Evasion of gaseous carbon (C) from streams is often poorly quantified in landscape C budgets. Even though the potential importance of the capillary network of streams as C conduits across the land-water-atmosphere interfaces is sometimes mentioned, low-order streams are often left out of budget estimates due to being poorly characterized in terms of gas exchange and even areal surface coverage. We show that evasion of C is greater than all the total dissolved C (both organic and inorganic) exported downstream in the waters of a boreal landscape. In this study evasion of carbon dioxide (CO2 ) from running waters within a 67 km(2) boreal catchment was studied. During a 4 year period (2006-2009) 13 streams were sampled on 104 different occasions for dissolved inorganic carbon (DIC) and dissolved organic carbon (DOC). From a locally determined model of gas exchange properties, we estimated the daily CO2 evasion with a high-resolution (5 × 5 m) grid-based stream evasion model comprising the entire ~100 km stream network. Despite the low areal coverage of stream surface, the evasion of CO2 from the stream network constituted 53% (5.0 (±1.8) g C m(-2)  yr(-1) ) of the entire stream C flux (9.6 (±2.4) g C m(-2)  yr(-1) ) (lateral as DIC, DOC, and vertical as CO2 ). In addition, 72% of the total CO2 loss took place already in the first- and second-order streams. This study demonstrates the importance of including CO2 evasion from low-order boreal streams into landscape C budgets as it more than doubled the magnitude of the aquatic conduit for C from this landscape. Neglecting this term will consequently result in an overestimation of the terrestrial C sink strength in the boreal landscape.

[1] Boreal streams represent potentially important conduits for the exchange of carbon dioxide (CO 2 ) between terrestrial ecosystems and the atmosphere. The gas transfer coefficient of CO 2 (K CO2 ) is a key variable in estimating this source strength, but the scarcity of measured values in lotic systems creates a risk of incorrect flux estimates even when stream gas concentrations are well known. This study used 114 independent measurements of K CO2 from 14 stream reaches in a boreal headwater system to determine and predict spatiotemporal variability in K CO2 . The K CO2 values ranged from 0.001 to 0.207 min −1 across the 14 sites. Median K CO2 for a specific site was positively correlated with the slope of the stream reach, with higher gas transfer coefficients occurring in steeper stream sections. Combining slope with a width/depth index of the stream reach explained 83% of the spatial variability in K CO2 . Temporal variability was more difficult to predict and was strongly site specific. Variation in K CO2 , rather than pCO 2 , was the main determinant of stream CO 2 evasion. Applying published generalized gas transfer velocities produced an error of up to 100% in median instantaneous evasion rates compared to the use of actual measured K CO2 values from our field study. Using the significant relationship to local slope, the median K CO2 was predicted for 300,000 km of watercourses (ranging in stream order 1-4) in the forested landscape of boreal/nemoral Sweden. The range in modeled stream order specific median K CO2 was 0.017-0.028 min −1 and there was a clear gradient of increasing K CO2 with lower stream order. We conclude that accurate regional scale estimates of CO 2 evasion fluxes from running waters are possible, but require a good understanding of gas exchange at the water surface.

Understanding of the processes that control CO 2 concentrations in the aquatic environment has been hampered by the absence of a direct method to make continuous measurements over both short-and long-term time intervals. We describe an in situ method in which a non-dispersive infrared (NDIR) sensor is enclosed in a water impermeable, gas permeable polytetrafluoroethylene (PTFE) membrane and deployed in a freshwater environment. This allows measurements of CO 2 concentration to be made directly at a specific depth in the water column without the need for pumps or reagents. We demonstrate the potential of the method using examples from different aquatic environments characterized by a range of CO 2 concentrations (0Ð5-8Ð0 mg CO 2 -C l 1 , equivalent to ca 40-650 µmol CO 2 l 1 ). These comprise streams and ponds from tropical, temperate and boreal regions. Data derived from the sensor was compared with direct measurements of CO 2 concentrations using headspace analysis. Sensor performance following long-term (>6 months) field deployment conformed to manufacturers' specifications, with no drift detected. We conclude that the sensor-based method is a robust, accurate and responsive method, with a wide range of potential applications, particularly when combined with other in situ sensor-based measurements of related variables.

No abstract