Abstract: To investigate the long-term evolution patterns of ecological impacts of flood disasters and their driving mechanisms, this study took a typical flood event in China in 2006 as a case study. Based on multi-source remote sensing and ecological datasets for the period 2003-2025, we constructed the time series of ecological deviation for five indicators, respectively: normalized difference vegetation index (NDVI), leaf area index (LAI), net ecosystem productivity (NEP), soil and water conservation (SWC), and water conservation (WC). Recovery trajectory fitting, recovery-state classification, XGBoost-SHAP analysis, and Bayesian network modeling were integrated to systematically analyze the temporal evolution, spatial heterogeneity, and transmission characteristics of post-flood ecological responses. The results showed that: Post-flood ecological recovery exhibited distinct stage-dependent characteristics, with NDVI, LAI, and SWC returning to their baseline levels approximately 2.58, 2.91, and 5.35 years after the flood, respectively, indicating a recovery pattern characterized by rapid structural restoration and gradual functional reconstruction. Ecological impacts displayed pronounced spatial heterogeneity. Recovery areas of NDVI, LAI, SWC, and NEP expanded continuously over time and showed varying degrees of overcompensation, whereas WC experienced degradation during the intermediate recovery stage and recovered in the later stage. Elevations, precipitation and temperature were identified as the dominant drivers of post-flood vegetation recovery. Clear threshold effects were detected, with optimal recovery occurring at precipitation levels of approximately 60 mm, elevations between 500 m and 1 000 m, and temperatures ranging from 15 ℃ to 22 ℃. Bayesian network analysis revealed a hierarchical correlation in post-flood ecological recovery, in which influences propagated gradually from vegetation structure to ecological functions. Overall, post-flood ecosystem recovery followed a hierarchical response pathway of environmental drivers-vegetation response-functional reconstruction, in which vegetation structure played a critical intermediary role linking environmental changes to ecosystem functional recovery. These findings can provide scientific support for post-disaster ecological restoration and watershed ecological risk management.