CO2 elicits long-term decline in nitrogen fixation.

Rising atmospheric carbon dioxide (Ca), a product of fossil fuel burning, land-use change, and cement manufacture, is expected to cause a large carbon sink in land ecosystems, partly mitigating human-driven climate change (1). Increasing biological nitrogen fixation with rising Ca has been invoked as a means to provide the N necessary to support C accumulation (2). As in many short-term experiments (3), we found that Ca enrichment increased N fixation during the first year of treatment in an oak woodland. However, the effect declined and disappeared by the third year. Ca enrichment consistently depressed N fixation during the 5th, 6th, and 7th years of treatment. Reduced availability of the micronutrient molybdenum, a key constituent of nitrogenase, best explains this reduction in N fixation. Our results demonstrate how multiple element interactions can influence ecosystem responses to atmospheric change and caution against expecting increased biological N fixation to fuel terrestrial C accumulation. We investigated the effects of elevated Ca on legume N fixation in a stand of scrub-oak vegetation in central coastal Florida, where the leguminous vine Galactia elliottii Nutt. occurs naturally (4). During the first year of the experiment, elevated Ca nearly doubled N fixation by G. elliottii (Fig. 1A), but this effect disappeared and later reversed (Ca time interaction, P 0.001). Elevated Ca significantly depressed N fixation by G. elliottii during the last 3 years of CO2 exposure (Fig. 1A, repeated measures analysis of variance for 2000 to 2002 only, P 0.049). During the 7-year period, the relative effect of elevated Ca on N fixation by G. elliottii declined exponentially (Fig. 1A). N-fixing plants are sensitive to light availability (5), so increased shading through greater leaf area could have been responsible for the declining response of G. elliottii N fixation to elevated Ca. However, N-fixing plants also require relatively high concentrations of other nutrients such as phosphorus, iron, and Mo (5). Elevated Ca increased nutrient accumulation in oak biomass and in organic forms in the soil (4 ), potentially reducing their availability to G. elliottii. Elevated Ca did not alter foliar P in G. elliottii (1.19 0.05 mg g 1 for ambient Ca versus 1.08 0.12 mg g 1 for elevated Ca, P 0.397), but it substantially decreased foliar concentrations of both Mo and Fe (Fig. 1B). This pattern may reflect reduced Mo and Fe concentrations in G. elliottii nodules, where N fixation occurs. Neither leaf area index (r – 0.192, P 0.476) nor foliar Fe (r 0.024, P 0.843) was significantly correlated with N fixation by G. elliottii, whereas foliar Mo showed a positive, significant relationship (Fig. 1C). Limitation of N fixation by light and nutrients could both occur, but we submit that the regression analysis provides correlative evidence in favor of the latter and specifically implicates limitation by the availability of Mo. Mo deficiency in Nfixing plants has been documented, particularly in sandy acidic soils (6 ) similar to the scrub-oak soil studied here (4 ). By causing soil pH to decline or organic matter to accumulate (both of which increase Mo adsorption to soil particles), elevated Ca could reduce the availability of Mo, causing a systematic decline in N fixation (Fig. 1A). We found that experimental Ca exposure initially increased but later suppressed N fixation, contrary to the expected response (2, 3). Biogeochemical feedbacks can substantially modify ecosystem responses to global change, and our results underscore the need to broaden the suite of elements typically considered when examining these feedbacks.