<p>Effective identification of the spatiotemporal heterogeneity of terrestrial carbon sinks is crucial for achieving carbon neutrality. However, current assessment efforts face challenges stemming from methodological differences and data inconsistencies, particularly in tracking micro-scale carbon stock dynamics and distinguishing between emission sources and carbon sinks. Therefore, this study developed a “multi-model estimation-multi-factor verification” (MEFV) framework that integrates carbon stock dynamics and source-sink equilibrium to address these limitations. The dynamic changes of carbon storage (CS) and net ecosystem productivity (NEP) in Xianyang from 2000 to 2020 were estimated. To enhance the reliability of the carbon storage results, carbon sinks and carbon sources in different regions were tracked and verified. Additionally, an ecosystem quality assessment model was introduced to analyze the causes of carbon sink conversion from an indicator perspective, exploring the interrelationship and regulatory mechanisms between ecosystem quality and carbon sinks. To deeply explore the underlying mechanisms, the XGBoost-SHAP framework was applied to quantify the non-linear contributions and interactions of eight factors, including SEI and DI. The results show that: (1) CS exhibited a modest decline of 7.325 × 10⁶ t between 2010 and 2020, in contrast to a more pronounced increase during 2000 ~ 2010. (2) The change in CS in the previous decade was significant, with 76.56% of the total area transferred by grade, with the largest area transferred by grade III. (3) The driving mechanisms have evolved from being dominated by human activities in 2000, to climate regulation in 2010, and finally to mandatory controls using pollution as a means in 2020. The Drought Index (DI) was the primary limiting factor (coefficient of determination R² ranging from 0.704 to 0.842 across the three study periods), while the synergistic effect between high PM₂.₅ concentrations and drought stress significantly exacerbated the decline in carbon sequestration. (4)The spatial evolution of ecological environmental quality fragility (EEQF) is strongly correlated with the redistribution of CS “sources” and “sinks”. The EEQF was primarily driven by the “stress” dimension, in which natural factors (SEI and DI) and anthropogenic factors (UR and ACE) interact to exacerbate fragility, thereby increasing the EEQF and consequently reducing the carbon sequestration capacity. This study provided a dynamic equilibrium perspective on ecosystem evolution analysis and established a modelling framework to achieve source-sink-balanced ecosystems.</p> Graphical abstract <p></p>