Air-water gas exchange governs fluxes of gas into and out of aquatic ecosystems. Knowing this flux is necessary to calculate gas budgets (i.e., O2) to estimate whole-ecosystem metabolism and basin-scale carbon budgets. Empirical data on rates of gas exchange for streams, estuaries, and oceans are readily available. However, there are few data from large rivers and no data from whitewater rapids. We measured gas transfer velocity in the Colorado River, Grand Canyon, as decline in O2 saturation deficit, 7 times in a 28-km segment spanning 7 rapids. The O2 saturation deficit exists because of hypolimnetic discharge from Glen Canyon Dam, located 25 km upriver from Lees Ferry. Gas transfer velocity (k600) increased with slope of the immediate reach. k600 was < 10 cm h-1 in flat reaches, while k600 for the steepest rapid ranged 3600-7700 cm h-1, an extremely high value of k600. Using the rate of gas exchange per unit length of water surface elevation (Kdrop, m-1), segment-integrated k600 varied between 74 and 101 cm h-1. Using Kdrop we scaled k600 to the remainder of the Colorado River in Grand Canyon. At the scale corresponding to the segment length where 80% of the O2 exchanged with the atmosphere (mean length = 26.1 km), k600 varied 4.5-fold between 56 and 272 cm h-1 with a mean of 113 cm h-1. Gas transfer velocity for the Colorado River was higher than those from other aquatic ecosystems because of large rapids. Our approach of scaling k600 based on Kdrop allows comparing gas transfer velocity across rivers with spatially heterogeneous morphology.