TL;DR
This paper introduces a novel sub-national disaggregation method (SDM) to create the first global 0.1°x0.1° resolution inventory of fuel combustion (PKU-FUEL) and CO2 emissions (PKU-CO2). This high-resolution approach significantly reduces spatial bias and uncertainty in emissions mapping compared to traditional national-level methods, offering crucial data for climate modeling and policy development.
What Problem Does This Paper Address?
Accurately mapping CO2 emissions from fuel combustion is vital for understanding the global carbon cycle, predicting pollutant transport, and designing effective climate mitigation strategies. However, existing global emission inventories often rely on national-level data and disaggregate emissions using broad proxies like population density. This approach can distort the true geographical distribution of emissions within countries, especially in regions with significant internal variations in fuel consumption and economic activity. This "spatial bias" limits the accuracy of climate models and the effectiveness of localized emission reduction policies. The authors aim to overcome this by developing a more granular, sub-national approach to emission mapping.
Methods
The core innovation is the Sub-national Disaggregation Method (SDM). This method leverages detailed sub-national fuel consumption data (e.g., county-level, state/province-level) for 45 major countries. For countries lacking such detailed data, regional regression models were developed. These sub-national data, along with national data for other countries, are then disaggregated to a 0.1°x0.1° grid resolution using various proxies specific to different fuel types and sectors.
To quantify uncertainties, the authors employed a Monte Carlo simulation approach, running the inventory calculations 1000 times with randomly varied input data. This allowed them to derive metrics like the 95th minus 5th percentile range (R90) and the relative uncertainty (R90/M) for both global emissions and their spatial distribution.
Datasets
The paper introduces the PKU-FUEL (fuel consumption) and PKU-CO2 (CO2 emissions) inventories. These inventories are built upon:
- Sub-national fuel consumption data: Collected for 45 countries at various sub-national levels (counties, states/provinces, 0.5°x0.5° grids).
- National fuel data: For countries lacking sub-national data.
- Various proxies: Including vegetation density, CO emission proxies, night-light data, crop production statistics, waste generation data, and population density.
- Emission Factors (EFs): Derived from literature for different fuel types and combustion processes.
- CarbonTracker inverse model output: Used for terrestrial ecosystem carbon flux calculations.
The authors also created a mock-up inventory (NAT-CO2) using national data and disaggregation to compare against their PKU-CO2 inventory, highlighting the benefits of their SDM. They also compared their results with two existing inventories: ODIAC (global, satellite nightlight-based) and VULCAN (US-specific, detailed process-based).
Key Insights and Ideas
- Sub-national data is crucial for accurate emission mapping: The stark differences in per-capita CO2 emissions between urban and rural areas in developing nations (2.09 vs. 0.600 Mg C cap⁻¹ yr⁻¹) highlight how national averages can mask significant local variations.
- SDM dramatically reduces spatial bias: By moving from national to sub-national disaggregation, the uncertainty range for CO2 emissions across grids was reduced by over 80% (from 417 to 68.2 Mg km⁻² yr⁻¹).
- Nightlight data has limitations for emission mapping: The ODIAC inventory, relying on satellite nightlights, showed a tendency to over-concentrate emissions in urbanized regions, potentially missing significant emissions from rural industrial zones or settlements.
- Urbanization is a major driver of CO2 emissions in developing countries: The large urban-rural gap in developing nations suggests that rapid urbanization is a key factor influencing future emission trends and needs to be a focus for mitigation policies.
Key Results
- Global CO2 emission in 2007: 11.2 Pg C yr⁻¹ (with a 5th-95th percentile range of 9.11–13.3 Pg C yr⁻¹).
- Reduction in spatial uncertainty: The average 95th minus 5th percentile range of CO2 emissions for all grids was reduced from 417 Mg km⁻² yr⁻¹ (using national disaggregation) to 68.2 Mg km⁻² yr⁻¹ (using SDM).
- Relative uncertainty reduction: The R90/M for CO2 emission was reduced from 364% to 64.8% by using sub-national disaggregation.
- Urban vs. Rural emissions: Globally, urban per-capita emissions were 2.45 Mg C cap⁻¹ yr⁻¹ compared to 0.795 Mg C cap⁻¹ yr⁻¹ for rural areas. This gap was much larger in developing countries (2.09 vs. 0.600 Mg C cap⁻¹ yr⁻¹) than in developed ones (3.57 vs. 3.42 Mg C cap⁻¹ yr⁻¹).
- Comparison with ODIAC: The PKU-CO2 inventory showed a more realistic spatial distribution, capturing emissions in rural industrial areas that were missed by the nightlight-based ODIAC.
- Comparison with VULCAN: The PKU-CO2 and VULCAN inventories showed reasonable agreement over the US, with localized differences attributed to VULCAN's more detailed traffic data and PKU-CO2's use of population proxy for area sources.
Limitations and Future Work
- Data availability: While the study compiled a large database of sub-national fuel consumption, more data for other large countries would further improve the inventory.
- Temporal resolution: The current inventory is for a single year (2007). Future work should aim to develop inventories with temporal resolutions, both intra- and inter-annual, to capture dynamic changes.
- Urban-rural classification: The authors acknowledge that variations in the threshold used for urban-rural classification were included in uncertainty calculations, suggesting that refining this classification could be beneficial.
- Future applications: The PKU-FUEL inventory is ready for estimating emissions of other greenhouse gases and air pollutants, which can enhance understanding of combustion-related climate forcing and health impacts. The authors also recommend explicitly using the error structure of F(x) in future atmospheric inversion studies.
Summary updated