Technical Documentation

Verdian Ag Methodology

How we process raw geodata into verifiable agronomic intelligence to support institutional risk assessment.

Data Ingestion Pipeline

Verdian Ag combines remote sensing, weather, farmer history, uploaded evidence, and lender-facing audit trails. Satellite data remains the primary independent observation layer, while farmer records and documents are treated as provenance-weighted facts rather than unqualified claims.

Remote Sensing

We ingest continuous optical and synthetic-aperture radar (SAR) data. This multi-modal approach ensures consistent data collection regardless of cloud cover, allowing for uninterrupted monitoring.

Climatic Inputs

Satellite baselines are enriched with historical rainfall data, 14-day forecasts, and temperature extremes to contextualize the physiological stress observed from space.

Farmer Evidence

Season history, receipts, buyer records, input invoices, and uploaded documents are extracted into structured facts. Each fact carries source type, confidence, evidence strength, extraction method, reviewer status, date reliability, and a provenance weight.

Agronomic Normalization

Local crop names, pests, fertilizer products, chemical brands, and active ingredients are resolved through controlled taxonomies and alias tables so messy farmer records can be compared across fields, seasons, and markets.

Algorithmic Processing

Rather than providing raw vegetation indices, Verdian Ag processes pixel-level data through specialized agronomic models. We model crop stress responses by tracking actual canopy cover growth against idealized logistic S-curves for specific crop types in specific climate zones.

When moisture levels drop below designated threshold parameters, our models apply localized water stress penalty functions to estimate biomass accumulation discrepancies.

Processing Logic Flow

1. normalize_sar_backscatter()
2. estimate_vigor_index()
3. apply_stress_penalty()
4. generate_verdian_score()

Provenance-Weighted Intelligence

Verdian does not treat every farmer statement equally. A self-reported claim, a WhatsApp message, a parsed spreadsheet, a buyer receipt, an input invoice, and a reviewer-approved evidence item each carry different trust values. Those trust values are used inside the CCP and Verdian Score as bounded underwriting signals.

This creates an audit trail that lenders can inspect: what was claimed, where it came from, how it was extracted, how confident the system is, whether it was reviewed, and whether it links to a known crop, pest, fertilizer, chemical product, or active ingredient.

Fact Capture

Examples include crop grown, yield, fertilizer used, pest event, pest response, field loss, amount sold, sale price, buyer, irrigation profile, and verified evidence.

Trust Weighting

Each fact receives a provenance weight based on source, confidence, evidence strength, reviewer status, and date reliability. Stronger evidence gets more influence.

Lender Audit

The CCP and lender dashboard expose verified farmer intelligence so credit teams can see which claims are evidence-backed and which still need review.

Pilot Validation Path

Shadow ModeLender keeps normal underwriting while Verdian scores the same farmers in parallel.
Decision SupportVerdian CCPs become part of credit memos, risk review, exposure sizing, and monitoring.
Policy IntegrationValidated score bands can support fast-track, review, haircut, or manual-underwriting rules.

Interactive Field Intelligence

The Verdian dashboard features an interactive satellite mapping interface that provides granular field-level visibility. On initial load, the map automatically fits all monitored fields within the viewport. When a user selects a specific field from the field tabs or map overlay, the map intelligently zooms to that field's boundaries with appropriate padding and a maximum zoom cap to ensure optimal visibility without excessive magnification.

This dynamic zoom behavior adapts to field size and geographic distribution—clustered fields maintain higher zoom levels for detail, while dispersed fields trigger wider contextual views. The interface displays real-time NDVI health indices, field status indicators, and satellite grid overlays, enabling immediate visual assessment of crop performance across the entire farm operation.

Dynamic Field Zoom

Clicking any field tab triggers an animated zoom transition to that field's precise geofenced boundaries. The system calculates optimal zoom levels based on field geometry, with a maximum cap at zoom level 16 to maintain context for small fields.

Health Visualization

Field boundaries render with color-coded health status—emerald for healthy, amber for warning, rose for critical—derived from continuous NDVI monitoring. Selected fields receive enhanced border weight and visual emphasis.

Situational Ground-Truth Integration

Satellite data alone is insufficient for assessing institutional lending risk. A high-yielding crop is a stranded asset if the farmer lacks market access, faces impassable infrastructure, or lacks off-take logistics. Verdian bridges this gap through continuous farmer interaction.

Proximity & Logistics

We algorithmically map the distance from a verified farm polygon to the nearest known processing hubs and markets. However, a map cannot see a washed-out bridge. Through our WhatsApp AI Agent, we run conversational loops querying farmers about the actual, on-the-ground conditions of those access routes during harvest season. This grounds abstract spatial data in logistical reality.

Behavioral Verification

The platform verifies crop interventions by cross-referencing farmer-reported actions (e.g., "I applied urea yesterday") via WhatsApp with subsequent NDVI shifts observed from space, establishing a reliable metric of grower honesty and capability.

Structured Data Outputs

Representative Output Schema

The following JSON structure represents the standard data payload generated by the Verdian platform. This is a representative schema intended to illustrate the structure and types of intelligence our models provide to support institutional pre-verification and risk assessment.

Response: field_performance_report.json
{"field_id": "uuid-v4-string","assessment_date": "2026-02-22T00:00:00Z","metrics": {"verdian_score": 82,"confidence_level": 0.88,"vigor_index": {"current_value": 0.65,"7_day_trend": "stable"},"water_stress": {"status": "nominal","estimated_depletion_percent": 35},"compliance_ratio": 0.90},"flags": [{"type": "heat_stress_warning","severity": "medium"}]}