Capability I
AI-driven methane monitoring and emissions quantification
A tiered monitoring framework combining satellite screening, continuous low-cost sensor networks, and machine-learning calibration to support methane hotspot identification, anomaly detection, trend analysis, and QMRV reporting support.
- For landfills, wastewater treatment facilities, oil and gas sites, industrial parks, and carbon-market teams.
- Designed to move from one-time estimates toward traceable monitoring evidence.
Legally binding reporting or carbon-credit verification may require site calibration, reference instruments, or third-party validation.
Capability II
Oil sands energy efficiency analysis and decision support
KDD-based industrial data analysis using Petrinex operating records, k-means clustering, Association Rules, and Chi-square testing to identify high-efficiency operating patterns and factors associated with Steam-Oil Ratio.
- Built around real operating data rather than simulation-only analysis.
- Supports discovery of operating patterns and optimization opportunities.
This is decision support and opportunity identification, not a real-time optimization control system or a promise of direct optimal operating parameters.
Capability III
Atmospheric pollution source attribution and regional transport identification
PMF and other receptor models can resolve major pollution sources and relative contributions when monitoring data are suitable. CMB may be used when reliable source profiles exist. CALIPSO aerosol extinction data and MERRA-2 reanalysis support regional particulate transport identification.
- Supports environmental management and regulatory decision-making.
- OFP analysis can support ozone-formation-contribution-oriented priorities.
Technical source attribution supports decisions; legal responsibility requires separate regulatory, permitting, enforcement, inventory, and evidentiary processes.
Capability IV
Predictive Emission Monitoring System for industrial facilities
AI-powered PEMS uses existing process parameters and transparent Keras/TensorFlow model architectures to predict combustion-related emissions, especially NOx, with site-specific model training and validation.
- Reference case: 28 months of continuous industrial field validation.
- Reported metrics include MAE = 0.5982, r = 0.9451, and 0.14% total-emissions difference in the test set.
The reference case reported 99.93% data availability; this is not prediction accuracy. Models are site- and equipment-specific.