Non-growing season plant nutrient uptake controls Arctic tundra vegetation composition under future climate

Earth's warming over the coming decades in response to human activity will depend strongly on the terrestrial carbon cycle, which has gross uptake (photosynthesis) and release (respiration) fluxes that are about an order of magnitude larger than anthropogenic fossil fuel CO₂ emissions (Ciais et al 2013). Currently, the terrestrial biosphere assimilates about a quarter of anthropogenic CO₂ emissions, but the extent to which this service will continue is very uncertain in Earth System Models (ESMs) used to predict future climate (Friedlingstein et al 2014). Of particular concern is the high-latitude permafrost region, which contains more carbon in the top 3 m of soil than is currently in the atmosphere (Schuur et al 2008), and for which models and observations currently diverge in assessments of 21^st century CO₂ exchanges with the atmosphere (McGuire et al 2018). Observational studies have found increases in shrub abundance due to climate warming (e.g. Elmendorf et al 2012), which may increase ecosystem carbon storage via increased litter with higher C:N ratios or decrease it because of priming (Parker et al 2015). An important component of uncertainty in how shrubs will respond to warming is the impact soil nutrients (i.e. nitrogen (N) and phosphorus (P)) have on photosynthesis and plant growth (Mekonnen et al 2021a). Most terrestrial ecosystems are limited by N, P, or both (Elser et al 2007, LeBauer and Treseder 2008, Zamin and Grogan 2012, Zamin et al 2014, Du et al 2020), and these limitations are expected to intensify with increasing plant growth under high atmospheric CO₂ levels (Hungate et al 2003, Wieder et al 2015). Although nutrient constraints on terrestrial carbon cycling were either inaccurately represented (Bouskill et al 2014, De Kauwe et al 2017) or absent in ESMs that participated in the Coupled Model Inter-comparison Project 5 (CMIP5), many modeling groups have recently incorporated nutrient dynamics in preparation for the ongoing CMIP6 (Riley et al 2018).

An important component of these new land models is their representation of soil nutrient competition between microbes, plants, and abiotic processes (e.g. leaching, sorption). This competition is commonly resolved in models using the relative demand (RD) approach, which defines plant nutrient demand based on instantaneous photosynthetic demand for fixation, thereby precluding non-growing season (NGS) plant nutrient uptake (Riley et al 2018). We define the NGS here with respect to aboveground plant growth, to distinguish from belowground growth, which can continue well past the time when aboveground growth has ceased, as reviewed below. Implications of these assumptions may be large for predicted high-latitude carbon cycling. For example, Koven et al (2015), using the CLM4.5 land model, concluded that N mineralized in currently frozen soils under 21^st century warming would not substantially benefit plant growth, largely because that model's RD implementation eliminates plant N uptake when photosynthetic demand ceases.

However, observations indicate that high-latitude belowground plant activity continues well past aboveground growth. Up to 90% of tundra vascular plant biomass is belowground, and root production continues weeks to months past when aboveground production has ceased (Iversen et al 2015). Substantial late-season root growth was found in several sub-Arctic plants, with woody plants showing larger asynchrony between aboveground and belowground growth than sedge communities (Sloan et al 2016). In a northern Scandinavian tundra site, exudation rates increased in autumn while allocation to aboveground plant components decreased (Olsrud and Christensen 2004). In three Arctic plant communities along an elevation gradient over two growing seasons, belowground growth continued 40–58 days longer than aboveground growth (Blume‐Werry et al 2016). Although high-latitude measurements of NGS belowground processes are difficult and therefore sparse, these studies clearly demonstrate continued root activity well past when aboveground growth has ceased.

