Estimates of soil nutrient limitation on the <scp>CO₂</scp> fertilization effect for tropical vegetation

Abstract

The CO2 fertilization effect in tropical forests is a key factor for the global land carbon sink. We show that the normalized CO2 effect on tropical vegetation carbon was c. 70% lower in seedling CO2 experiments without nutrient fertilizers and c. 50% and 70% lower in models that consider nitrogen and phosphorus cycles, based on two model ensembles. The inadequate representation or lack of nutrient cycles in Earth System models likely leads to overestimating future tropical carbon gains. Multiple lines of evidence support the hypothesis that the contemporary land carbon sink is primarily driven by carbon dioxide (CO2) fertilization of vegetation growth (Walker et al., 2021). Intact tropical vegetation contribute about ¼ to this terrestrial carbon sink, but their growth appears to level-off in recent decades and their carbon sink strength declines (Hubau et al., 2020). Earth System Models (from the Coupled Model Intercomparison Projects Phase 6 [CMIP6]), however, project an increase in the strength of the tropical vegetation sink for the coming decades, since elevated CO2 (eCO2)-induced vegetation growth continues to outweigh climate-induced vegetation losses in their simulations (Koch et al., 2021). This mismatch between forest inventories and model projections could be the result of an incomplete representation of the limiting factors of vegetation growth to eCO2 in models, such as the low soil nutrient availability in tropical forests. Tropical forests are predominantly late-succession forests, and found on highly weathered, nutrient impoverished soils (Quesada et al., 2012), so that their vegetation responses to eCO2 may be strongly constrained by nutrient availability. CO2 experiments in other climate zones indicate that nutrients strongly regulate the magnitude of the CO2 fertilization effect, and further emerging evidence suggests little or no growth response in mature, late-succession forests (Walker et al., 2021). Tropical CO2 experiments in tropical forests are limited to young plants and short durations, to date, but they offer empirical insights into the interaction between CO2 and soil nutrients. Next to estimates from process-based models, these studies add direct evidence on the magnitude of soil nutrient effects on CO2 fertilization in tropical vegetation. The CMIP6 models derive a vegetation β of 14 ± 2% (mean ± SE, n = 9) for the tropical biome. The estimate is derived from the “1pctCO2-bgc” experiment, which simulates a gradual increase in atmospheric CO2 of 1% each year, going from 372 to 616 ppm over 50 years. The change in CO2 is not affecting the climate in this model experiment so that it discerns the pure CO2 effect. Vegetation β was higher in the C-only models (18 ± 2%, n = 5) than in the nitrogen-enabled models (12 ± 1%, n = 3; Figure 1a). Inclusion of nitrogen cycling thus reduces the carbon-concentration feedback, that is, the land carbon uptake in response to CO2 (Arora et al., 2020). Consideration of phosphorus cycling in Earth System Models could further reduce vegetation β in the tropics, as shown by the one model in the ensemble, ACCESS-ESM1-5, that considers coupled carbon, nitrogen and phosphorus cycles and derives a vegetation β of 5% (Figure 1a; n = 1). Tropical CO2 experiments indicate a vegetation β of 10 ± 2% (n = 18). Some of these experiments artificially added soil nutrients. We found vegetation β was more than three times higher in experiments with combined nutrient fertilization (16 ± 3%, n = 7) than in experiments without added nutrients (5 ± 2%, n = 11; Figure 1a). Experiments without artificial nutrient-rich soils better mimic natural conditions in tropical forests, integrating the low availability of soil phosphorus and other limiting macronutrients (Quesada et al., 2012). CO2 experiments have predominantly been carried out in temperate experiments, with comparably few in tropical climate, and even fewer under very wet conditions (Figure 1b). Nevertheless, results from these globally distributed CO2 experiments and their relationship with experimental site factors, including soil phosphorus, can be extrapolated to estimate the CO2 effect across climate zones (Terrer et al., 2019). Vegetation β from this extrapolation of empirical estimates is 5 ± 2% for the tropical biome (Figure 1a). Site-scale simulations from a terrestrial biosphere model ensemble derive a vegetation β of 4 ± 1% (n = 14; Figure 1a). The simulated CO2 treatment was a step increase from 400 to 600 ppm in CO2 concentrations over a 15-year period, mimicking the planned AmazonFACE experiment (Fleischer et al., 2019). Vegetation β was highest for the C-only models (6 ± 1%, n = 3), followed by models accounting for nitrogen (5 ± 1%, n = 5) and for combined nitrogen and phosphorus (3 ± 1%, n = 6; Figure 1a). Initially, higher productivity in response to