Diagnosing CO 2 Anomalies in a Subtropical Forest in Eastern China: Atmospheric Filtering by Radiation, Boundary‐Layer State, and Air‐Mass Transport

Meteorological normalization is often used to reduce weather‐driven variability in atmospheric composition records, but deweathered CO 2 residuals may still retain structure related to boundary‐layer mixing and transport. Here we examine daytime post‐normalization CO 2 anomalies at a subtropical Moso bamboo forest site in Anji, eastern China, using machine‐learning normalization, boundary‐layer height (BLH) stratification, seasonal Hybrid Single‐Particle Lagrangian Integrated Trajectory trajectory clustering, and generalized additive modeling (GAM). Random forest (RF) showed the strongest agreement with observed hourly CO 2 among the tested normalization models ( R 2 = 0.91) and was used to define a statistical concentration baseline. The residual anomaly field retained systematic structure. BLH‐stratified smooths indicated approximate positive‐to‐negative anomaly transitions near 153 W m −2 under shallow BLH and 124 W m −2 under medium BLH, whereas no comparable transition was detected under deep BLH. Radiation‐normalized negative‐anomaly intensity decreased from shallow to deep BLH classes, but not within their common surface solar radiation downwards (SSRD) range, consistent with coupled radiation‐mixing conditions. Season‐specific trajectory clusters were also associated with systematic differences in CO 2 anomaly and GAM residuals, and these differences remained detectable within broad BLH classes. In the integrated GAM, SSRD, BLH, their interaction, and the season‐specific transport‐regime were all associated with remaining anomaly structure; adding trajectory clusters reduced residual deviance by about 3.8% beyond the local radiation‐mixing terms. Overall, deweathered daytime CO 2 anomalies behaved as atmospherically conditioned concentration residuals rather than direct local‐flux proxies. The results highlight that near‐surface CO 2 residuals require explicit boundary‐layer and transport context before they are used to interpret local surface‐atmosphere exchange or inform regional carbon‐flux constraints.