Consistent with the observed asynchrony in belowground and aboveground plant production, NGS root N and P uptake has been widely observed. For example, in a coastal heath system, soil N acquired by roots during winter was comparable in magnitude to N uptake during the growing season (Andresen and Michelsen 2005). In a series of studies of mountain birch in Sweden, (a) seedlings increased their N content by 73% over the winter season, accounting for 15% of total annual N uptake (Weih and Karlsson 1997); (b) the NGS accounted for ∼10%–20% of annual N uptake (Weih et al 1998); and (c) winter N uptake varied between 10% and 29% of annual N uptake at two sites (Weih 2000). N uptake over winter increased growing season productivity in a grassland system in Ontario Canada (Malyshev and Henry 2012). Mullen et al (1998) found that roots and total plant N content increased after aboveground growth ceased, and concluded that the root N increase was not attributable to aboveground translocation. P uptake during the NGS has also been observed. For example, for tundra sedges near Utqiaġvik (formerly Barrow) Alaska, USA, about half of their annual phosphate uptake occurred after net translocation of aboveground P (i.e. from leaves) to belowground storage had begun, a process plants use to conserve their limited nutrient supplies (Chapin and Bloom 1976). Overall, these and other (supplementary information (available online at stacks.iop.org/ERL/16/074047/mmedia), Mullen and Schmidt 1993, Bilbrough et al 2000, Grogan and Jonasson 2003, Miller et al 2009, Edwards and Jefferies 2010, Larsen et al 2012) studies suggest that up to half of annual plant nutrient uptake occurs outside of the aboveground growing season. Unfortunately, however, these measurements are insufficient to (a) extrapolate across geographical, temporal (seasonal to multi-decadal), and plant functional type (PFT) variations and (b) estimate their effects on high-latitude C cycling.

To investigate how high-latitude NGS plant nutrient uptake affects Arctic tundra vegetation composition (i.e. shrub expansion) and ecosystem carbon uptake, we further evaluate and apply a well-tested ecosystem model (ecosys) that explicitly links plant and microbial traits for nutrient uptake with prognostic soil N and P transformations (e.g. Grant (2013), Mezbahuddin et al (2020)). The ecosys model does not apply the RD approach, but rather applies a 'capacity-based' approach (Tang and Riley 2021). In particular, each of the modeled ecosystem's plant and abiotic (e.g. mineral sorption) competitors for nutrients have dynamic capabilities that affect the relative nutrient acquisition rates. With this approach the model is able to incorporate NGS processes important for evaluating impacts on vegetation dynamics over multi-decadal timescales. After describing the relevant modeled processes and simulation design, we (a) compare observed and modeled NGS uptake at a polygonal tundra site in Alaska; (b) quantify the spatially-explicit NGS fraction of annual N uptake for each of our five simulated PFT's; and (c) perform model experiments excluding NGS nutrient uptake to analyze the effects on 21^st century tundra shrub dynamics. The large impacts on modeled net primary productivity (NPP) and 21^st century shrub expansion are explained by differences in plant traits relevant to nutrient acquisition and retention. We conclude with a discussion of needed observations and model improvements.

Ecosys has been evaluated successfully over the past three decades in many ecosystems, including high-latitude tundra sites (e.g. Chang et al 2020, Grant 2015, Grant et al 2015, 2017a, 2017b, 2019a, 2019b). For this study, we first performed site-level simulations at a polygonal tundra site in Utqiaġvik, Alaska, USA (71.3° N, 156.5° W), which has a mean annual air temperature and precipitation of −12 °C and 106 mm yr⁻¹, respectively. In recent studies at this site (Grant et al 2019a), ecosys accurately captured landscape-feature-specific CO₂ (R² = 0.7–0.9) fluxes, CH₄ (R² ∼ 0.9) fluxes, leaf area index (LAI) (|bias| = 0.01–0.35 m² m⁻²), and plant biomass (|bias| = 0.08–19.6 gC m⁻²); landscape-scale latent (R² = 0.71–0.77) and sensible (R² = 0.78–0.88) heat fluxes; soil temperatures (R² = 0.92); active layer depth (RMSE = 2–5 cm); and soil moisture (RMSE = 0.05–0.09 m³ m⁻³). We also analyze 25 km resolution simulations across the North America (NA) tundra in which ecosys accurately reproduced observed inter-annual variability in LAI spatial patterns (R² = 0.71), long-term mean annual gross primary productivity (GPP) (R² = 0.78) (Mekonnen et al 2018), and active layer depth from 28 Circumpolar Active Layer Monitoring sites (R² = 0.63; RMSE = 10 cm) (Mekonnen et al 2021b).

Here we provide a brief description of the model's process representations important for NGS biogeochemistry and plant activity and to evaluate our model experiments. A detailed description of inputs, parameters, and algorithms used in ecosys can be found in Supplemental Information.

