Decoupling of spatiotemporal variations of lake CO2 emissions in China: Inconsistency between seasonal and spatial dominant factors

Lake CO 2 emissions have become a critical component shaping the stability of terrestrial carbon sinks and regional carbon budgets, yet existing studies often neglect the spatial heterogeneity and scale effects of the explanatory variables, resulting in substantial uncertainties in the assessment of lake CO 2 emissions. In this study, we examined seasonally dominant explanatory variables and their scale-dependent spatial associations with lake CO 2 fluxes across China. The results showed that lake CO 2 fluxes exhibited a spatiotemporal pattern, with higher mean values in winter (−0.36 to 10.53 mmol m −2 d −1 ; 1.54 ± 1.45 mmol m −2 d −1 ) and lower values in summer (−0.40 to 7.00 mmol m −2 d −1 ; 1.01 ± 1.32 mmol m −2 d −1 ; p < 0.001); annual mean values were highest in the Northeast Plain and Mountain Lake Zone (2.27 ± 1.55 mmol m −2 d −1 ) and lowest in the Inner Mongolia-Xinjiang Lake Zone (0.63 ± 0.69 mmol m −2 d −1 ). The global spatial autocorrelation of CO 2 fluxes strengthened from spring to winter (Moran's I = 0.348–0.411). The multiscale geographically weighted regression (MGWR) model achieved the best goodness of fit in all seasons (R 2 = 0.68–0.86), effectively capturing the spatial heterogeneity of the explanatory variables. Lake area exhibited the highest explanatory power in spring and winter (Q = 0.489 and 0.287, respectively), whereas DO ranked highest in summer and autumn (Q = 0.381 and 0.196, respectively). OPGD interaction detection showed that nonlinear enhancement was the predominant interaction type, suggesting that spatial heterogeneity in CO 2 fluxes reflects the joint influence of multiple explanatory variables. Scale-effect analysis based on MGWR bandwidths revealed that lake attributes (lake area, average depth, and average discharge) exhibited broad-scale effects, with relatively stable influences across larger spatial ranges (bandwidths: 159–181 km). In contrast, water quality indicators (DO and NH 4 + ) and climatic factors (PRE and SSR) showed localized effects due to their smaller bandwidths (23–40 km). The detector analysis further identified high-emission zones associated with small lakes and conditions of low DO, underscoring the necessity of differentiated management for small waterbodies and environmentally sensitive regions in lake‑carbon governance. Overall, this study demonstrates that explicitly incorporating spatial heterogeneity and multiscale effects is essential for accurately assessing lake CO 2 emissions and for designing spatially and seasonally targeted carbon-mitigation strategies.