【Objective】Stable carbon and nitrogen isotopes (δ13C and δ15N) have been widely applied due to their excellent source specificity and relative conservatism. They are particularly valuable in studies such as land-use identification and quantification of soil erosion effects. However, in the context of global climate change, the mechanisms underlying water-driven fractionation of stable carbon and nitrogen isotopes in soils remain unclear, which limits the accuracy of organic carbon source apportionment. Thus, this study aimed to investigate the mechanisms by which moisture conditions influence stable carbon and nitrogen isotope signals in soil and to analyze how the non-conservative behavior of stable carbon and nitrogen isotopes under different moisture conditions affects the accuracy of organic carbon source apportionment. 【Method】In this study, we designed soil incubation experiments under four typical moisture scenarios (drying, wetting, flooding, and drying-wetting cycles). We systematically investigated the patterns of stable carbon and nitrogen isotope signals driven by moisture and their implications for organic carbon source identification. In addition, we supplemented our analysis with Bayesian mixing models (MixSIAR model) and random forest models.【Result】The results showed that when using δ13C for tracing, the contribution bias for each source was generally below 10 %. Its significant conservatism makes it an excellent tracer for carbon source identification. In contrast, the δ15N signal is sensitive to moisture. Under extreme moisture conditions such as flooding and alternating wetting and drying, δ15N showed strong fractionation effects. Using δ15N alone for source identification results in substantial errors, with source contribution biases ranging from 0.067% to 24.77%. The combined application of δ13C and δ15N achieved higher apportionment accuracy under most moisture scenarios. However, it was still affected by the fractionation of δ15N under extreme moisture fluctuations.【Conclusion】The robustness of δ13C highlights its potential as a core indicator for analyzing organic carbon sources, while correction methods should be incorporated for δ15N to enhance result reliability. This study not only deepens the understanding of the processes and mechanisms of δ13C and δ15N signal fractionation in the context of climate change but also breaks through the inherent black box theory in traditional source apportionment methods for organic carbon sources at a mechanistic level. It also provides scientific support for optimizing organic carbon source apportionment methods under climate change.