Each increment was sectioned longitudinally into two halves. One half of the soil was homogenized, large roots and green mosses were removed, and a sub-sample of soil was extracted with 2 M KCl to determine the initial soil depth profile of plant available NH4-N and NO3-N . Nutrient concentrations were standardized per unit dry soil mass using gravimetric water conversion from samples oven dried for more than 48 h at 70°C. The distribution of total carbon and N throughout the soil profile was determined by elemental analysis on similarly collected and processed soil cores collected in 2012 from the nearby NGEE Arctic Intensive site [Iversen et al., 2015b]. Species- specific below ground biomass was determined from the remaining half of the soil from each depth increment. Living roots and rhizomes associated with the target species in each plot were removed from each depth increment using forceps, oven dried at 70°C for more than 48 h, and weighed to determine species specific living below ground biomass, which was converted to unit ground area using the bulk density of the core depth increments. After the initial harvest, we injected a solution of 15NH4Cl in the soil beneath newly located 9 cm ff 9 cm plots in homogenous species patches. We separately targeted three soil horizons for each species: organic, shallow mineral, or the permafrost boundary . Injections were made in a grid pattern of 16 points per plot to ensure homogenous delivery of the tracer solution at a given soil depth. The organic horizon injections were targeted at 3 cm depth, the mineral horizon injections were targeted at 10 cm depth,vertical grow rack and the permafrost boundary injections were targeted at 1 cm above the permafrost boundary .

One week later, the vegetation in the labeled plots was clipped to the moss surface, and soil cores were taken in the center of each labeled plot. Above ground and below ground vegetation was processed and quantified as above, and oven-dried, ground plant tissues were analyzed for 15N using continuous-flow isotope ratio mass spectrometry . Duplicate samples and standards of known 15N concentration were used to ensure the precision and accuracy of the data. The field experiment provides edaphic and vegetation data collected prior to 15N addition to initialize and drive the N-COM model. Then, the total amount of 15N acquired by the plants after the experimental tracer addition was used to test the predictions from the three nutrient competition concepts . The major difference among the three competition models is that only ECA explicitly considers essential root traits for plant-microbe competition. Therefore, first, a comparison between the ECA model and the other two models will inform how root traits control plant-microbe competition. Second, the 15N tracer experiment quantififies the vertical distribution of plant N uptake, which is an emergent pattern of plant-microbe competition. By comparing model predictions with different plantmicrobe competition hypotheses with the observations, we can evaluate how plant-microbe competition hypotheses affect plant N uptake. Third, since Relative Demand and Microbes Win competition hypothesesare widely used by prevailing ESMs, the discrepancy between these two concepts and observations can inform future modeling efforts.N-COM is a process-based model originally developed to represent coupled ecosystem carbon, nitrogen, and phosphorus cycles [Zhu and Riley, 2015; Zhu et al., 2016] based on Equilibrium Chemistry Approximation kinetics [Tang and Riley, 2013], although its structure is sufficiently generic to include any number of substrates and consumers.

The modeling framework mechanistically represents nutrient competition assuming plants and microbes produce specialized nutrient transporter enzymes to react with soil inorganic nitrogen substrates, enzyme-substrate complexes are then formed, these complexes can be transported into cells, and finally, the transporter enzymes are liberated [Button, 1985; Williams and Miller, 2001]. Thus, the binding of substrates to plant nutrient transporter enzymes inhibits the binding between substrate and microbial nutrient transporter enzymes and vice versa. While nutrient diffusivity limitation may constrain plant uptake by affecting substrate affinity [Tang and Riley, 2013], we did not consider diffusivity limitation in this study because 15N was directly added in the rooting zone and the spatial scale of the plots were just a few centimeters. As applied here, N-COM quantifies tundra C and N fluxes in three model layers: organic layer , mineral layer , and near the permafrost boundary layer for Carex aquatilis, Eriophorum angustifolium, and Salix rotundifolia . We focus here on plant 15NH4 + uptake; other nutrient uptake fluxes are described in Zhu et al. [2016]. Competition for NH4 + occurs among roots, nitrifiers, and microbial decomposers. However, nitrifier activity is typically very small in tundra soils [Giblin et al., 1991; Schimel et al., 1996]. We therefore assumed in this study that competition only occurred between roots and microbial decomposers. The model does not represent microbial community and diversity . Since different microbial functional groups may have different enzymatic kinetics, we also assessed uncertainties stemming from this model simplification in our uncertainty analysis .Most model parameters were directly taken from field data . However, several key model parameters were not measured in the field experiment due to logistical constraints, and our derivation of those parameters from literature may have introduced uncertainties.