Ecosys represents multiple canopy and soil layers and solves the fully coupled carbon, energy, water, and nutrient cycles with a dynamic time stepping algorithm to ensure numerical accuracy. Surface energy and water exchanges drive soil heat and water transfers that affect temperature and water content in multiple soil layers. These transfers drive soil freezing and thawing and, hence, active layer depth, through the general heat flux equation.

Eleven microbial functional groups that affect soil carbon, nitrogen, and phosphorus dynamics are simulated in the model (Grant 2013). Decomposition rates of different organic matter substrates are combined functions of active biomass in microbial populations and substrate concentrations (Grant et al 2006). Microbial respiration rates are controlled by soil water potential and concentrations of microbial nitrogen and phosphorus, dissolved organic carbon, and O₂. Decomposition rates determine net nutrient mineralization and thereby soil mineral nutrient content.

We initialized five coexisting PFTs (deciduous shrub, evergreen shrub, graminoid, moss, lichen) with equal amounts of total seed carbon across the simulation spatial domain. Modeled PFT-specific traits (e.g. specific leaf area, leaf optical properties, leaf clumping, leaf turnover, foliar nutrient content, foliar nutrient retention, root hydraulic conductivity) result in different responses to environmental and competitive controls. Plant N and P uptake from soil is modeled from convective and diffusive transfers of ammonium, nitrate, and phosphate to root and mycorrhizal surfaces coupled with active uptake from these surfaces to nonstructural pools within roots and mycorrhizae. Path lengths and surface areas of roots and mycorrhizae used to model uptake are calculated from a root and mycorrhizal growth sub-model driven by dynamic nonstructural C, N, and P concentration gradients within the plant (Grant 1998). Energy requirements for active root nutrient uptake is driven by oxidation of root non-structural C, which is generated during the growing season by CO₂ fixation. Remobilization of shoot nonstructural C, N, and P under cooling temperatures and shortening photoperiods during autumn in deciduous PFTs is stored and facilitates leafout the following spring. This modeled remobilization into storage provides resources for overwinter respiration and growth in Spring. A product inhibition function based on root and mycorrhizal nonstructural C:N:P ratios is included to avoid uptake in excess of nutrient requirements determined by C availability. Thus, root C respiration can drive N and P uptake even when shoots are not photosynthetically active (i.e. during the NGS) and can reduce nutrient accumulation from overwinter mineralization that could be lost during the following spring.

Deciduous shrubs are modeled to have greater specific leaf area and less leaf clumping and thus greater light interception as compared to evergreen shrubs. Deciduous shrubs have full annual leaf turnover, while evergreen shrubs retain their leaves over several years. These lower evergreen shrub nutrient losses enable them to compete more effectively in nutrient-limited environments (Aerts 1995). Deciduous shrubs also have higher potential leaf nitrogen and phosphorus concentrations under non-limiting conditions than evergreen shrubs, which affect maximum carboxylation rates (V_max) and electron transport (J_max). Under nutrient rich conditions, greater nutrient demand and investment in uptake capacity allows for higher CO₂ fixation in deciduous vs. evergreen shrubs. Deciduous shrubs are also modeled to have lower axial resistivity, thus facilitating faster water and nutrient uptake and growth (Grant 2013).

These differences in plant traits result in emergent variations in phenology, irradiance, CO₂ fixation rate, and water uptake among PFTs. These processes drive vertical profiles of leaf area and root length that determine competition for radiation, water, and nutrients within each canopy and rooted soil layer. The vertical profiles are generated from allocations of plant nonstructural carbon, nitrogen, and phosphorus to each organ of each PFT.

Allocation rates are determined by non-structural carbon, nitrogen, and phosphorus concentration gradients within each PFT arising from leaf CO₂ fixation, and by root nitrogen and phosphorus uptake versus carbon consumption through maintenance and growth respiration. Temporal changes in nutrient availability, modeled through soil nutrient transformation processes, and the relative competitive ability of PFTs (as affected by their inherent traits) determine plant nutrient uptake, and thus growth.

