Long-term (2000-2020) global 0.05° continuous atmospheric carbon dioxide mapping combining OCO-2 observations and...

Area of Science:

• Earth Science
• Atmospheric Science
• Remote Sensing

Background:

• Accurate, long-term global atmospheric CO2 monitoring is crucial for understanding climate change and carbon cycle dynamics.
• Existing datasets often lack global seamless coverage or sufficient spatial resolution.
• Remote sensing offers a promising avenue for continuous global CO2 observation.

Purpose of the Study:

• To develop a high-resolution, continuous global column-averaged CO2 (XCO2) dataset from 2000-2020.
• To improve upon existing CO2 products in terms of spatial coverage, resolution, and accuracy.
• To provide a valuable tool for climate change research and carbon balance studies.

Main Methods:

• Reconstruction of an 8-day global XCO2 product (GCXCO2) using a stacking machine learning method.
• Integration of terrestrial/marine remote sensing data with model simulations.
• Development of a dynamic normalization strategy to address temporal variations and ensure model temporal expansion.

Main Results:

• The GCXCO2 product exhibits seamless global coverage with a 0.05° spatial resolution, outperforming CarbonTracker and CAMS simulations.
• Validation via 10-fold cross-validation yielded R² = 0.974 and RMSE = 0.551 ppm.
• The dataset accurately predicted XCO2 in regions and periods lacking satellite observations (R² = 0.958 and R² = 0.886).
• Comparison with TCCON ground stations confirmed GCXCO2's superior accuracy and resolution over model data.
• An upward global XCO2 trend of ~2.105 ppm/year was identified between 2000-2020, with distinct Northern and Southern Hemisphere seasonal patterns.

Conclusions:

• The GCXCO2 dataset provides a reliable, high-precision, long-term record of global XCO2.
• It serves as an essential tool for advancing the understanding of climate change and carbon balance.
• The dataset aids in detecting CO2 concentration anomalies and supports climate modeling efforts.