<p>Streamflow prediction is essential for sustainable water resource management, supporting ecosystems, human livelihoods, and disaster resilience. However, characterizing rainfall-runoff predictability remains a critical challenge in hydrology, particularly under intensifying human interventions. This study aims to characterize rainfall-runoff predictability under the influence of human interventions by a novel integration of chaos theory and information entropy. The methodology employs phase space reconstruction and Largest Lyapunov Exponent (LLE) to quantify chaotic dynamics and predictability in rainfall and streamflow, alongside transfer entropy (TE) to evaluate directional information transfer between these variables. Key results reveal distinct spatial patterns: streamflow predictability deteriorates systematically from upstream to downstream, whereas rainfall predictability exhibits uniformly chaotic across the basin (LLE: 0.33 ~ 0.67). Information entropy analysis demonstrates stronger dynamical connectivity in upstream regions (TE: 0.45 ~ 0.74), highlighting tighter rainfall-runoff coupling. Conversely, downstream areas experience disrupted information transfer due to reservoir operations, with Reservoir Index (RI) values exceeding the critical threshold (e.g., Daheiting: RI = 0.653), leading to a 60 to 78% decline in TE compared to upstream systems. Additionally, streamflow complexity correlates positively with rainfall variability and potential evapotranspiration (<i>ET</i>_p) but negatively with the aridity index. Demonstrated through a case study of the Luanhe River Basin, this framework underscores how reservoirs diminish hydrological predictability and alter information pathways. By integrating chaos theory with entropy-based metrics, it offers a robust approach for diagnosing predictability loss in human-impacted river basins and yields valuable insights for adaptive water management.</p>