[17] Partial pressures of carbon dioxide (pCO_{2}, in *μ*atm = 10^{−6} atm) were calculated from published measurements of pH, total alkalinity (TA, in *μ*eq L^{−1}) or dissolved inorganic carbon (DIC, in *μ*mol L^{−1}), and water temperature (°C) at 0–1 m depth in the water column, using the program CO2SYS [*Lewis and Wallace*, 1998]. Freshwater dissociation constants were used for all lakes except Van and Great Salt [*Millero*, 1979], where salinity exceeded 15 p.p.t. so the seawater constants of *Roy et al.* [1993, 1994, 1996] were deemed more appropriate. Because the salinity of lakes Turkana and Issyk Kul falls between the salinity ranges appropriate for analysis with freshwater and seawater constants, some error is introduced to the pCO_{2} and gas exchange calculations for these lakes. The quality of published water chemistry data used to calculate pCO_{2} is variable, with assessments of data quality shown in auxiliary material Table S4. An exception to this was Lake Baikal, where pCO_{2} and flux values were published without a description of methods [*Kozhova and Izmest'eva*, 1998].

[18] Using calculated pCO_{2} values, we calculated rudimentary estimates of lake-atmosphere CO_{2} exchange as follows. The wind speed-gas exchange relationship of *Cole and Caraco* [1998] was used to estimate the CO_{2} gas transfer velocity for freshwater at a temperature of 20°C (i.e., *k*_{600}, cm h^{−1}):

where *U*_{10} is the wind speed (m s^{−1}) normalized to a height of 10 m above the water surface. Wind speed data were obtained from a variety of publications and other sources (listed in auxiliary material Table S4), including the National Data Buoy Center (NDBC) web site for the Laurentian Great Lakes. In order to convert from the *k*_{600} value to the gas transfer value at the observed temperature in each lake (*k*_{T}), we used the equation

where *Sc*_{T} is the Schmidt number at temperature *T* (°C) and the exponent *n* depends on water surface conditions and varies from −0.67 to 1 [*Cole and Caraco*, 1998]. Consistent with numerous other authors, we use a value of −0.5 for *n*, typical of a wavy water surface free of films, which can inhibit gas exchange [*Jähne et al.*, 1987]. The Schmidt number for CO_{2} in freshwater varies with temperature according to the following equation [*Wanninkhof*, 1992]:

Finally, lake-atmosphere exchange fluxes (F_{CO2}) were calculated using the flux equation,

where Δ*pCO*_{2} is the water-air pCO_{2} gradient (pCO_{2water} − pCO_{2air}, in10^{−6} atm) and *α* is CO_{2} solubility in freshwater (mol L^{−1} atm^{−1}), calculated as

where *T*_{K} is temperature in degrees Kelvin [*Weiss*, 1974], and converted to units of mol cm^{−3} atm^{−1} by dividing by 1000. Flux values were converted from mol C cm^{−2} h^{−1} to units of g C m^{−2} yr^{−1} by multiplying by the appropriate conversion factors. Air pCO_{2} values were not reported, so we used a value of 377.4 *μ*atm for 2004 and corrected it for the year of measurement by subtracting 1.4 *μ*atm per year prior to that datum. Finally, solubility was corrected for atmospheric pressure at the elevation of each lake, calculated using an approximation of the barometric formula

where *P*_{Z} and *P*_{SL} (atm) represent atmospheric pressure at height Z (m) above sea level (SL), and 8000 approximates the value of the temperature term and constants in the exponent at 273.15° K [cf. *Berberan-Santos et al.*, 1997].

[19] In all cases, pCO_{2} and wind data were used to calculate flux estimates at the highest resolution the data would afford (i.e., biweekly, monthly, or seasonal values where sufficient data existed) then combined by weighted averaging to generate an annual flux estimate. For most lakes, sufficiently resolved water chemistry and wind speed data do not exist at this juncture for the flux estimates to adequately reflect interseasonal or even intraseasonal variation in CO_{2} concentrations and fluxes. These flux estimates should be viewed as rudimentary.