Unexpected large evasion fluxes of carbon dioxide from turbulent streams draining the world’s mountains

Inland waters, including streams and rivers, are active components of the global carbon cycle. Despite the large areal extent of the world’s mountains, the role of mountain streams for global carbon fluxes remains elusive. Using recent insights from gas exchange in turbulent streams, we found that areal CO2 evasion fluxes from mountain streams equal or exceed those reported from tropical and boreal streams, typically regarded as hotspots of aquatic carbon fluxes. At the regional scale of the Swiss Alps, we present evidence that emitted CO2 derives from lithogenic and biogenic sources within the catchment and delivered by the groundwater to the streams. At a global scale, we estimate the CO2 evasion from mountain streams to 167 ± 1.5 Tg C yr−1, which is high given their relatively low areal contribution to the global stream and river networks. Our findings shed new light on mountain streams for global carbon fluxes.


Supplementary Note 1
show that small streams (catchment sizes <30 km 2 ) have high geochemical variability but are skewed towards the carbonate end-member. This it related to short residence times in small headwater catchments, which favors fast weathered bedrock, such as carbonate bedrock weathering, rather than silicate weathering 1 . We repeated the analysis from Marx and colleagues using the same database, the Global River Chemistry database (GLORICH) 2 , together with data from our Swiss monitoring stations. The geochemical characterization shows that the Swiss sites are indeed influenced by carbonate bedrock weathering (Supplementary Figure 4).

Supplementary Note 2
We evaluated differences between geopredictors for Swiss streams derived from data sets with different resolutions (highly resolved Swiss dataset and lower resolved globally available dataset; Supplementary Figure 6) to be confident that the stream channel slopes (Supplementary Figure 6A) stream altitude (Supplementary Figure 6B) and stream discharge (Supplementary Figure 6C) did not deviate considerably depending on data resolution. The distributions of the data were similar for the two datasets, although slope was somewhat lower in the high-resolution data set (median 0.039 m m -1 , CI: 0.004 and 0.176 m m -1 ) compared to the low-resolution data set (median 0.055 m m -1 , CI: 0.004 and 0.222 m m -1 ). Altitude was slightly higher in the high-resolution data set (median 902 m, CI: 415 and 2185 m) compared to the low-resolution data set (median 834 m, CI: 406 and 2141 m). Discharge was similar between the high-resolution data set (median 0.38 m 3 s -1 , CI: 0.09 and 1.77 m 3 s -1 ) and the low-resolution data set (median 0.38 m 3 s -1 , CI: 0.12 and 1.82 m 3 s -1 ).
Moreover, for all our monitoring stations in the Swiss Aps (Supplementary Figure 1) we measured stream channel slope every 10 meters' distance in the field using a dGPS system. We then compared the difference in stream channel slopes calculated for whole stream reaches with channel slopes calculated from segments of 10 m in length each. We found that in average the slope is underestimated by 0.022 m m -1 when using mean reach slope and not considering the variations in slopes along a stream reach. The maximum slope measured along the reach (10 m sub-reaches) deviated significantly from the mean reach slopes, and was up to 0.60 m m -1 higher. Across all 12 catchments, the maximum slope was 0.42 m m -1 higher compared to the reach slope. This suggests that the kCO2 values may be even higher than estimated in this study for Switzerland (median 86.4 m d -1 , CI: 6.0 and 461.9 m d -1 ) and for mountain streams worldwide (median 25.6 m d -1 , CI: 3.5 and 410.6 m d -1 ). Predicted streamwater pCO2 was similar in Swiss streams (median 705 µatm, CI: 380 and 1224 µatm) and at the global extent (median 737 µatm, CI: 317 and 1644 µatm). 10.8% of the Swiss mountain streams, and the same proportion of the mountain streams worldwide has negative ΔCO2 meaning that they fall below atmospheric saturation. The median areal CO2 fluxes are higher from Swiss mountain streams (median 3.6 kg C m -2 yr -1 , CI: -0.5 and 23.5 kg C m -2 yr -1 ) compared to mountain streams worldwide (median 1.1 kg C m -2 yr -1 , CI: -0.5 and 32.1 kg C m -2 yr -1 ) (Supplementary Figure 7).

Supplementary Note 3
Estimating CO2 fluxes based on mean or median values of CO2 and discharge, instead of highly resolved data, may induce errors due to the temporal variability. We compared CO2 fluxes derived from measured data at 10-minute time steps with CO2 fluxes predicted using our CO2 prediction model combined with Q from GloRiC. Due to data availability, this analysis was possible at 7 of our 12 high-altitude Alpine monitoring stations. Despite high temporal variability in CO2 fluxes, we found median areal fluxes at the monitoring stations corresponding relatively well to the areal fluxes that we predicted in our study where the slope between FCO2 predicted by the model and FCO2 calculated from 10-minute time series was -0.921 ± 0.284 (R 2 = 0.68, n = 7, P = 0.0022) (Supplementary Figure 9).

Supplementary Figures
Supplementary Figure 1. We continuously monitored pCO2 in 12 streams located in 4 Alpine catchments in Switzerland. Our 12 mountain stream monitoring stations measured streamwater pCO2 levels close to saturation throughout the year (median pCO2 397 to 673 µatm; Supplementary Table 1).

Supplementary Figure 2. Data input of the CO2 model.
Sampling locations for the mountain stream CO2 data used for the prediction model (A). Shown are also density distributions of elevation (B), discharge (C) and soil organic carbon (SOC) (D), used as input parameters for the prediction of streamwater CO2 concentrations (E). Observed versus predicted CO2 followed the 1:1 line (blue) and fell within the 95% prediction confidence intervals (dashed blue lines) except for at very high CO2 concentrations were CO2 was underpredicted.