In our December REALM team meeting, we discussed how to represent nutrients in Land Surface Models (LSMs). Below is a summary of the following talks from the team:
- Colin Prentice: How NOT to represent nutrients in land-surface models.
- Huiying Xu: Controls of leaf C:N ratio
- Yunke Peng: Downregulation of photosynthetic capacity under carbon dioxide enrichment
- Ruijie Ding. A parsimonious model for carbon allocation
- Cai Wenjia: Soil effects on forest growth
.
Colin Prentice: How NOT to represent nutrients in land-surface models.
Back in 2001 (see Cramer et al. 2001 Global Change Biology), the main ecosystem models, such as LPJ, were “carbon-only” models and ignored nutrients and “nitrogen limitation” (which is better expressed as “costs of nitrogen acquisition” in my view, but we’ll come to that). Hungate et al (2003, Science) made the point that when you take into consideration the nitrogen supply limits, i.e., the assumed rates at which nitrogen can be supplied to ecosystems, all the models predicted nitrogen uptake above the upper threshold – leading to unrealistically high land carbon uptake by 2100 under high-end emissions scenarios.
That short but influential paper started a race to add nutrient cycling to Earth System models (ESMs). Firstly, the modelling community decided there was an urgent need to include “nitrogen limitation” in the models in order to reduce their projections of the future carbon sink (an unfortunate outcome in some cases was also to reduce the present carbon sink, although this had been correctly simulated before). Then by 2015, the same modelling groups – having incorporated nitrogen – decided that it was necessary also to include phosphorus cycling; because the growth of many tropical forests, at least, does appear not to be limited by nitrogen, but rather by phosphorus.
So now we have a situation where many ESMs include representations of both the nitrogen and phosphorus cycles, coupled to the carbon cycle. The trouble is that this modelling effort has raced ahead of our understanding of how these cycles are linked, while several major research questions – essential to the modelling effort – have not been satisfactorily addressed in the plant functional ecology literature. They are given extra importance and urgency by the need to develop reliable representations of plant-nutrient interactions in ESMs.
- What are the mechanisms of “nitrogen limitation” as observed today?
“Nitrogen limitation” has been demonstrated experimentally in many ecosystems, especially in colder climates. This simply means that fertilization with nitrogen increases plant growth. But we also need to know how this happens. Many nitrogen addition experiments have documented increases in carbon allocation above ground (including to leaves) and decreasing root-shoot ratios, allowing more leaves to be produced (and thus more photosynthesis carried out) with less effort by roots. Many current models do not represent this control of carbon allocation, although it seems to be the most important mechanism by which plants benefit from nitrogen addition in the real world.
- What determines leaf nitrogen content?
Numerous papers have shown a relationship between photosynthetic capacity and leaf nitrogen content and inferred that leaf nitrogen content controls photosynthetic capacity. But the relationship is only a correlation. It says nothing about which is the cause, and which is the effect. A strong case can be made, instead, that photosynthetic capacity is optimized to the growth environment (Peng et al. 2021 Communications Biology) and that it, along with leaf mass-per-area, controls leaf nitrogen content (Dong et al. 2017, Biogeosciences). Further confusion has been sown by the habit of expressing both quantities on a mass basis. Leaves with higher leaf mass-per-area typically have both lower photosynthetic capacity, and lower nitrogen, per unit of leaf mass, so high correlations between mass-based quantities tell us nothing we didn’t know already (see the perceptive, if somewhat provocative, analysis by Lloyd et al. 2013 New Phytologist).
- Is the CO2 fertilization effect influenced by nitrogen supply?
There seems to be a widespread belief to this effect. Indeed, it has often been assumed that “nitrogen limitation” implies a reduction in the CO2 fertilization effect. Yet provided this effect is appropriately expressed (as a response ratio, rather than an absolute amount) then there is no obvious reason why it should be true in general. Moreover, many experimental data don’t support it. A revealing case study is provided by Fig. 9 in Nowak et al. (2004) New Phytologist, whose caption states that:
“The ratio of response under elevated [CO2] to that under ambient [CO2] (E/A) for low nitrogen (N) availability … are production under elevated [CO2] and low N availability divided by production under ambient [CO2] and low N. The E/A ratio for high N availability … are production under elevated CO2 and high N availability divided by production under ambient [CO2] and low N.” (italics mine).
