The Science
Behind the Data
Combining satellite observations with atmospheric science to deliver independent, audit-ready emission assessments anywhere in the world.
Atmospheric Gases Tracked
Geographic Coverage
Satellite Platforms
Data Turnaround
Uncertainty (1σ)
THE SAGE PIPELINE
Five Stages. One Pipeline.
From scope definition to final delivery, the SAGE pipeline transforms raw satellite observations into auditable, regulator-ready emission measurements.
Scope & Boundary
Define pollutants, geographic coverage, temporal granularity, and emission scopes for each asset.
Data Acquisition
Automated pipelines pull satellite data from public archives, verified with SHA-256 checksums and tamper-proof audit records.
Pre-Processing
Four-stage conditioning: radiometric calibration, geolocation alignment, atmospheric correction, and cross-sensor harmonisation.
Detection & Quantification
A multi-stage inversion framework filters noise, traces readings backward through 3D wind fields, removes background concentration, and resolves emission rates with quantified uncertainty.
Quality & Delivery
Three-tier QA review, two-person sign-off, and structured deliverables aligned with ISO and GHG Protocol standards.
Technology Stack
A Virtual Constellation
“Floodlight uses a virtual constellation approach, pulling together data from several independent satellite missions and emissions inventories to get the broadest possible spatial and temporal coverage while keeping the accuracy of the best sensors.”
When one satellite has a blind spot, the others fill it in. By combining instruments across NASA, ESA, and JAXA with global emissions inventories, SAGE keeps spatial and temporal coverage broad while preserving the accuracy of the best sensors.
| Source | Agency | Target Gases | Resolution | Revisit | XCO₂ Accuracy |
|---|---|---|---|---|---|
OCO-2 | NASA/JPL | CO₂ | 1.3 × 2.2 km | 16-day | 0.1–0.3 ppm |
OCO-3 | NASA/JPL | CO₂ | ~1–2 km | Variable (ISS) | 0.2–0.5 ppm |
Sentinel-5P | ESA | CH₄,NO₂,CO,SO₂ | 5.5 × 3.5 km | Near-daily | — |
GOSAT-I | JAXA | CO₂,CH₄ | ~10.5 km | 3-day | 0.3–0.6 ppm |
GOSAT-II | JAXA | CO₂,CH₄,CO | ~8.5 km | 6-day | 0.2–0.5 ppm |
GOSAT-GW | JAXA | CO₂,CH₄ | ~910 km swath | 6-day | 0.1–0.4 ppm |
ODIACInventory | NIES/NASA | CO₂ | 1 × 1 km | Monthly | — |
EDGARInventory | EC-JRC/PBL | CO₂,CH₄,NO₂,CO,SO₂,SF₆/NF₃/CF₄ | 0.1° (~11 km) | Annual | — |
From Satellite Signal to Emission Rate
Every data point passes through a rigorous chain of atmospheric modelling and statistical analysis before it becomes an emission estimate.
Atmospheric Back-Projection
Each satellite reading is traced backward through the 3D wind field using HYSPLIT/STILT Lagrangian particle dispersion to determine where the measured gas originated.
Background Subtraction
Natural background gas concentration is removed using either regional median or forward trajectory analysis, isolating the facility’s actual emission contribution.
Inverse Modelling & Flux Estimation
A Bayesian statistical model calculates scale factors to align simulated and observed values, producing calibrated emission rate estimates with full uncertainty ranges.
simulations per estimate
Each estimate is the product of a Monte Carlo ensemble: thousands of simulations with perturbed inputs, producing a full probability distribution rather than a single number.
Variance decomposition typically attributes the largest share to wind transport uncertainty, with retrieval noise a distant second. Every point estimate ships with a 95% confidence interval.
Composite Uncertainty Budget
| Component | Typical Range | Main Drivers |
|---|---|---|
| Instrument accuracy | 2–8% | Satellite platform, soundings, atmospheric conditions |
| Temporal sampling | 3–15% | Usable overpasses, emission variability |
| Meteorological transport | 2–20% | Terrain complexity, atmospheric stability |
| Inverse modelling | 3–10% | Particle count, integration time |
| Background determination | 1–5% | Regional gradients, season |
| Prior emission inventory | 3–10% | Data availability, inventory quality |
| Composite: 5–20% (1σ). Best-case: 5–10%. | ||
From Asset to Neighbourhood.
The same atmospheric machinery that quantifies a single industrial facility scales upward. ECO CITY turns satellite overpasses into block-by-block emission intelligence, so urban planners can target the streets, sectors, and infrastructure that actually move the number.
Grid-Based Hotspotting
Cities are partitioned into a uniform grid (typically 3x3 km, refined to 1x1 km in dense districts). SAGE runs independently inside every cell, surfacing the neighbourhoods that drive the citywide budget instead of hiding them in an aggregate total.
