Surface waters contribute substantially to carbon dioxide (CO2) emissions to the atmosphere. However, global estimates remain uncertain due to methodological difficulties, such as in precisely estimating gas transfer in steep upland streams. Here, we addressed the question of what drives CO2 evasion from steep mountainous stream network of the European Alps by assessing the spatial and temporal variation of partial pressure of CO2 (pCO2) for 148 streams and the gas transfer coefficient for CO2 (kCO2) for 88 locations within this 254 km2 watershed. Results show that log kCO2 can be predicted reasonably well (r2 = 0.71, p<0.001, n = 88) using a statistical model based on slope, average width, flow velocity and stream discharge. Also, most sites were supersaturated in CO2 with significant variation in pCO2 due to season (September vs. December) and time of day (day vs. night), but not stream order. Resulting median CO2 evasion rates were 145, 119, 46, 43, and 50 mg C m−2 h−1 at 1st to 5th order streams, respectively. CO2 evasion was dependent on season and time of day, with the highest evasion (184.0 kg C h−1) during growing season at nighttime, followed by 124.6 kg C h−1 during daytime. Dormant season nighttime evasion was 30.9 kg C h−1 and daytime evasion only 17.1 kg C h−1. Overall we conclude that CO2 evasion of steep mountainous streams depends on seasonal and diurnal variation in pCO2 and reach-specific variability in kCO2. These controls strongly alter landscape-scale CO2 evasion estimates, with implications for regional to global carbon budgets.