Canada's climate goals require reliable methods to evaluate how industrial emission reductions translate into measurable atmospheric changes. Traditional bottom-up emission inventories identify emission sources but do not capture atmospheric transport, dispersion, or carbon uptake by natural sinks. While airborne and drone-based measurements provide detailed observations, they are expensive, cover limited geographic areas, and cannot support continuous, large-scale monitoring. This research addresses these limitations by integrating facility-level emission inventories with satellite-derived CO₂ observations using a machine learning framework.

The key innovation is the integration of industrial emissions and satellite-derived CO₂ fluxes into a unified spatial model that captures complex, non-linear relationships between emission sources and atmospheric response. The framework enables counterfactual simulations to estimate how specific emission reduction strategies affect atmospheric CO₂, providing a practical tool for evaluating mitigation scenarios. By moving beyond emission reporting to quantify real atmospheric impacts, this research supports more effective, evidence-based climate policy and industrial decarbonization planning.