Abstract: Against the backdrop of intensified climate change and human activities, hydrological processes in rivers primarily fed by snow and ice meltwater have become markedly non-stationary, rendering ecological flow thresholds derived under stationary-runoff assumptions inadequate for dynamic water resources management. Using the Buha River on the Qinghai–Tibet Plateau as a case study, this study applies change-point tests to the 1956—2020 runoff series to identify regime shifts and delineate runoff periods. Based on the segmented records, estimates monthly ecological flows using hydrological methods including the Tennant method, the 90% exceedance monthly mean flow method (Q90 method), and driest monthly average flow method in recent 10 years. In parallel, a hydraulic method is used to establish the stage–discharge relationship; ecological flows required during the migration period of Gymnocypris przewalskii (naked carp) are back-calculated from threshold water depths, and flow velocities corresponding to non-scouring and non-silting conditions are computed to verify riverbed stability. The results show that changes in multi-year mean discharge between periods lead to an overall increase in ecological-flow thresholds, with more pronounced differences during the main flood season. Compared with estimates based on the 1956—2000 flow series, the recommended ecological flows for 2001—2020 increase from 2.44m³/s to 4.88m³/s for January—April and December (Tennant method), from 8.61m³/s (hydraulic method) to 14.65m³/s for May—June (Tennant method), and from 3.27~30.66m³/s to 7.91~42.70m³/s for July—November (Q90 method). The threshold discharges at the study cross-section are 1.70m³/s and 8.61m³/s, and the riverbed remains in a relatively stable state under ecological-flow conditions during the migration period. This study provides a scientific basis for period-based, dynamic, and adaptive ecological-flow management and regulation in snow- and ice-melt-fed rivers discharging into lakes.