Fossil fuel CO2 emissions over metropolitan areas from space: A multi-model analysis of OCO-2 data over Lahore, Pakistan

• DOI: 10.1016/j.rse.2021.112625
• Publication Date: 2021-08-09T23:04:06Z
• Authors: Ruixue Lei; Sha Feng; Alexandre Danjou; Grégoire Broquet; Dien Wu; John C. Lin; Christopher W. O'...; Thomas Lauvaux
• Keywords: Emission inventory

Urban areas, where more than 55% of the global population gathers, contribute 70% anthropogenic fossil fuel carbon dioxide (CO2ff) emissions. Accurate quantification CO2ff emissions from urban areas is great importance for formulating warming mitigation policies to achieve neutrality by 2050. Satellite-based inversion techniques are unique among "top-down" approaches, potentially allowing us track emission changes over cities globally. However, their accuracy still limited incomplete background information, cloud blockages, aerosol contamination, and uncertainties in models priori inventories. To evaluate current potential space-based techniques, we present first attempt monitor long-term based on OCO-2 satellite measurements column-averaged dry-air mole fractions CO2 (XCO2) a fast-growing Asian metropolitan area: Lahore, Pakistan. We examined data availability at scale. About 17% soundings 70 most populated 2014 2019 marked as high-quality. Cloud blockage contamination two main causes loss. As an recover additional soundings, evaluated effectiveness quality flags city level comparing three flux methods (WRF-Chem, X-STILT, cross-sectional integration method). The satellite/bottom-up (OCO-2/ODIAC) ratios high-quality tracks with reduced better agreed across compared all-data tracks. This demonstrates that useful filters low-quality retrievals local scales. All consistently suggested ratio medians greater 1, implying ODIAC slightly underestimated Lahore. Additionally, our estimation posteriori trend was about 734 kt C/year (i.e., annual 6.7% increase). 10,000 Monte Carlo simulations Mann-Kendall upward test showed less 10% prior uncertainty 8 (or 20% 25 tracks) required greater-than-50% significant possibility 95% confidence level. It implies driven not due assimilation retrievals. key improving role detection lies collecting frequent near constraints XCO2