2.1. Simulation design

3.1. Model testing

Modeled and observed (Chapin and Bloom 1976) seasonality of plant P uptake at the Utqiaġvik site were broadly consistent (figure 1(a)). That observational study reported that tundra sedges absorbed about half of annual P uptake after net nutrient translocation belowground began while the model predicted a 30%–55% variation in NGS P uptake across landscape position in the polygons (i.e. trough, rim, center). Modeled N and P uptake follow approximately the same seasonal cycle (supplementary information figure S1). Both observations and simulations indicate that the large volume of soil, which remains unfrozen well past the end of the growing season (figure 1(b)) allows for continued root and microbial activity and plant nutrient uptake. Immediately after the growing season ends, the proportion of N acquired by plants decreases concurrently with the ratio of root to total soil respiration (figure 1(c)). After about day-of-year 300, this ratio increases, indicating continued root activity, although both components of belowground respiration fluxes are small during the winter. While this comparison provides some confidence in the model's process representations, it cannot provide an observational constraint for all PFTs in Alaska. Therefore, more temporally-resolved observations of plant nutrient uptake could substantially benefit model development and evaluation.

Zoom In Zoom Out Reset image size

To examine the effects of nutrient uptake seasonality on NPP, we performed an analogous Utqiaġvik site-level model experiment, but precluded plant nutrient uptake during the NGS (figure 2). Over each subsequent winter, the lack of NGS plant uptake led to higher mineralized nitrogen stocks in the soil, which were then lost in the subsequent growing season. These enhanced losses were dominated by N₂O gaseous losses, compared to N leaching losses. Root N content was lower compared to the baseline simulation at the end of the NGS, and therefore less N is available to fuel springtime plant growth. By the fifth simulation year, compared to the baseline simulation, spring NPP decreased by ∼23%, end of growing season NPP decreased by ∼5%, and annual NPP decreased by ∼16%.

Zoom In Zoom Out Reset image size

Because our goal is to estimate large-scale implications of NGS nutrient dynamics, we next analyzed their importance across the NA tundra region. Under current climate, simulations indicate large spatial heterogeneity and differences between PFTs in the fraction of N acquired during the NGS ( $f_{\text{N}}^{{\,\text{NGS}}}$; figure 3). These predictions fall within the range of measurements across a wide variety of systems (∼5%–50%; described above and in supplemental information). The graminoid PFT has the largest modeled zonal-mean $f_{\text{N}}^{\,{\text{NGS}}}$ values, ranging between 20% and 40% (figure 3(D)), evergreen shrubs have the smallest values (∼1%–25%), and deciduous shrubs are intermediate. There were also spatially- and PFT-dependent heterogeneous differences in the proportion of NGS N uptake that occurred early versus late in the NGS, both under current and end-of-century conditions (figure S2).

Zoom In Zoom Out Reset image size

Variations in modeled NGS nutrient uptake in Arctic tundra PFTs emerge from differences in inherent plant traits that control modeled PFT nutrient acquisition and growth strategies (Mekonnen et al 2018). We found that the PFT growth strategy is related to the resulting NGS nutrient uptake. Evergreen shrubs are represented as the most conservative PFT (i.e. slower growing, slower leaf turnover), while deciduous shrubs employ more rapid nutrient uptake capacity (and are better competitors under more nutrient rich conditions expected as climate warms) and more rapid leaf turnover. Our modeling results show that PFTs with less conservative strategies have higher potential for NGS nutrient uptake.

To explore interactions between climate change and the importance of NGS nutrient uptake, we first performed a baseline simulation forced with an RCP8.5 scenario over the 21^st century (section 2). Consistent with observational studies (Myers-Smith et al 2018) and simulations (Zhang et al 2013), this baseline simulation indicates that shrubs will continue to increase in dominance throughout the 21^st century (figure 4(a); solid lines). As has been reported elsewhere (Lader et al 2018), we found that growing season lengths will increase by between ∼1 and 2 months under this climate change scenario (figure S3), with large differences among PFTs and location (Mekonnen et al 2018). Evergreen shrubs and graminoids experienced the longest and shortest increases in growing season length, respectively, with large differences across latitude. The 21^st century decreases in NGS length caused decreases in $f_{\text{N}}^{\,{\text{NGS}}}$ (including up to ∼70% decreases in evergreen and deciduous shrubs and much smaller changes in graminoids) and changes in ecosystem composition.

