Particle-size-coupled tracing (PSCT): A source-size classification framework for sediment source apportionment under particle sorting.

Reliable apportionment of sediment sources is an important prerequisite for understanding sediment dynamics and can support future assessment of sediment-associated non-point source pollution and catchment water-quality management. However, hydrodynamic particle sorting during erosion and transport disrupts tracer conservativeness, thereby reducing sediment source fingerprinting accuracy. To address this issue, we developed and evaluated a particle-size-coupled tracing (PSCT) framework that explicitly embeds particle-size classes into source classification to reconstruct source-sediment comparability under sorting stress. Using a representative headwater catchment on the Loess Plateau, we evaluated the PSCT framework with four-class (FC) and six-class (SC) schemes via artificial mixtures and rainfall simulations with known source contributions. Its performance was compared against the traditional classification (TC) scheme using geochemical elements (GE) and mid-infrared spectroscopy (MIR) as tracers. Results indicated that the TC scheme failed to account for source particle-size heterogeneity, yielding mean absolute errors (MAEs) of 29.3% for GE and 18.4% for MIR in artificial mixtures. In contrast, the PSCT framework substantially reduced sorting-induced bias, with the SC scheme performing best. It reduced MAEs to 6.8% for GE and 4.5% for MIR in artificial mixtures. Under rainfall simulations, the SC scheme maintained average MAE values below 10%, indicating strong robustness under near-natural hydrodynamic conditions. These findings demonstrate that explicitly coupling particle size with source classification effectively enhances tracer conservativeness, reduces sorting-induced bias, and improves sediment source apportionment reliability under dynamic erosion and transport conditions. The proposed framework provides a mechanistically grounded basis for more reliable sediment source apportionment and may support future applications aimed at quantifying sediment-associated pollutant pathways.