Isolating the effect of subterranean ventilation on CO2 emissions from drylands to the atmosphere

Recent studies show the importance of CO2 exchange processes by subterranean ventilation, not directly related to concurrent biological activity, in net ecosystem exchanges between drylands and the atmosphere. We present seven years of eddy covariance measurements over a carbonate ecosystem in southern Spain and offer a practical, empirical approach that isolates such processes. Net fluxes represent emissions by ecosystem respiration and subterranean ventilation (Vs), reduced by the gross photosynthetic flux (Fc = −Fgp + Reco + Vs). Large daytime emissions due to decreases in subterranean storage (Ss < 0) – destocking subterranean CO2 (−Ss = Vs) – are hypothesized as the dominant process decoupling Fc from biological sources (Reco) and sinks (Fgp) during dry periods. Since the latter are regulated by stomatal conductance (gs), and can be characterized in terms of evapotranspiration (E), our approach exploits the extreme seasonality of gas exchanges in such ecosystems to disentangle the role of Vs in eddy fluxes. Diurnal trends in Fc and gs were strongly linked during growing periods. During dry periods however, daytime CO2 emissions greatly exceeding modelled Reco when gs was low, precluded physiological interpretation of the negative quantity (Reco − Fc) as Fgp. Since physiological inhibition of photosynthesis by drought coincided with large CO2 emissions, Vs was isolated as Fc − Reco. Analysis of different environmental factors showed that wind speed is most strongly correlated with Vs over the dataset. Although the ecosystem under study was nearly carbon neutral annually (Reco ≈ Fgp + Ss; all respired CO2 is either offset by photosynthetic uptake or stored underground), during long periods with dry soils and wind, Vs represented up to 62% of annual emissions. Once the effects of subterranean ventilation are considered, net exchanges correspond credibly to a net flux due to concurrent biological processes that can be better decomposed into Fgp and Reco. These findings suggest that the flux-partitioning and gap-filling models used by the FLUXNET community require adaptation to explicitly account for such processes in drylands.