The real-world effect of accumulated surface roughness on wind-turbine power production is not well understood. To isolate specific blade roughness features and test their effect, field measurements of turbine-blade roughness were made and simulated on a NACA 633-418 airfoil in a wind tunnel. Insect roughness, paint chips, and erosion were characterized then manufactured. In the tests, these roughness configurations were recreated as distributed roughness, a forward-facing step, and an eroded leading edge. Distributed roughness was tested in three heights and five densities. Chord Reynolds number was varied between 0:8 to 4:8 � 10<
sup>
6<
/sup>
. Measurements included lift, drag, pitching moment, and boundary-layer transition location. Results indicate minimal effect from paint-chip roughness. As distributed roughness height and density increase, the lift-curve slope, maximum lift, and lift-to-drag ratio decrease. As Reynolds number increases, natural transition is replaced by bypass transition. The critical roughness Reynolds number varies between 178 to 318, within the historical range. At a chord Reynolds number of 3:2 � 10<
sup>
6<
/sup>
, the maximum lift-to-drag ratio decreases 40% for 140 ?m roughness, corresponding to a 2.3% loss in annual energy production. Simulated performance loss compares well to measured performance loss of an in-service wind turbine.