eCO2 was downregulated in the nutrient-enabled models, in particular through progressive phosphorus limitation of decomposition processes or through constraints on plant phosphorus acquisition. Models accounting for nitrogen, or nitrogen and phosphorus cycles, reproduce the constraining role of soil nutrients on vegetation β from experiments, with stronger control of combined nitrogen and phosphorus limitation in tropical forests. Assumptions on plant nutrient acquisition and plant stoichiometric plasticity in models are often barely constrained by observations, but determine their projections of soil nutrient feedbacks to vegetation growth under eCO2 (Fleischer et al., 2019). The relatively higher vegetation β in the CMIP6 models, compared to the biosphere models, partly results from the longer simulation time, and the fact that vegetation turnover is a function of vegetation growth and carbon pool sizes in most models. Other mechanisms may potentially accelerate carbon losses but are currently not captured by models, such as CO2-induced faster plant turnover, which would lead to reductions in vegetation β (Brienen et al., 2017). Nitrogen and phosphorus limitation in process-based models and the omission of nutrient fertilization in CO2 experiments thus reduces the CO2 effect on tropical vegetation carbon notably (Figure 1a). Model estimates unconstrained by nutrients seem to present unrealistic scenarios of nutrient availability for future vegetation carbon gains, equivalent to assuming that artificial fertilizers are loaded to whole biomes. The tropical CO2 experiments provide evidence that soil nutrients limit vegetation β but do not allow separating limitations by individual nutrients. Vegetation β from CO2 experiments is based on experiments with short-term CO2 exposure of young plants, most of them not exceeding a year of treatment (Figure 1b). Much slower carbon and nutrient dynamics are expected in slow-growing mature forests, and the full response of vegetation to CO2, as well as secondary effects, may only develop over a longer time period (Quesada et al., 2012). Plants may slowly improve nutrient use and/or upregulate nutrient acquisition in response to eCO2 with time, potentially also with associated changes in species composition. Seedlings and young plants in CO2 experiments may also exhibit stronger responses than mature trees, due to their faster growth, and in particular in late succession forests where low vegetation growth responses are expected (Walker et al., 2021). Our findings point to limited CO2-induced vegetation growth in the tropics when considering soil nutrient availability. We show that the normalized CO2 effect on tropical vegetation carbon was c. 70% lower in seedling CO2 experiments without fertilizers, and c. 50% and 70% lower in models that consider nitrogen and phosphorus, from two model ensembles. The inadequate or lacking representation of nutrient cycles in models likely leads to overestimating CO2 effects on tropical vegetation growth. If nutrients would limit the CO2 fertilization effect in Earth System Models, climate-induced vegetation losses (which may also be underestimated) would be less offset and potentially turn the vegetation into a carbon source. Next to nutrient limitation on vegetation growth, the projection of a strong CO2 fertilization effect on the tropical land carbon sink may additionally be compromised by the omission of negative feedbacks from vegetation turnover, and nonlinearity between vegetation and soil carbon accrual under eCO2 (Terrer et al., 2021). Process-based modelers need to continue on the challenging task to incorporate our process-based understanding of carbon and nutrient dynamics in tropical forests in meaningful models, and their integration in Earth System Models. In parallel, there is an urgent need for long-term tropical CO2 experiments in mature forests for capturing relevant processes of terrestrial carbon and nutrient feedbacks at relevant scales. Katrin Fleischer and César Terrer conceived the study and wrote the manuscript together. The authors acknowledge the work of many people that were involved in conducting and analyzing CO2 experiments, and who worked on developing models and performing model simulations that made this synthesis possible. The authors further explicitly thank Huanyuan Zhang for the extraction of CMIP6 model results, and Sönke Zaehle and Alexander Winkler for constructive comments on a previous version of the manuscript. The authors declare no conflict of interest. CMIP6 model simulations are freely accessible from https://esgf-index1.ceda.ac.uk/search/cmip6-ceda/, meta-data and responses from CO2 experiments are accessible on https://zenodo.org/record/6913071#.YuEoVC-B0Q0 (doi: https://doi.org/10.5281/zenodo.6913071), R code to synthesize data and produce the figures displayed in this paper can be accessed in https://github.com/cesarterrer/tropical-CO2.

Topics & Concepts