Surface reflectance prior uncertainty remains a major source of error in land aerosol optical depth (AOD) retrieval, particularly over complex surfaces and for newly launched satellite sensors lacking long-term observations for constructing stable surface priors. Here, we develop a hybrid AOD retrieval framework for the DQ-1 Wide Swath Imager (WSI) that integrates a static surface reflectance prior, machine-learning-based blue-band surface residual correction, and a physically based lookup-table (LUT) inversion. A multi-band static surface prior is first constructed from quality-controlled clear-sky observations under low-aerosol-loading conditions. Rather than directly predicting AOD, a random forest model is used to estimate the residual between the static and reference surface terms at 0.443 μm from the static blue-band prior, top-of-atmosphere reflectance, NDVI, viewing geometry, and elevation. The corrected blue-band surface term is then propagated through the conventional LUT retrieval, thereby retaining the  physical aerosol inversion framework while using machine learning only to mitigate surface-prior  mismatch. MCD19A2 AOD is used exclusively in the offline stage to support the construction of  LUT-derived reference surface terms and for independent cross-product spatial consistency assessment, whereas AERONET observations provide the primary ground-based validation. Relative to the static-prior LUT retrieval, residual correction reduces the AERONET-based RMSE from 0.256 to 0.107 and mean bias error from 0.187 to 0.059, while increasing the fraction of retrievals within the expected error envelope from 14.8 % to 55.6 %. Evaluation on a held-out scene further demonstrates that the model generalizes the relationship between static-prior error and the LUT-derived reference surface term, reducing the blue-band surface-term MAE from 0.0088 to 0.0044. On dates excluded from model training and internal validation, the retrieved DQ-1 WSI AOD also exhibits strong regional spatial consistency with MCD19A2 across 200,507 collocated pixels (R = 0.863, RMSE = 0.131, MAE = 0.093, MBE = 0.012, and regression slope = 0.99). These results indicate that the performance improvement primarily arises from alleviating blue-band surface-prior mismatch within the physical LUT inversion rather than from direct empirical fitting of AOD. The proposed framework therefore provides a physically interpretable strategy for improving land AOD retrieval from DQ-1 WSI over springtime complex surfaces and offers a transferable solution for new wide-swath sensors for which robust long-term surface  reflectance priors are not yet available.