It is thus made to appear as if the relative enhancement of productivity by elevated CO2 is greater at high N; but that is not what these experimental data show. There is, nonetheless, evidence that this enhancement really is generally greater at high nitrogen levels in plants with arbuscular mycorrhizal symbionts (see Terrer et al. 2016 Science and subsequent discussion). Some land models are now starting to incorporate the distinction among mycorrhizal types.
- What determines the rates of nitrogen fixation and loss?
Both quantities are extremely hard to measure, and widely different results can be obtained depending on the method used – making it difficult to gain any insight into their controls.
All the same questions apply to phosphorus, with the key difference that phosphorus can only be obtained from rock weathering (or dust deposition) – there is no process analogous to nitrogen fixation, whereby ecosystems can gain additional nitrogen from the air through the action of free-living and symbiotic microbes. Overall, we know even less about the phosphorus cycle than the nitrogen cycle.
A paper by Du et al. (2018 GMD) examined three contrasting carbon-nitrogen models and described the various assumptions made. I am not convinced by any of them, for the following reasons:
| Assumption | Problems with assumption |
| “Downregulation” of photosynthesis with rising CO2 is caused either directly through limitation by the rate of nitrogen mineralization, or indirectly via low leaf nitrogen content. | Reduction of Vcmax is a common response to CO2 enrichment, but it has a simple explanation in terms of optimality theory (Dong et al. 2022, New Phytologist); it has nothing to do with “nitrogen limitation”; and it is always accompanied by an increase in photosynthesis when this is measured at actual (rather than saturating!) light levels. It is also relevant that nitrogen fertilization does not generally result in enhanced Vcmax (see e.g., Fig. 4a in Liang et al. 2020 GCB). When nitrogen is added, Vcmax sometimes increases; more often, it doesn’t. |
| “N limitation” of photosynthesis takes place either via a reduction in Vcmax, or (in one model) a reduction in the ci:ca ratio (χ). | See above regarding the controls of Vcmax. There is no evidence for a reduction of χ at low nitrogen levels. |
| The JULES model includes the concept of rapid respiration of “excess” carbon, which is fixed and then quickly returned to the atmosphere because it can’t be used. | This has never been observed. It is also highly unlikely, as plants do not generally engage in wasteful metabolic activity. |
| Nitrogen fixation rate is determined by Net Primary Production (NPP) or Actual Evapotranspiration (AET). | We do not know how nitrogen fixation is controlled at the ecosystem level, but it is surely not determined by either NPP or AET. The present carbon sink can be well simulated simply by assuming that nitrogen fixation adjusts to meet stoichiometric demands (Xu-Ri & Prentice 2017, Biogeosciences). |
| Phosphorus is treated in parallel with nitrogen. | Little is known about how phosphorus supply affects GPP (Gross Primary Production) or NPP. But we do know that phosphorus is not parallel to nitrogen, either in the manner of its use by plants, or its acquisition; therefore, it should not be treated in the same way. |
| “Liebig’s Law” for limitation by nitrogen or phosphorus. | Liebig’s law is generally false. But that would be a whole other blog post! |
The take-home message is that many of the assumptions about how to model nutrients in current land-surface models are probably wrong. We don’t have firm knowledge yet. There is a pressing need to acquire it, even if it involves jettisoning some prevalent assumptions about the controls of photosynthetic traits. A promising line of enquiry involved re-casting “nutrient limitation” in an optimality context, in terms of the costs and benefits of nutrient acquisition – see e.g., Paillassa et al. (2020) New Phytologist and Westerband et al. (2022) Global Change Biology. This approach has now also been adopted in the energy exascale Earth System land model (Braghiere et al., 2022 JAMES). Perhaps a paradigm shift is on the way?
