Fluvial sediment changes in cold regions are key research focus under global warming. However, the understanding in glacier basins remains constrained by sparse observations. Accurate turbidity–suspended sediment concentration (SSC) relationships are essential but challenging for real-time monitoring of SSC, especially in basins supplied by multi-source runoff. In this study, in-situ turbidity monitoring and SSC sampling were conducted at two cross sections, the Kuoqionggangri glacier terminus and the basin outlet, respectively. The main controls on turbidity–SSC relationships were examined via laboratory mixing experiments. Contrasts in sediment grain size distributions, shaped by varied sediment source zones during periods dominated by snowmelt, glacier melt, and rainfall, respectively, are the primary driver of turbidity–SSC relationship variability. Relative to a full-period curve, period-specific curves improved the fit markedly, with R² increasing from 0.86 to 0.98. However, direct extrapolation of low-turbidity curves (<1400 NTU) to the medium range (1400~3000 NTU) performed poorly (NSE=-0.67~0.11). In contrast, extrapolation of grain-size composition achieved much higher accuracy (NSE=0.91~0.98). A grain-size informed framework was therefore developed to extend turbidity–SSC relationships by tracking systematic changes in grain-size fractions. SSC estimates for the medium-turbidity range derived using this framework achieved NSE values of 0.96~0.98, allowing the relationships to be extended from the sampled range (<2840 NTU) to an unsampled high-turbidity range (<5580 NTU). The framework provides a feasible approach for deriving high temporal resolution SSC from turbidity records in remote glacier basins and improves process-based understanding of sediment dynamics in high altitude or latitude rivers.