Agriculture

Satellite imagery enables remote crop monitoring: 1) Multispectral sensors capture reflectance in different wavelengths, 2) NDVI (Normalized Difference Vegetation Index) = (NIR - Red) / (NIR + Red), ranges from -1 to +1, 3) Healthy vegetation has high NDVI (0.6-0.9) due to chlorophyll absorbing red and reflecting NIR, 4) Stressed crops show lower NDVI, 5) Time-series analysis reveals growth patterns, anomalies, 6) Spatial variation within field guides variable rate application. Sources: Sentinel-2 (free, 5-day revisit), Planet (daily), Landsat. Challenges: cloud cover, resolution vs cost, ground truth validation.

Smallholder challenges: 1) Low digital literacy - need voice/vernacular interfaces, 2) Limited smartphone penetration - feature phone/USSD support, 3) Fragmented land holdings - average 1.08 hectares, 4) Connectivity issues - offline-first design essential, 5) Trust deficit - need human touchpoints (field agents), 6) Unit economics - low ARPU needs aggregation models, 7) Heterogeneous crops and practices across regions, 8) Last-mile delivery for inputs in remote areas. Solutions: agent-assisted models (DeHaat), vernacular apps, IVR/WhatsApp bots, FPO (Farmer Producer Organization) aggregation, hybrid online-offline models.

Alternative credit scoring for farmers: 1) Traditional bureau data often unavailable for farmers, 2) Alternative data sources: land records (ownership, area), satellite imagery (crop health, sowing date), transaction history (inputs purchased, produce sold), weather data, mobile usage patterns, social connections, 3) ML models correlate these with repayment behavior, 4) Psychometric assessments for behavioral indicators, 5) Seasonal cash flow modeling based on crop cycle. Benefits: financial inclusion for unbanked farmers. Challenges: data quality, model accuracy, regulatory acceptance. Examples: Samunnati, FarmFundr, Jai Kisan.

Traceability system architecture: 1) Data capture layer - mobile apps for farmers, IoT sensors for cold chain, QR scanners at checkpoints, 2) Unique identifiers - farm ID, lot numbers, batch codes linked through chain, 3) Event logging - immutable record of each custody transfer, transformation, 4) Storage - database or blockchain for tamper-evidence (blockchain adds trust but complexity), 5) Integration - APIs for supply chain partners, 6) Consumer interface - QR code scanning, transparency portal. Key events: harvest, quality test, transport, processing, packaging. Standards: GS1 for identification, EPCIS for event data. Challenge: ensuring data entry compliance across fragmented supply chain.

PMFBY technology stack: 1) Enrollment - farmer data linked with land records, crop sowing report, 2) Premium calculation - actuarial rates by crop and area, government subsidy applied, 3) Crop Cutting Experiments (CCE) - sample yield estimation (traditional but being replaced by tech), 4) Remote sensing - satellite imagery for crop health assessment, sowing confirmation, 5) Weather data integration - automatic triggers for weather-based payouts, 6) Claims processing - yield loss calculation, 7) Settlement - direct benefit transfer to farmer bank accounts. Tech improvements: WINDS (Weather-based Insurance using NDVI and Data from Space), CCE app for digital yield estimation, drones for rapid damage assessment.

IoT in precision agriculture: 1) Soil sensors - moisture, temperature, pH, nutrients at multiple depths, 2) Weather stations - hyperlocal temperature, humidity, rainfall, wind, 3) Water management - flow meters, pressure sensors for irrigation scheduling, 4) Crop sensors - canopy temperature, NDVI (proximal sensing), 5) Equipment telematics - GPS, usage tracking, maintenance alerts, 6) Livestock - GPS tracking, health monitors, feed automation. Architecture: sensors → edge gateway (LoRa, NB-IoT) → cloud platform → analytics → farmer app/automation. Challenges: power (solar), connectivity (LPWAN), cost, farmer adoption. ROI: 15-20% input savings, yield improvement.

eNAM architecture: 1) Central platform hosted by SFAC, 2) Integration with state APMC systems (varies by state), 3) Stakeholder modules: farmer registration, trader registration, commission agent, mandi officials, 4) Key flows: gate entry → assaying → lot creation → auction → payment → delivery, 5) Inter-mandi trading - lots visible across mandis, logistics coordination, 6) Payment gateway integration for online settlement, 7) Mobile app for farmers to track, 8) Analytics dashboard for price discovery, trade volumes. Challenges: APMC reform varies by state, physical infrastructure gaps, adoption by traditional commission agents. Impact: 1.75 crore+ farmers, 1,000+ mandis integrated.

AI for pest/disease detection: 1) Image classification - CNN models trained on pest/disease images (Plantix, AgroStar), 2) Data collection - crowdsourced images from farmers, expert-labeled training data, 3) Mobile deployment - TensorFlow Lite, on-device inference for offline, 4) Recommendation engine - treatment suggestions based on identification, crop stage, severity, 5) Integration with spray records, input procurement. Advanced: 6) Drone/satellite imagery for field-level detection, 7) Early warning models using weather + historical outbreak data. Challenges: image quality from farmer phones, regional variation in pest species, new pest emergence, language for recommendations. Accuracy: top models achieve 90%+ on common diseases.