Huiying Xu: Controls of leaf C:N ratio
Leaf carbon (C) and nitrogen (N) are key elements of plant cells, which influence productivity, respiration and litter decomposition rates. Leaf C:N ratio is an important parameter in land surface models (LSMs) to couple carbon and nitrogen cycling, however treated as fixed values for each plant functional type (PFTs). The knowledge of leaf stoichiometric trait variations along climate gradients help us understand how plant adjust stoichiometry to maintain different functions, and improve the representation of leaf C:N in dynamic global vegetation models (DGVMs). We first investigated the driving factors of leaf stoichiometric traits including taxonomy, phylogeny and environmental variables. Then we extended the leaf nitrogen framework from area basis to mass basis to predict the leaf nitrogen content (Nmass) and C:N ratio. The result shows that leaf stoichiometric traits are phylogenetically conserved, and explained nearly 40% of their variations by species and phylogeny at species level. The climate variables increased almost fivefold of explained variations at community level, indicating the importance of species turnover along climate gradients. We found that Nmass determines the variation in leaf C:N and nitrogen partitioning between metabolic and structural components shifts as leaf area index changes. The optimality-based model captures 30% of Nmass variation and predicted C:N is within the range of observations.
Fig 1. The comparison between observations, optimality-based predictions and fixed PFTs of leaf C:N ratio in LSMs.
Yunke Peng: Downregulation of photosynthetic capacity under carbon dioxide enrichment: experimental support for demand-side control.
Our understanding of photosynthetic acclimation to CO2 is incomplete. Observed reductions in maximum rates of carboxylation (Vcmax) and electron transport (Jmax) under elevated CO2 (eCO2) have been variously explained by limited nitrogen (N) supply or reduced leaf-level N demand. Here we show that Vcmax, Jmax and leaf N decreased, while Jmax/Vcmax increased, with eCO2. These responses were independent of experimental N fertilization, species’ N fixation status, and eCO2-related soil inorganic N decline. Net primary production (NPP), leaf area index (LAI), and belowground allocation increased with eCO2, with larger increases accompanying larger declines in Vcmax. At increasing CO2, additional photosynthate is produced, with higher root allocation shown to transport more N required for higher NPP. These findings, taken together, do not support the control of Vcmax by N supply. They are consistent with a reduction in leaf-level N demand in response to eCO2.
Fig. 1: Sensitivity coefficients of Vcmax, Jmax and Jmax / Vcmax under eCO2 with different N availability (n = 12). Two types of comparison are referred to by the axis labels ‘Low nitrogen’ and ‘High nitrogen’: non-fertilization versus fertilization (n = 5: black) and low versus high N fertilization (n = 7: blue). ‘High – low’ indicates the difference in sensitivity coefficient values between high and low N treatments.
Fig. 2: Bivariate relationships among the sensitivity coefficients S for different observed quantities with respect to eCO2. Red points are experiments with N-fertilization (including experiments with low and high levels of fertilization, and experiments with and without N fertilization, as shown in Fig. 3). Green points are experiments with N-fixing species. Blue points are experiments with no N-fertilization and no N-fixing species. Vcmax and Jmax responses are compared to those of all other quantities. For leaf traits (Narea, Nmass, LMA) versus other variables (ANPP, BNPP, root/shoot ratio, LAI, soil inorganic N) only relationships with p ≤ 0.1 are shown. Relationships with p ≤ 0.1 are shown with solid lines and confidence intervals (95%), and those with 0.1 < p ≤ 0.2 with dotted lines. No lines are shown for cases with p > 0.2. p values refer to a two-sided t-test for the slope of the respective bivariate relationship being significantly different from zero.