Multi-Overpass Convergence
Every additional satellite pass approaches the source from a different wind direction. As trajectories accumulate, the probable source footprint is refined. Spatial precision is not built in; it emerges from the data.
Multi-Gas Source Attribution
Different sources leave different chemical signatures. By co-locating retrievals across CO2, CH4, NO2, CO, and SO2, SAGE distinguishes combustion from fugitive emissions, heavy industry from transport, and clean fuel from dirty.
Chemical Fingerprints
How co-located multi-gas retrievals reveal what kind of source produced a given enhancement.
| Gas Signature | Inferred Source Type | Typical Examples |
|---|---|---|
| CO₂ + NO₂ | Combustion source | Power plant, heating system, traffic corridor |
| CH₄ only | Fugitive release | Landfill, gas distribution leak, wastewater facility |
| CO₂ + CO | Incomplete combustion | Older industrial facility, biomass burning |
| SO₂ elevated | High-sulphur fuel | Heavy fuel oil, coal-fired plant, refinery flare |
Infrastructure upgrades and inspections where each tonne of avoided emission costs the least.
Low-emission zones, transit corridors, and zoning policy backed by observed concentration data, not models alone.
The real-world impact of climate action plans at the neighbourhood level, in months not years.
QUALITY & GOVERNANCE
Three Tiers of Quality Assurance
“Floodlight keeps false positives below one percent by layering automated filters, structured human review, and a tamper-evident audit trail into three nested QA tiers.”
Scene-Level
Cloud shadow, sun glint, high aerosol, and retrieval warning flags checked. ~35% of raw soundings rejected. Passing data is re-projected, bias-corrected, and checksummed.
Project-Level
QA Manager verifies parcel shapefile consistency, raster alignment, and sufficient overpass coverage for the target. A Quality Score is assigned based on root-sum-square error propagation.
Release-Level
Senior Analyst and independent reviewer perform full inspection. Logged in timestamped audit ledger. Any post-sign-off change blocks delivery until resolved.
Standards Alignment
The methodology aligns with ISO 9001:2015, ISO 14064-1, the GHG Protocol, giving investors, regulators, and city governments a transparent, auditable picture of real-world emissions.
PEER-REVIEWED FOUNDATIONS
References
The SAGE methodology builds on peer-reviewed atmospheric science and validated satellite data products.
Taylor, T. E., O’Dell, C. W., Baker, D., et al. (2023). “Evaluating the consistency between OCO-2 and OCO-3 XCO₂ estimates derived from the NASA ACOS version 10 retrieval algorithm.” Atmospheric Measurement Techniques, 16, 3173–3209.
DOI: 10.5194/amt-16-3173-2023Imasu, R., Matsunaga, T., Nakajima, M., et al. (2023). “Greenhouse gases Observing SATellite 2 (GOSAT-2): mission overview.” Progress in Earth and Planetary Science, 10, 33.
DOI: 10.1186/s40645-023-00562-2van Geffen, J., Eskes, H., Compernolle, S., et al. (2022). “Sentinel-5P TROPOMI NO₂ retrieval: impact of version v2.2 improvements and comparisons with OMI and ground-based data.” Atmospheric Measurement Techniques, 15, 2037–2060.
DOI: 10.5194/amt-15-2037-2022Dowell, D. C., Alexander, C. R., James, E. P., et al. (2022). “The High-Resolution Rapid Refresh (HRRR): An hourly updating convection-allowing forecast model. Part I.” Weather and Forecasting, 37(8), 1371–1395.
DOI: 10.1175/WAF-D-21-0151.1Nassar, R., Moeini, O., Mastrogiacomo, J.-P., et al. (2022). “Tracking CO₂ emission reductions from space: A case study at Europe’s largest fossil fuel power plant.” Frontiers in Remote Sensing, 3, 1028240.
DOI: 10.3389/frsen.2022.1028240Laughner, J. L., Toon, G. C., Mendonca, J., et al. (2024). “The Total Carbon Column Observing Network’s GGG2020 data version.” Earth System Science Data, 16, 2197–2260.
DOI: 10.5194/essd-16-2197-2024Crippa, M., Guizzardi, D., Pagani, F., et al. (2024). “Insights into the spatial distribution of global, national, and subnational greenhouse gas emissions in the EDGAR v8.0.” Earth System Science Data, 16, 2811–2830.
DOI: 10.5194/essd-16-2811-2024Zanaga, D., Van De Kerchove, R., Daems, D., et al. (2022). “ESA WorldCover 10 m 2021 v200.” Zenodo.
DOI: 10.5281/zenodo.7254221Jacob, D. J., Varon, D. J., Cusworth, D. H., et al. (2022). “Quantifying methane emissions from the global scale down to point sources using satellite observations.” Atmospheric Chemistry and Physics, 22, 9617–9646.
DOI: 10.5194/acp-22-9617-2022
The Standard for Climate Risk Intelligence.
Rigorous engineering meets climate science. Join leading institutional investors and corporations in building a resilient future.