The three numbers in that scaling do not move together, and the mismatch is the interesting part. A tenfold increase in aquatic surface area came with about 35 times longer residence time, but only 1.8 times more emission per unit of carbon exported downstream [4]. Run the elasticity: the emission-to-export ratio rises as roughly the 0.17 power of residence time [17]. Measured against surface area, residence scales near area^1.5 while the partitioning scales near area^0.26 [15][16]. Lakes are a strong lever on how long water sits and a weak one on where its carbon ends up.
That distinction matters for how the finding gets quoted. The 1.8 is a ratio of two fluxes, emission over downstream export, not an absolute emission total, so the phys.org opening line that networks with ten times more lake area had almost twice the carbon emissions [13] is looser than the result it is describing [4]. In this dataset, streams produced more emissions overall [6]. A lake-rich network is not a bigger emitter so much as a leakier conduit.
Nobody can randomise a catchment's lake fraction, so the comparison across the 32 networks is observational, and the four sampling campaigns [3] are doing the work a control would otherwise do. The ratio was lowest in spring and autumn and highest in summer, when flows were low and water warm [7], and the lake-area relationship persisted through every period sampled, including an unusually dry summer [8]. A cross-sectional correlation that survives that much hydrological variation is worth considerably more than one measured once.
The mechanism is inferred rather than isolated. Alriksson and colleagues propose that longer residence gives microbes more time to break dissolved organic carbon down into carbon dioxide that then escapes the water [9], and they state plainly that they did not measure all the responsible processes and cannot say how much of the pattern comes from that breakdown rather than from other differences between lake-rich and lake-poor networks [10]. That is the honest ceiling on the causal claim.
What survives the ceiling is still usable. If almost all of a network's residence time is lake residence time [5], then an inventory that multiplies lake area by one per-area flux and stream area by another carries no term for where those lakes sit relative to the water moving through them, which is the argument the authors make for treating the two as a single connected network [12]. Note what that does not license. A ratio constrains the fate of carbon, not the size of any existing total, so nothing here moves a published flux number up or down. And one destination stays uncounted, because lakes also bury carbon in their sediments [11].
Residence time inside a single boreal region spanned roughly 360,000-fold, from 42 minutes to 29 years [14][5]. Any estimate of how much terrestrially derived carbon reaches the ocean instead of the atmosphere [18] is sitting on that spread.
