The process of dissolved oxygen (DO) dynamics in response to temperature is usually highly complex in climate transition zones (CTZs). The intrinsic mechanism of extreme temperature events driving riverine DO dynamics, especially its significant seasonal variability, has not been fully resolved. In this study, an integrated analytical framework was constructed in the Fen River Basin (FRB), a typical semi-arid-semi-moist climate transition zone in China. This framework integrated Seasonal-Trend decomposition using Loess (STL), XGBoost and Shapley Additive explanation (SHAP) coupled model, and Partial Least Squares-Structural Equation Modeling (PLS-SEM). Through time-series decomposition, factor contribution quantification, and path analysis, this study effectively resolves the challenges in identifying complex seasonal linkages between extreme temperatures and DO fluctuations, as well as the coupled mechanisms of multiple drivers. The results show that: (1) The FRB has been warming significantly (0.368 degrees C/10a) over the past 51 years, far exceeding global and Chinese regional averages. After 1994, the amplitude of temperature-extreme fluctuations has continued to expand, indicating that the basin has entered an accelerated warming phase; (2) Quantitative analysis based on the STL-XGBoost-SHAP model (test set R2 = 0.787, RMSE = 0.365 mg/L) demonstrated that temperature extremes were the primary drivers of the heightened DO fluctuations in the FRB, with the minimum temperature being the most significant influencing factor, contributing 52.7 %; (3) The seasonal effects of temperature extremes on DO fluctuations were characterised by more complex mechanisms during the summer high-temperature period. The study reveals that temperature extremes in the CTZs exacerbate seasonal fluctuations in DO and provide scientific references for water quality management in response to climate change.