Bed material abrasion is a major control on the partitioning of basin-scale sediment fluxes between coarse and fine material. While abrasion is traditionally treated as an exponential function of transport distance and a lithology-specific abrasion coefficient, experimental studies have demonstrated greater complexity in the abrasion process: the rate of abrasion varies with clast angularity, transport rate, and grain size. Yet, few studies have attempted to assess the importance of these complexities in a field setting. Here, we develop a new method for rapidly quantifying baseline abrasion rate in the field via Schmidt Hammer Rock Strength (SHRS). We use this method, along with measurements of gravel bar lithology, to quantify abrasion in the Suiattle River, a basin in the North Cascades of Washington State dominated by a single coarse sediment source: recurrent, debris flows from a tributary draining Glacier Peak stratovolcano. Rapid downstream strengthening of river bar sediment and a preferential loss of weak, low-density vesicular volcanic clasts relative to non-vesicular ones suggest that abrasion is extremely effective in this system. The standard exponential model for downstream abrasion fails to reproduce observed downstream patterns in lithology and clast strength in the Suiattle, despite accounting for the heterogeneity of source material strength and systematic underestimate of abrasion rates by tumbler experiments. Incorporating transport-dependent abrasion into our model largely resolves this failure. While a simplified approach to characterizing abrasion is tempting, our findings show that sediment heterogeneity and transport-dependent abrasion are important controls on the downstream fate of coarse sediment in fluvial systems.