HERBARIUM LABSThe Verified Data Layer for Agricultural Finance
Machine-generated, edge-signed, chain-anchored. Every signal Herbarium produces is independently verifiable.
Architecture Overview
The Herbarium protocol consists of three interconnected pillars that together form the verified data layer for agricultural finance.
DePIN Layer
Decentralized Physical Infrastructure Network of IoT sensors deployed on farms, capturing microclimate conditions and generating continuous data streams.
TraceOrigins
Composable on-chain provenance protocol. Every harvest batch becomes a verifiable, chain-anchored record connecting farmers to consumers and data to capital markets.
ParaFi
API-accessible risk signals for agricultural finance. Parametric insurance, yield pre-purchase, and verified agricultural data connected to on-chain capital markets.
Three-Stage Pipeline
Sense
IoT sensor networks deployed directly on farms capture microclimate conditions in real-time: temperature, humidity, soil moisture, light levels, CO2 concentration, and more. Data is collected at the edge and cryptographically signed before transmission.
- Real-time environmental monitoring
- Edge-signed data integrity
- Tamper-evident sensor feeds
Model
AI-driven analysis transforms raw sensor data into actionable farm state models. Machine learning algorithms trained on agricultural datasets detect anomalies, predict disease outbreaks, and assess crop health with documented 98.33% accuracy in disease detection.
- 98.33% ML accuracy (disease detection)
- Anomaly detection and alerts
- Crop health assessment
Predict
Continuous yield prediction generates API-accessible risk signals for financial products. These signals power parametric insurance triggers, yield pre-purchase contracts, and credit risk assessments for agricultural lenders.
- Continuous yield forecasting
- API-accessible risk signals
- Financial product integration
Hardware Specifications
Sensor Array
- TemperatureDHT22 / SHT31
- HumidityDHT22 / SHT31
- Soil MoistureCapacitive v2.0
- Light (PAR)BH1750 / TSL2591
- CO2MH-Z19B / SCD40
- Air QualityBME680
Edge Compute
- ProcessorESP32-S3 / RPi CM4
- ConnectivityWiFi / LoRa / 4G
- Storage32GB eMMC
- PowerSolar + Battery
Vision Module
- Camera5MP OV5647 / 12MP IMX477
- ProcessingOn-device ML inference
- Use CaseDisease detection, growth tracking
Trust Architecture
Data that cannot be independently verified is not data. Herbarium's trust architecture ensures every signal is tamper-evident and auditable.
Machine-Generated
All data originates from calibrated sensors, not human input. This eliminates the possibility of manual data entry fraud and ensures consistency across the network.
Edge-Signed
Data is cryptographically signed at the point of collection using hardware security modules. Any tampering between sensor and chain is detectable.
Chain-Anchored
Merkle roots of sensor data batches are anchored to public blockchains, creating an immutable audit trail that anyone can verify independently.
Anti-Manipulation
Cross-validation between sensors, anomaly detection algorithms, and physical tamper-evident enclosures make gaming the system economically irrational.
Research
Blockchain-Enabled Cyber-Physical Systems for Agricultural Risk Management
Our research proposes a blockchain-enabled human-cyber-physical system that integrates multimodal sensing, AI-driven inference, and verifiable data pipelines to enable predictive agricultural risk management. The system achieves 98.33% accuracy in disease detection using convolutional neural networks trained on agricultural image datasets.