Ruijie Ding. A parsimonious model for carbon allocation
The distribution of assimilated carbon among different compartments is critical in explaining the exchanges between multiple functional traits to respond to climate change. Machine learning techniques make progress in characterizing the environmental drivers of the variation in biomass allocation at large spatial scales. However, C allocation is treated simplistically in current ecosystem models. Quantitative aspects of C allocation require theoretical analysis, and theoretical predictions require testing against the available data from experiments and large-scale surveys to provide a better foundation for modelling. Eco-evolutionary optimality (EEO) concepts have already proved their worth in generating testable – and tested – hypotheses about plant and ecosystem function at the leaf and canopy levels. This project aims to apply and extend eco-evolutionary optimality (EEO) principles into a parsimonious model to quantify C allocation partitioning and its interaction with different environmental variables.
We synthesize a database of 29896 root: shoot biomass ratio (R:S) worldwide with gridded datasets of climatic and edaphic conditions to investigate environmental controls on the partitioning of below- versus aboveground biomass using Ordinary Least Squares (OLS) multiple linear regression modelling. Our results show that the variation of R:S across and within vegetation types (woody and herbaceous) is attributable to resource availability. R:S is significantly governed by growing-season mean temperature (Tg), root zone water capacity (RZWC), gross primary production (GPP), pH and soil sand content across vegetation types (woody and herbaceous). LAI is optimized for the environment, while a certain quantity of roots is essential to maintain the leaves. As GPP increases, so does the potential to allocate carbon to stems, which are predominantly above ground, leading to lower R/S. Warmer temperatures lead to faster turnover of nutrients in soils, then a smaller requirement for carbon allocation to roots. A reduced need for roots to take up nutrients on acid soils (low pH) because of the lower optimal photosynthetic capacity. An increased need for root stability in sandy soil due to the low water-holding capacity. More carbon is allocated to roots in climates with seasonal mismatches between water supply and demand to increase their root-zone water capacity (RZWC) to buffer the experienced water deficit for the dry season.
Fig.1 Partial residual plots from the regression of R:S against explanatory variables for the combination of two vegetation types. The plots are from ordinary least squares multiple linear regressions. a: growing-season mean temperature (Tg, ˚C). b: the natural-log transformed root zone water capacity (ln RZWC, mm).c: the natural-log transformed gross primary production (ln GPP, gC/m2/d). d: the pH. e: the soil sand content (Sand, % of weight). Predicted relationships in pink with 95% confidence interval in grey, the density of points is the frequency of data in geographic space.
Cai Wenjia: Soil effects on forest growth
Vegetation growth depends on investment on functional organs in return for carbon assimilation. Vegetation investments on leaf photosynthesis not only consist of leaf construction and maintenance cost, but also that for fine roots to acquire nutrient which is the source of chlorophyll. Soil nutrient is thus a critical factor in determining both overall and relative carbon allocation to leaf and root. Despite numerous studies on nutrient impact on leaf traits, few of them focused on leaf area. Based on eco-evolutionary optimality principles, we hypothesized that the capacity of root for nutrient forage should equal the ability of leaf carbon acquisition (functional equilibrium), and that the ratio of fine roots biomass to leaf area would shift along the nutrient gradient to optimize net carbon gain, consequently the overall investments of leaf and fine root would be linked to soil nutrient status.
Here we use over 1600 site measurements on fine root/leaf biomass from FRED3 across different soil and climatic conditions, together with specific leaf area (SLA) of different species from TRY database to calculate the ratio of fine root biomass to leaf area at these sites, and compare against soil properties, climate variables, mycorrhizal fungi and phenology traits. Plants tend to exhibit lower fine root to leaf area ratio with increasing temperature, suggesting more energy for acquiring nutrients is needed from tropic towards high latitudes. Species associated with ectomycorrhizal (EcM) have significant negative relationship with ratio of fine root biomass to leaf area, which indicated species, with the help of nutrient fixing fungi, could invest less to reach same carbon assimilation efficiency. This ratio is also negatively correlated with evergreen species, partly explained by temperature, but also demonstrated that phenology traits should be accounted for when considering nutrient effects on plant growth.
Fig.1. Relationship between ratio of fine root biomass to leaf area and soil properties using multiple regression. Values show regression coefficient, with error bars showing standard error. Yellow solid circle indicates positive relationship, while green indicates negative correlation. Significances of relationship are shown in asterisks.
Lab Prentice 