China's county-level monthly CO₂ emissions during 2013-2021

Area of Science:

• Environmental Science
• Earth Science
• Data Science

Background:

• County-level CO2 emission estimation in China often uses top-down methods.
• Previous studies relied on single nighttime light brightness, assuming a uniform correlation with CO2 emissions, which is often inaccurate.
• Heterogeneous relationships between CO2 emissions and light intensity across regions are overlooked.

Purpose of the Study:

• To develop a more accurate method for estimating China's county-level CO2 emissions.
• To construct a refined dataset of instrumental variables for CO2 emission prediction.
• To support the achievement of China's county-level "dual carbon" strategy.

Main Methods:

• Constructed a dataset using improved nighttime light data, human settlement data, and socioeconomic indicators.
• Developed a hybrid regression algorithm combining deep neural networks and CatBoost.
• Generated instrumental variables with a stronger linear relationship to CO2 emissions than total nighttime brightness.

Main Results:

• Estimated China's monthly CO2 emissions at the county level from 2013 to 2021.
• The new instrumental variables showed a stronger correlation with CO2 emissions compared to traditional methods.
• The hybrid model provided more accurate and granular CO2 emission data.

Conclusions:

• The developed method offers a more robust approach to county-level CO2 emission estimation.
• This approach can be generalized to other global regions for similar environmental monitoring.
• The findings provide a strong basis for evaluating progress towards China's "dual carbon" goals.