Preliminary · not final The figures shown are a county-level proxy (v0) built from ERA5 reanalysis and official INSSE yield statistics. Field-level indices on 13.8M+ DigiFarm field boundaries are pending HPC processing. 2025 yields are not yet published by INSSE; 2026 is provisional. Read these as risk-signal evidence and trigger-design proof — not exact insured loss or quality figures.
DigiFarm — a Fronterax Technologies company

The DigiFarm HeatRisk Engine

Field-level heatwave intelligence for crop insurance, underwriting, and agricultural supply chains — turning crop-stage-aware heat exposure into priceable, auditable risk.

field boundaries indexed· countries· 5 heat indices· backtest· live RO sunflower parametric precedent (r=0.829, blind 2024)
01

Romania growing-season heat exposure, by county

Choropleth = days above 35°C during the crop growing season (ERA5). Pick a year; click a county for its heat & yield history.
Heat exposure — 2024
Growing-season days >35°C · ERA5 reanalysis
Year 2024
02

Backtest proof — heat exposure tracks yield loss

Pooled county×year, Romania 2017–2024. Each point is one county in one year. INSSE actuals vs ERA5 growing-season heat.
Sunflower
Maize
Crop-stage-aware heat exposure tracks county yield losses. Each additional growing-season day above 35°C carries a measurable, statistically significant yield penalty for both crops (sunflower r=−0.44, maize r=−0.46, p<0.001). The signal strengthens in the worst heat year (2024) and reverses cleanly in cool years (2018: zero hot-days → +26% sunflower, +54% maize). Honest / preliminary — county proxy
03

Five crop-stage heat indices

Each index isolates a distinct physiological damage pathway, gated to the crop’s own phenological stage from DigiFarm CropID.
RHSI Reproductive Heat Stress Index
MeasuresAccumulated exposure above crop-specific heat thresholds (>32–35°C) during flowering & grain/seed fill — the window where heat sterilises florets and aborts kernels.
Why it mattersReproductive heat is the dominant driver of catastrophic yield collapse; calendar-day metrics miss it.
Customer useCore trigger variable for parametric sunflower/maize covers.
TNRI Tropical-Night Respiration Index
MeasuresRun of warm nights (Tmin above threshold) that drive respiratory carbon loss and prevent overnight recovery during grain fill.
Why it mattersNight heat erodes yield & grain weight even when daytime maxima look benign — invisible to daytime-only triggers.
Customer useBasis-risk reducer; explains losses daytime indices leave unpriced.
VPDI Vapour-Pressure-Deficit Index
MeasuresAtmospheric drought stress — the evaporative demand the canopy faces — combining temperature and humidity, with soil-moisture context.
Why it mattersSeparates “hot but humid” from “hot and desiccating”; high VPD compounds heat damage.
Customer useDistinguishes heat-only from combined heat×drought events for cover design.
HQDI Heat-Quality Downgrade Index
MeasuresHeat exposure timed to quality-critical windows — oil content in sunflower, protein/test-weight in cereals, kernel integrity in maize.
Why it mattersHeat hits grade and price, not just tonnage; processors are exposed to quality, not yield.
Customer useQuality-downgrade cover & procurement margin protection.
OHDI Onset / Heat-Duration Index
MeasuresTiming, length and intensity of heatwave spells relative to phenology — early-onset vs prolonged vs compound events.
Why it mattersA 3-day spike and a 14-day plateau price very differently; duration & onset shape the loss curve.
Customer useTrigger shape & payout-curve calibration; portfolio accumulation.
04

Country readiness

Where the engine is field-level today, where yields carry a caveat, and where coverage is on a public-data enablement path. Honest status.
RO
Romania
Field-level ready
Full CropID + field boundaries + ERA5/AgERA5 + INSSE backtest. Live sunflower parametric precedent already in market.
HU
Hungary
Field-level ready
CropID 92.8% (ecrop CTY); field boundaries and heat-index stack ready for client backtests.
BG
Bulgaria
Yield caveat
CropID 91.9% and heat indices ready; official yield series thinner — backtest validity carries a documented caveat.
SK
Slovakia
Yield caveat
CropID 93.5% and heat indices ready; smaller cropped area and yield-series depth limit backtest strength.
FR
France
Region-tier · deferred
Region-tier indices today; field-level deferred pending public CropID (RPG/IGN parcel data) enablement path.
05

