Are missing nitrogen inputs in grasslands due to N2 fixation or sampling artifacts?

Carmella Vizza and Sarah Roley, School of the Environment, Washington State University, Richland, WA, Carmella Vizza, Stephen K. Hamilton and G Philip Robertson, W. K. Kellogg Biological Station, Michigan State University, Hickory Corners, MI, Di Liang, DOE Center for Advanced Bioenergy & Bioproducts Innovation, University of Illinois at Urbana-Champaign, Urbana, IL

Background/Question/Methods
Over the last 30 years, ecosystem mass balances for both forests and grasslands have documented nitrogen (N) deficits—where net N outputs, after being corrected for changes in N storage, exceed total inputs. The missing inputs could be unrecognized N2 fixation or a sampling artifact. Solving this decades-old puzzle could help reveal the ecosystem importance of non-symbiotic N2 fixation.
We conducted a meta-analysis for grassland N mass balances using 41 (of 1208 possible) data sources that contained plant and soil N data for at least two time points (and harvested N as appropriate), excluding agricultural systems with cereal crops or legumes. We calculated both a ratio (inputs/outputs) and a delta value (inputs – outputs). Ratios of less than 1 and negative delta values suggest missing N inputs or deficits.
Results/Conclusions
Preliminary results from 28 sources and 108 grassland budgets showed that input/output ratios (1.8 ± 2.5; mean ± SD) and delta values (88 ± 1571 kg N ha–1 yr–1) were widely variable. Forty-five percent of the budgets exhibited deficits. When we examined the effect of various inputs and outputs on both the ratios and delta values, change in soil N was the most important driver, yet the way soil N was measured (e.g., total versus mineral) did not significantly affect ratios, delta values, or even the change in soil N. We were able to predict input/output ratios well using a mixed-effects model where study was a random factor and study scale (mesocosm vs. field), grass life history (perennial vs. annual), N fertilizer amount, and change in soil N were fixed factors, which together explained 65% of the variation, while 21% was random. Field studies and perennial grasses were more likely to exhibit deficits than were mesocosm studies and annual grasses. Additionally, deficits were more likely to be found in unfertilized systems, which might be expected if missing inputs were due to N2 fixation. Because the input/output ratios were well-predicted by ecosystem characteristics and deficits were more prevalent in unfertilized perennial grasslands, we conclude that the observed deficits are not sampling artifacts and that unmeasured N2 fixation—most likely associative—is the logical source of missing N inputs in at least 45% of the grasslands examined.