This paper evaluates the predictive power of nine bedload equations, comparing the results provided by the equations with the bedload rates obtained in a previous field-based tracer experiment accomplished in River Pigüeña and River Coto, two coarse bed streams from NW Spain. Rivers from NW Spain draining the northern watershed of the Cantabrian Mountain range flow into the Bay of Biscay in a short path (50–60km). In this region, they are developed forested catchments featured by fluvial networks with relatively steep slopes, single-thread sinuous channels, and where bed sediment is typically coarse (cobble and gravel).Tagged stones were used to trace bed sediment movement during flood events in River Pigüeña and River Coto, the two main tributaries of the Narcea River basin. With the tracer results, bedload transport rates between 0.2 and 4.0kg/s were estimated for six flood episodes.The tracer-based bedload discharges were compared with the bedload rates estimated with the bedload formulae (DuBoys–Straub, Schoklitsch, Meyer Peter–Müller, Bagbold, Einstein, Parker–Klingeman–McLean, Parker–Klingeman, Parker and Wilcock-Crowe). Our assessment shows that all of the bedload equations tend to overestimate when compared with the tracer-based results, with the Wilcock and Crowe (2003) equation the only exception in River Pigüeña.We linked these results to the particular geomorphology of coarse-bed rivers in humid and forested mountain environments. Within these rivers, armored textures and structural arrangements in the bed are ubiquitous; these features, together with a low sediment supply coming from upstream forested reaches, define a supply-limited condition for these channels limiting the potential use of bedload equations. The Wilcock and Crowe (2003) equation introduces complex corrections into the ‘hiding function’, and this could explain why it performs better.