Bibliometric mapping of machine learning applications for satellite derived carbon-related variables

Carbon monitoring is a critical component of climate change mitigation, and satellite-based observations provide consistent, near-global insights into atmospheric carbon dynamics. In this context, machine learning (ML) has emerged as a key tool for pattern extraction and prediction from increasingly complex Earth system datasets. This study provides a systematic assessment of ML applications to XCO₂ and other satellite-derived carbon-related response variables, based on a structured review of the literature published up to August 2025. The methodology was based on a search in Web of Science, conducted in three stages. Initially, 2,707 articles were identified and 94 contained the “artificial intelligence” or “machine learning” descriptors in key metadata fields. After a detailed analysis, 42 studies on orbital carbon modeling with ML were selected. The results indicate a rapid expansion of ML applications in carbon monitoring over the past decade, reflecting the growing integration of data-driven approaches into Earth system science. Decision tree-based and ensemble methods, particularly those based on gradient boosting, are the most widely adopted and perform well across environmental datasets. However, the lack of consistent relationships between model types and target variables highlights the methodological diversity of carbon-related applications. The analysis reveals important spatial and thematic gaps, including geographic concentration and limited integration of anthropogenic variables, which are essential for accurately representing emissions and improving predictions. These findings highlight the need for greater spatial diversity and integration of heterogeneous data sources. In this context, the study outlines key directions for advancing the field, including improved data integration, expansion to underrepresented regions, and the development of more robust and interpretable modeling approaches to support environmental monitoring and decision-making.