI compute on the 10-meter visible and near-infrared bands; NDMI uses the 20-meter shortwave-infrared band alongside NIR. Two satellites in orbit since mid-2017, ~5-day revisit; our archive starts 2017. Free under the Copernicus Open Data Policy.

USDA Cropland Data Layer (CDL)

30-meter annual raster that tells us which pixels were actually planted to corn or soy. Without it we'd be averaging in pasture, woods, and roads. CDL lags by one growing season — a 2025-season read uses the 2024 CDL.

USDA NASS Quick Stats

County-level historical yields (the training target for our yield model) and weekly Crop Progress.

IEMRE — Iowa Environmental Mesonet

Daily ~14 km gridded weather (temperature, precipitation, GDD, VPD) reanalyzed by Iowa State. Drives the weather context next to every reading.

A raw NDVI of 0.74 in Iowa in early July is meaningless on its own. What matters is whether 0.74 is high, low, or normal for that county at that point in the season.

For every county-week combination, we hold a 5-year average — the same calendar week across the previous five years. The deviation we publish is the percentage difference from that baseline. So Iowa, week 28, −3% means corn is reading 3% below what Iowa's typical week-28 corn looks like.

This baseline rolls forward — we don't anchor to one good year and never update. As new seasons come in, the oldest year drops out.

Every weekly reading carries a confidence score. It's a function of three things: how many usable pixels were captured (cloud cover knocks this down), how late in the growing season we are, and how stable the underlying read has been week-over-week.

• High — clear pass, full canopy, multiple consistent reads. Trust the number.
• Medium — partial coverage or shoulder-season. Useful as direction, less useful as a precise level.
• Low — heavy clouds, very early or very late in the year, or sparse pixel count. We still show it; treat it as a hint, not a fact.

Regional commentary aggregates the underlying state-level reads — NDVI, drought, weather anomalies — across the member states below. WASDE balance-sheet figures are national; we report them once and surface them in every region's read.

Western Corn Belt

IA · MN · NE · SD · ND

Eastern Corn Belt

IL · IN · OH · MI · WI

Northern Plains

MT · WY · CO

Southern Plains

KS · MO · OK · TX

Delta

AR · MS · LA

Southeast

GA · AL · NC · SC · TN · KY

Regional commentary aggregates these 26 corn-belt states. Field-level crop-health and weather coverage extend across all 48 contiguous states wherever corn or soy pixels appear in the latest USDA CDL.

• Not investment advice. Nothing here is a buy/sell recommendation, hedge instruction, or risk-management directive. Traders use us as one input among many.
• Not real-time. Sentinel-2 has a ~5-day revisit and we publish weekly. If a hailstorm hit yesterday, we'll see it next Monday, not this afternoon.
• Not field-level on the free tier. Free shows one rotating state and four rotating counties. County-level coverage with no rotation requires Pro; whole-state coverage requires Pro+.
• Not a yield guarantee. Yield forecasts (Pro+) are a model output with a confidence band, not a promise. We've trained on 2017–2024 USDA actuals across 32 corn-growing states (29 for soy); the band reflects how tightly the model has fit history.
• Not raw imagery. We don't sell pixels or tiles. If you need rasters, go to the Copernicus Open Hub directly.

If any of this changes, this page changes with it. See the full disclaimer for the legal version.