Crop Stress Index API

Why a composite score

Individual agricultural signals each capture a piece of crop risk. Drought severity reflects cumulative moisture deficit. Soil moisture shows what's actually in the root zone. VPD captures heat-driven transpiration stress. Precipitation tells you what fell. NDVI anomaly shows how the vegetation is actually responding to all of it.

Any one of these signals alone can mislead. A county in D2 drought with recently replenished soil moisture is in a very different situation from one that's been dry for 12 consecutive weeks. A wet spring followed by a high-VPD July can be more damaging than moderate drought with mild temperatures.

The Crop Stress Index combines all five signals into a single number — reducing five API calls and a manual blending step to one call that returns the aggregated view. Use the individual signal endpoints when you need to understand what's driving the score; use the index when you need to rank, screen, or alert across a portfolio.

How the index is built

Each of the five component signals is scored on its own scale, then combined using crop-specific weights. The weights were derived by computing Pearson correlations between each signal's growing-season summary and county-level yield deviations from trend, using USDA NASS data for 2001–2024 across roughly 40,000 county-year observations for corn alone. Only signals with a meaningful correlation (|r| ≥ 0.05) are included in each crop's model.

The five components

• Drought — USDM drought category (D0–D4), with bonus points for worsening trajectory over the past 4 weeks and sustained 52-week burden. Drought is the single strongest individual predictor of yield loss across most crops.
• Vapor pressure deficit (VPD) — 30-day average VPD from PRISM, in hPa. High VPD drives crop water stress even when soil moisture is adequate. Particularly important for soybeans during pod fill and corn during pollination.
• Precipitation — 90-day cumulative rainfall from PRISM, with a short-term recency check (7-day vs. 90-day daily average) to penalize counties that received rainfall early but have been dry recently.
• Soil moisture — NASA SMAP anomaly z-score. A z-score of –2.0 means the county is 2 standard deviations drier than its historical baseline for that week of year. Not included in wheat or sorghum models where the correlation with yield was too weak.
• NDVI anomaly — MODIS vegetation index z-score vs. a multi-year baseline. The only signal that captures how plants are actually responding — upstream of yield but downstream of weather. Strong predictor for wheat and cotton.

Auto-selected crop model

The API automatically selects the crop model for a county based on its dominant crop by harvested acreage (5-year NASS average). A McLean County, IL request automatically uses the corn model; a Lubbock County, TX request uses cotton. Counties without NASS data fall back to the GENERAL model. The crop field in every response tells you which model was applied.

Partial scoring and data coverage

When a signal is temporarily unavailable (e.g., NDVI is a 16-day composite with cloud gaps), the index rescales over the available components rather than returning null. The data_coverage_pct field tells you what share of the model's total weight was covered by valid data — a score computed at 60% coverage should be treated with more caution than one at 100%.

Crop models and weights

Component weights (shown below) are proportional to the absolute Pearson correlation between each signal and county yield deviations, rescaled to sum to 100 per crop. A dash (—) means the signal was excluded from that model due to insufficient correlation.

The differences are meaningful: VPD carries 26 weight for soybeans vs. 21 for corn, reflecting soybeans' acute sensitivity during pod fill. NDVI anomaly dominates the winter wheat model (39) because canopy greenness is a direct measure of crop condition in a crop that winters in the ground. Precipitation is excluded from wheat because timing matters more than cumulative total and the raw correlation was near zero.

Score ratings

Every score comes with a plain-English rating field:

Use cases

Crop insurance: early-season loss screening

Crop insurers typically wait for adjuster reports to identify stressed counties — by which time losses have already materialized. The Crop Stress Index provides an early-warning signal: a county whose score climbs from 30 to 60 over three consecutive weeks is showing deteriorating conditions across multiple signals, not a single noisy reading.

Use the history endpoint to build a time series of index values through the growing season. Flag counties where the score has risen 20+ points in four weeks and cross-reference against the policy book to prioritize adjuster deployment before claims are filed.

Agricultural lending: portfolio-wide risk scoring

A farm loan portfolio secured by agricultural land in 150 counties needs a consistent way to rank current-season risk. The Crop Stress Index gives you exactly that — one number per county, comparable across geographies, updated weekly. Sort your portfolio by score, filter to "elevated" and above, and you have a prioritized watchlist for the current season.

Combine with the /portfolio endpoint to pull the Crop Stress Index alongside drought, soil moisture, crop progress, and NRI scores for every county in your book in a single request.

AgTech platforms: risk-aware agronomic recommendations

Field-level recommendations — irrigation scheduling, fertilizer application timing, pest scouting priority — are more accurate when placed in regional context. A field that looks fine on internal sensors but sits in a county with a score of 70 is in a very different situation than one in a score-15 county. The Crop Stress Index gives AgTech platforms a reliable county-level baseline to calibrate against.

Commodity analytics: in-season yield signal

County-level crop stress during the critical growing window (corn: June–August; winter wheat: April–June) is a leading indicator of yield deviations from trend. The Crop Stress Index aggregates the same signals that professional crop scouts use into a structured, queryable API — useful for building proprietary yield forecasting models or cross-referencing with futures price moves.

API endpoints

Both endpoints require an API key in the X-API-Key header and a Starter plan or above. The Crop Stress Index is not available on the free tier.

Current score

Returns the current Crop Stress Index for a county — score, rating, crop model used, and each component's individual contribution. One request.

# Current Crop Stress Index for McLean County, IL (FIPS 17113 — top corn county)
curl https://r2data2.com/county/17113/crop-stress-index \
  -H "X-API-Key: agr_your_key_here"

{
  "fips": "17113",
  "county_name": "McLean",
  "state_abbr": "IL",
  "as_of": "2025-07-10",
  "score": 58,
  "rating": "elevated",
  "crop": "CORN",
  "components": {
    "drought": 26,
    "vpd": 11,
    "precipitation": 7,
    "soil_moisture": 7,
    "ndvi_anomaly": 5
  },
  "data_coverage_pct": 100,
  "model_note": null
}

The components object shows each signal's raw score on its own scale — useful for understanding what's driving the index. In this example, drought (D2 with worsening trajectory) is the dominant driver, with VPD and precipitation adding secondary pressure.

Historical time series

Returns the Crop Stress Index for each weekly USDM observation date in a date range — useful for tracking stress progression through the growing season, flagging multi-week deterioration, or building actuarial inputs for historical loss modeling.

# Crop stress through corn's critical window: June–August 2024
curl "https://r2data2.com/county/17113/crop-stress-index/history?start_date=2024-06-01&end_date=2024-08-31" \
  -H "X-API-Key: agr_your_key_here"

{
  "fips": "17113",
  "county_name": "McLean",
  "state_abbr": "IL",
  "crop": "CORN",
  "model_note": null,
  "records": [
    {
      "as_of": "2024-08-29",
      "score": 42,
      "rating": "moderate",
      "components": { /* ... */ },
      "data_coverage_pct": 100
    }
  ]
}

See the Data Dictionary for the full field reference and API Docs for interactive request testing.

Frequently asked questions