Remote-sensing-driven GHG emission models for large-scale agriculture

Overview

This solution leverages remote-sensing and satellite image data to develop on-site greenhouse gas (GHG) emission models for agricultural environments. By pairing satellite-derived vegetative indices with data from a proprietary wireless chamber sensor network, deep learning models can estimate N2O, CH4, and CO2 fluxes at the high spatiotemporal resolution required for large-scale agriculture. The approach reduces the need for dense, invasive, and maintenance-heavy on-site sensor deployments while enabling emissions monitoring across broad geographic areas.

The underlying chamber-sensor framework uses LoRa-based wireless connectivity and NDIR (non-dispersive infrared) sensors for soil gas collection, providing high-quality ground-truth measurements. When combined with remote-sensing inputs, the resulting models offer a scalable and cost-effective pathway for quantifying agricultural GHG emissions, supporting sustainability reporting, carbon accounting, and regulatory compliance for farms and agri-businesses.

Technical specifications

Core components:

• Wireless LoRa-enabled NDIR sensor and soil-gas collection chamber platform for on-site GHG flux measurement
• Integration of satellite remote-sensing data, including vegetative indices such as NDVI, GDVI, and GCI, to capture spatiotemporal patterns in nutrient usage, biomass production, and soil reflectance
• Deep learning models trained to estimate N2O, CH4, and CO2 fluxes from remote imagery alone, with chamber sensor data used for training and baseline validation
• Scalable architecture designed to operate with limited infrastructure costs across varying field sizes and agricultural contexts

Key advantages:

• Reduces dependence on dense physical sensor networks by extending coverage through remote sensing
• Captures flux dynamics at high spatial and temporal resolution across large agricultural areas
• Supports multiple gas species (N2O, CH4, CO2) within a unified modeling framework

Technology readiness level

The chamber-sensor framework has been developed and validated in field settings, with associated publication work pending. The current stage focuses on building large crop datasets from satellite-sourced imagery and training deep learning models using paired chamber measurements. The team is actively seeking partners with N2O NDIR sensor production capabilities to scale sensor deployment and accelerate model training and validation.