Abrasion in a concrete-lined sediment bypass tunnel is estimated using a Japanese state-of-the-art prediction model and validated by measured invert abrasion data at Asahi Reservoir, Japan. The model is described in detail, certain shortcomings are disclosed, and a revised version is proposed. The model consists of a kinetic energy term accounting for the impact by saltating particles, and a friction work term accounting for the grinding stress. It is found that the latter term yields concrete abrasion values being consistently a multiple compared to its kinetic term contradicting other research. Based on that, and a possible particle impact angle inconsistency, it is proposed to omit the friction work term. It is shown that the calculated abrasion is overestimated by 138% on average compared with that measured, if both terms are accounted for. However, promising results are obtained with only 30% overestimation by neglecting the friction work term.