Mapping daily 1-km resolution XCO2 in China using deep learning and multi-source data

Abstract

High-resolution spatiotemporal column-averaged CO ₂ (XCO ₂) data is essential for understanding anthropogenic carbon emissions, but current satellite limitations hinder detailed analysis. To address this, we develop the Spatial-Temporal Attention XCO ₂ Network (STAXN) to improve prediction accuracy by capturing spatial-temporal variability and multiscale influences of auxiliary variables. Monte Carlo validation demonstrates robust performance, with an RMSE of 0.90 ppm and an R² of 0.97. Using this model, we generate a 1-km resolution daily XCO ₂ dataset for China (2015–2020) and analyze XCO ₂ anomaly patterns. Seasonal XCO ₂ anomalies peak in summer and winter, with nighttime light exhibiting strong positive effects (β = 0.134, 0.107), and GPP exerting the most substantial adverse influence in winter (β = −0.200). The centroid trajectories of XCO ₂ anomalies exhibit consistent seasonal shifts, shaped by regional disparities in carbon efficiency, industrial structure, and emission intensity. These findings offer valuable insights into China's carbon emission dynamics, informing policy and management strategies.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata

Altmetrics