Zoom In Zoom Out Reset image size

We next performed a 21^st century RCP8.5 simulation that excluded NGS nutrient uptake for all PFTs (section 2). Comparisons with the baseline simulation indicate that wintertime nutrient dynamics and plant acquisition explain a substantial portion of shrub expansion rates modeled by year 2100. Excluding NGS nutrient uptake led to about a 50% decline in shrub dominance (figure 4(a); black dashed lines), concurrent with a comparable increase in graminoid (red dashed lines) and non-vascular (blue dashed lines) plant dominance. The decrease of evergreen shrubs was larger than of deciduous shrubs (figure 4(b)). As described above, deciduous shrubs have a relative advantage over evergreen shrubs under this (unrealistic) scenario because of their autumn remobilization of nutrients and greater ability to take advantage of the nutrient rich environment during snowmelt and before leafout. In the absence of NGS nutrient uptake, shrubs were unable to grow tall enough to fully benefit from their inherent capacity for light competition conferred by their taller stature (Bjorkman et al 2018). Finally, the lack of NGS nutrient uptake, and the subsequent decline in modeled shrub expansion, resulted in about a 25% reduction in ecosystem NPP compared to baseline by year 2100 (figure 4(c)). These results suggest that models that ignore NGS nutrient uptake are likely to predict lower rates of recent and future tundra shrub expansion, lower ecosystem NPP, and therefore biased high-latitude carbon-climate feedbacks.

There are, of course, uncertainties in our predictions of high-latitude biogeochemical, hydrological, and thermal processes that could affect vegetation dynamics and nutrient uptake predictions. In particular, we did not consider thermokarst, or variation and uncertainty in landscape hydrology, and soil biogeochemistry below the model's spatial resolution, all of which could affect our results. Although ecosys represents coupling between plants and soil processes (e.g. Grant (2013)), more work needs to be done to evaluate these interactions in high-latitude systems (Parker et al 2015, 2021).

We show here, consistent with observations, that modeled North America tundra NGS plant nutrient uptake has large spatial heterogeneity, differences among PFTs, and trajectories across the 21^st century. Our modeled graminoid PFT had the highest zonal-mean $f_{\text{N}}^{\,{\text{NGS}}}$ values, reaching ∼40% at some latitudes and being highest in the eastern half of our study region. These differences in timing of nutrient acquisition by PFTs strongly affect modeled vegetation composition of Arctic tundra. We found that NGS plant nutrient uptake is a substantial factor in controlling 21^st century shrub expansion. The enhanced 21^st century NGS nutrient availability strongly affects competition between deciduous and evergreen shrubs, largely benefiting the faster growing, less conservative deciduous shrubs. Ignoring NGS plant nutrient uptake led to a dramatically different 21^st century tundra ecosystem, one dominated by moss and graminoid PFTs instead of shrubs.

The substantial NGS tundra plant nutrient acquisition found here and in observations, and its large impacts on vegetation composition, implies that land models must implement realistic representations of plant root and soil biogeochemical processes. Since most models that will participate in the next version of climate change assessments adopt the RD approach to resolve nutrient competition, they do not account for NGS plant nutrient uptake (Riley et al 2018). As we demonstrate here, excluding these processes unrealistically limits modeled vegetation access to nutrients, which then substantially affects predicted high-latitude ecosystem nutrient dynamics and vegetation composition and growth, and therefore likely carbon-climate interactions. We conclude that NGS dynamics require models to represent the coupling of plant growth with carbon and nutrient allocation, root nutrient acquisition based on root properties, and the belowground competitive environment with microbes. Finally, although we attempted to synthesize as many observations of NGS nutrient acquisition as possible, more datasets and their metadata are needed for model development and evaluation. Needed observations to develop and test models include temporally-resolved plant nutrient acquisition, storage, allocation, and use. Accurate representation of NGS biogeochemical and plant root dynamics requires both model improvements and new observations.

This research was supported by the Director, Office of Science, Office of Biological and Environmental Research of the US Department of Energy under contract no. DE-AC02-05CH11231 to Lawrence Berkeley National Laboratory as part of the Next-Generation Ecosystem Experiments in the Arctic (NGEE-Arctic) project.

The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.5440/1785957. Data will be available from 28 July 2021.