Abstract:Research on carbon dynamics and source processes in terminal lakes contributes to a precise characterization of their role as "sources" or "sinks" in the regional carbon cycle. This study selected Lake Ulansuhai, a terminal lake in northern China, as the research object. Using methods such as the Bayesian mixing model (MixSIAR) and a carbon isotope two-end-member mixing model, the sources of different forms of carbon in the water body and their main influencing factors were investigated. The main results are as follows: Both dissolved organic carbon (DOC) and particulate organic carbon (POC) showed a spatially decreasing concentration trend from north to south (p<0.05). In April, the primary source of both DOC and POC was phytoplankton, whereas in July and October, the main source was irrigation return flow. For DOC, the overall source contribution proportions were irrigation return flow (72%) > phytoplankton (15%) > terrestrial C3 plants (7%) > aquatic plants (6%). For POC, the overall proportions were phytoplankton (39%) > irrigation return flow (34%) > aquatic plants (15%) > terrestrial C3 plants (12%). Spatially, no significant differences were found for dissolved inorganic carbon (DIC) and particulate inorganic carbon (PIC) (p>0.05). The sources of DIC varied significantly across months: in April, it was primarily contributed by soil organic matter (92.7%) and atmospheric exchange (7.3%); in July, the main sources were irrigation return flow (74.1%) and biological activity (25.9%); in October, the main sources were irrigation return flow (71.9%) and atmospheric exchange (28.1%). PIC mainly originated from endogenous autogenic precipitation. Multiple regression analysis indicated that Chl-a and TN are the main drivers of DOC concentration (R2=0.662, p<0.001); SD, TN, and SPM collectively explained the variation in POC concentration (R2=0.566, p<0.05); NH4+-N is the core regulating factor for DIC concentration (R2=0.370, p<0.001); and both TN and pH have a dual promoting effect on PIC formation (R2=0.573, p<0.05). The findings of this study can provide a scientific basis for carbon cycle research in terminal lakes within agricultural irrigation areas.