Built for two sides of the heat-risk market

The same crop-stage heat indices feed underwriting and procurement — wherever heat risk is currently underpriced or invisible.
01Underwriting score
A per-field / per-portfolio heat-risk score combining the five indices with crop, phenology and soil context — pricing input before binding.
02Trigger design
Index-trigger and payout-curve design backed by the county backtest, minimising basis risk against realised loss.
03Basis-risk audit
Independent audit of an existing parametric trigger vs field-level heat & yield — quantify residual basis before renewal.
04Post-event certificates
Objective, auditable heat-exposure certificates after a heatwave for fast, disputable-free parametric settlement.
05Portfolio accumulation
Spatial accumulation & correlation of heat exposure across a book — concentration and PML insight for reinsurance.
Proof in market: DigiFarm is the independent calculation agent on a live Romania sunflower parametric product (risk carrier Syngenta / Marsh). Measured basis 7.67 pp with blind-2024 validation r = 0.829 — the same crop-stage heat methodology shown here, already settling real contracts.
01Procurement shortfall
Early-season heat-driven volume-shortfall signal across sourcing catchments — months before harvest actuals.
02Quality downgrade
HQDI flags heat hitting oil content / protein / test-weight windows — grade & price risk, not just tonnage.
03Harvest timing
Phenology-aware heat & senescence signals to sequence harvest logistics and storage allocation.
04Catchment heat exposure
Field-resolved heat exposure aggregated to each plant / silo / port catchment — supply-security mapping.
05Margin-at-risk
Translate catchment heat exposure into volume × quality × price margin-at-risk for hedging & sourcing.
Same evidence base: the methodology validated on the live RO sunflower parametric cover (DigiFarm = calculation agent, Syngenta / Marsh; basis 7.67 pp, blind-2024 r = 0.829) reuses directly for procurement — heat that triggers an insurance payout is the same heat that erodes a processor’s contracted volume and grade.
06

Methodology

From field boundaries to a backtested, auditable heat-risk signal.
1
DigiFarm field boundaries
13.8M+ delineated fields across RO/HU/BG/SK (118M pan-European, euds v7) as the spatial unit of risk.
2
CropID per field
ecrop crop classification (RO 86.2% / HU 92.8% / BG 91.9% / SK 93.5%) assigns crop & thus the relevant heat thresholds.
3
Crop-stage phenology
Sentinel response curves locate flowering, grain/seed fill and quality windows — heat is scored only when it bites.
4
Heat & water forcing
ERA5-Land / AgERA5 hot-days, tropical nights, VPD and growing-season water deficit, joined to soil & terrain context.
5
Five crop-stage indices
RHSI · TNRI · VPDI · HQDI · OHDI computed per field × season.
6
Public-yield backtest
Validated against official INSSE county yield statistics (2017–2024) — the evidence shown above (r=−0.44 / −0.46).
What this is — and isn’t
The engine produces risk indicators, backtests and trigger-design evidence. With public data alone it does not output exact insured loss or quality figures.
  • Current public numbers are a county-level proxy (v0); field-level indices are pending HPC.
  • 2025 INSSE yields are unpublished; 2026 is provisional.
  • Exact insured-loss & quality calibration requires the client’s private loss / settlement labels — which we run under a private backtest.

See where your heat risk is underpriced — or invisible.

Send us your insured portfolio or procurement catchments and we’ll run a private HeatRisk backtest against field-level, crop-stage heat exposure — and show you exactly where current pricing or sourcing assumptions miss it.

Request a private backtest → konstantin@digifarm.io