We quantified the uncertainty associated with the unobserved parameters using a Monte Carlo approach. For ECA model simulations, VMAX and KM were randomly sampled from their observed ranges 500 times for each plant species and for soil microbes. For the Relative Demand and Microbes Win competition models, which do not use kinetics parameters, uncertainty stemmed from the estimate of total plant N demand calculated by dividing plant net biomass production by its C to N ratio. Similarly, we randomly sampled plant net biomass production 500 times from the literature reported range for each species. Uncertainty ranges associated with the above mentioned parameters are reported as error bars in results and figures.The three dominant tundra plant species observed here were dramatically different in terms of their maximum rooting depth, vertical rooting profile , and their prescribed root NH4 + uptake kinetics . However, the distribution of soil NH4 + throughout the soil profile was relatively similar across the three plant monocultures , being higher in the organic layer and near the permafrost boundary , and lower in mineral soil layers . Carex aquatilis is a relatively deep-rooting species . Moreover, it had the highest total root density of the species, most of which was in the organic and surface mineral layers . Eriophorum angustifolium is also a deep-rooting species . However, its root density was much lower compared with Carex aquatilis. Salix rotundifolia also has a relatively high density of roots, of which the vast majority are in the upper 10 cm of soil and ~90% are in the organic layer . This pattern indicates that Salix rotundifolia lacks the ability to access deep soil nutrients and must directly compete with microbial decomposers in surface soils. The 15N tracer data show that Carex aquatilis took up most of its NH4 + from mineral soil layers,vertical planting tower which agrees with its root density profile . Similar to Carex aquatilis, Eriophorum angustifolium took up most of its NH4 + from mineral soil layers . However, the observed uptake profile was in contrast to its root density profile . Although roughly equalamounts of roots existed in the organic layer and mineral layer , mineral layer NH4 + uptake was tenfold higher than in the organic layer. Salix rotundifolia took up 80% of NH4 + from the organic layer . Our modeling of plant-microbe competition using the ECA approach in N-COM generally captured the observed vertical patterns of plant N uptake for all three tundra species . Differences between ECA predictions and observations were small: for Carex aquatilis, ECA overestimated plant N uptake near the permafrost boundary layer and for Eriophorum angustifolium, ECA overestimated plant N uptake at organic layer . Our uncertainty analysis showed that plant-microbe competition was sensitive to the choice of kinetics parameters . However, the plant N uptake patterns were conservative even when considering the full range of uncertainties in derived kinetics parameters. Therefore, we conclude that using literature-derived kinetics parameters introduced modest uncertainty in our model analysis but did not undermine the fidelity of the ECA approach.In addition to the ECA competition hypothesis, two other prevailing hypotheses are employed by ESMs: root nitrogen uptake is based on plant demand and the competition between root and decomposer microbes is scaled with their relative demand and the Microbes Win hypothesis , which assumes roots are completely out competed by microbes and that roots take up nitrogen after microbial demand has been satisfied. Model setup for RD and MIC are described in supporting information Method S3.They each predicted substantial nitrogen uptake near the permafrost boundary layer , in contrast to observed uptake.

These discrepancies occurred because, in these hypotheses, root uptake in shallow soil layers are either completely suppressed or largely suppressed by microbial decomposers . Therefore, they tend to acquire soil nitrogen from deeper in the soil profile, where microbial competition stress is lower. Both ECA and observations indicate that Eriophorum shifts its nitrogen uptake from the organic layer to mineral layer, but not down to the permafrost boundary layer. Overall, by explicitly considering maximum rooting depth, biomass density, and uptake kinetics, ECA is the only hypothesis that captured NH4 + uptake patterns for all three tundra species.For Carex aquatilis and Salix rotundifolia, the observed NH4 + uptake profiles were consistent with the prevailing hypothesis that fine-root biomass density, as functionally absorptive tissues, exerts first-order control on nutrient uptake [De Baets et al., 2007; Vamerali et al., 2003]. For Eriophorum angustifolium, however, the observed uptake profile did not follow the prevailing hypothesis. We showed that this pattern resulted from decreased competition between roots and microbial decomposers in mineral soils. The ECA competition hypothesis as integrated in the N-COM model explicitly represents these competitive interactions and accurately predicted the NH4 + uptake profile . The model results indicate that since Eriophorum angustifolium is a relatively poor competitor for NH4 + , it shifts its uptake profile deeper in the soil, in order to avoid NH4 + competition with microbial decomposers in the organic layer. Root physiology traits suggest that the Eriophorum angustifolium root system is less carbon efficient. In particular, compared with Carex aquatilis, maintenance and growth respiration per gram of root are higher but root longevity is much shorter for Eriophorum angustifolium [Billings et al., 1977; Shaver and Billings, 1975]. Although root morphological traits suggest that Eriophorum angustifolium has higher root length per gram root biomass [Eissenstat et al., 2000], total root density is much lower than Carex aquatilis . Furthermore, the Eriophorum angustifolium NH4 + uptake pattern is also consistent with the idea that microbial activity and N immobilization are highly limited by carbon availability. Compared with mineral soil layers, relatively higher carbon availability in surface organic layer will lead to higher potential of microbial activity and consequently higher microbial N immobilization demand and stronger nitrogen competition between plant and microbe [Booth et al., 2005]. Although both gross nitrogen mineralization and immobilization rates are commonly high in surface soils, net immobilization typically occurs because of strong microbial demand [e.g., Iversen et al., 2011]. Overall, our 15N tracer measurements and modeling analysis at Barrow, Alaska, showed that plant nitrogen uptake patterns emerge from root and soil biotic competition, which could be predicted by essential root traits and appropriate treatment of microbial competitive interaction. Although not studied here, mineral surfaces are also effective competitors for enzymes [Sulman et al., 2014; Tang and Riley, 2015], and further research is required to determine when those processes need to be included in nutrient and carbon cycle models.In this study, we showed that an important complication in predicting arctic tundra vegetation species responses to warming is associated with their different root characteristics, which can affect their ability to compete for elevated nitrogen availability throughout the soil profile. In this sense, explicitly considering key root functional traits is particularly important for studying warming-induced fertilization effects on arctic vegetation. Here we highlight the importance of several essential root traits in controlling nitrogen uptake patterns. First, maximum rooting depth is an important plant functional trait in modeling plant nitrogen uptake and response to arctic warming.