Spatial patterns in CO2 evasion from the global river network

CO2 evasion from rivers (FCO2) is an important component of the global carbon budget. Here we present the first global maps of CO2 partial pressures (pCO(2)) in rivers of stream orders 3 and higher and the resulting FCO2 at 0.5 degrees resolution constructed with a statistical model. A geographic information system based approach is used to derive a pCO(2) prediction function trained on data from 1182 sampling locations. While data from Asia and Africa are scarce and the training data set is dominated by sampling locations from the Americas, Europe, and Australia, the sampling locations cover the full spectrum from high to low latitudes. The predictors of pCO(2) are net primary production, population density, and slope gradient within the river catchment as well as mean air temperature at the sampling location (r(2)=0.47). The predicted pCO(2) map was then combined with spatially explicit estimates of stream surface area A(river) and gas exchange velocity k calculated from published empirical equations and data sets to derive the FCO2 map. Using Monte Carlo simulations, we assessed the uncertainties of our estimates. At the global scale, we estimate an average river pCO(2) of 2400 (2019-2826) mu atm and a FCO2 of 650 (483-846) Tg C yr(-1) (5th and 95th percentiles of confidence interval). Our global CO2 evasion is substantially lower than the recent estimate of 1800 Tg C yr(-1) although the training set of pCO(2) is very similar in both studies, mainly due to lower tropical pCO(2) estimates in the present study. Our maps reveal strong latitudinal gradients in pCO(2), A(river), and FCO2. The zone between 10 degrees N and 10 degrees S contributes about half of the global CO2 evasion. Collection of pCO(2) data in this zone, in particular, for African and Southeast Asian rivers is a high priority to reduce uncertainty on FCO2.