Overestimated nitrogen loss from denitrification for natural terrestrial ecosystems in CMIP6 Earth System Models

Maoyuan Feng, Shushi Peng, Yilong Wang, Philippe Ciais, Daniel S. Goll, Jinfeng Chang, Yunting Fang, Benjamin Z. Houlton, Gang Liu, Yan Sun, Yi XiNature Communications 2023 · DOI

TL;DR

Earth System Models (ESMs) used for climate projections significantly overestimate nitrogen loss through denitrification, nearly doubling the actual amount. This overestimation leads to an exaggerated view of nitrogen limitation on plant growth, potentially skewing our understanding of how ecosystems will respond to rising CO2.

What Problem Does This Paper Address?

Accurately modeling the global nitrogen (N) cycle is crucial for understanding how terrestrial ecosystems will respond to climate change, particularly rising atmospheric CO2. Nitrogen is a key nutrient that regulates plant growth. However, quantifying N losses from ecosystems, especially through denitrification (the microbial conversion of nitrate to gaseous nitrogen), is challenging. These losses are poorly constrained in current Earth System Models (ESMs). Previous studies have suggested that some ESMs, like the Community Land Model (CLM) used in CMIP5, might overestimate denitrification. This paper investigates whether this issue persists in the latest generation of ESMs (CMIP6) and quantifies the extent of this overestimation, exploring its implications for predicting plant responses to elevated CO2.

Methods

The researchers employed an "isotope-benchmarking" method to estimate natural soil denitrification N loss. This approach leverages the natural abundance of the nitrogen isotope 15N (denoted as δ15N) in soils.

  1. Global Soil δ15N Map: They compiled a large dataset of 5887 direct soil δ15N measurements from natural ecosystems. Using a Random Forest (RF) machine learning model, they created a high-resolution (0.1° x 0.1°) global map of soil δ15N. This RF model significantly improved upon previous methods by achieving a higher R² (0.92) and lower RMSE (0.77%) in predicting soil δ15N.
  2. Isotope Mass Balance: They then used this δ15N map and established isotope mass balance equations to derive a global map of the "fraction of denitrification" (fdenit). This fraction represents the proportion of total N loss attributed to denitrification. This derivation involved considering various N input sources (rock weathering, atmospheric deposition, biological nitrogen fixation) and their respective isotopic signatures.
  3. ESM Comparison: Finally, they compared their isotope-benchmarking-derived fdenit with the fdenit simulated by 13 CMIP6 ESMs. They also used their derived fdenit to re-allocate the total N losses simulated by ESMs into denitrification and leaching components.

Datasets

  • Soil δ15N Observations: A global dataset of 5887 direct measurements from natural ecosystems (compiled from Craine et al. and Sena-Souza et al.).
  • Predictor Variables for RF Model:
    • Climate drivers: Precipitation (P), temperature (T), aridity index.
  • Soil properties: Bulk density, soil pH, clay, silt, sand fractions, organic carbon, C:N ratio.
  • Microbial symbiont abundance: Arbuscular mycorrhizal (AM), ectomycorrhizal (ECM) fungi, and N-fixing bacteria (Nfix).
  • Gross Primary Production (GPP).
  • Atmospheric deposition (NHx and NOy).
  • CMIP6 ESM Outputs: Historical simulation outputs for nitrogen fluxes (denitrification, leaching) and land N stocks from 13 ESMs.
  • N Input Datasets: Multiple global maps for N deposition, biological N fixation (BNF), and rock weathering were used to account for uncertainties.

Key Insights and Ideas

  • Persistent ESM Overestimation: The study confirms that a significant overestimation of denitrification N loss is a persistent problem in ESMs, even in the latest CMIP6 generation, suggesting limited improvement since CMIP5.
  • Isotope Benchmarking as a Powerful Tool: The refined isotope-benchmarking method, utilizing a large observational dataset and machine learning, provides a more robust estimate of natural denitrification rates, serving as a critical benchmark for ESM evaluation.
  • Exaggerated N Limitation: The overestimation of denitrification in ESMs leads to an underestimation of available soil nitrogen, which in turn exaggerates the predicted N limitation on plant growth, particularly in boreal regions. This has direct implications for how ESMs project the future terrestrial carbon sink.
  • Implications for Climate Projections: Biases in representing N losses can propagate into climate feedback loops, potentially affecting the accuracy of future climate projections.

Key Results

  • Denitrification Overestimation: CMIP6 ESMs simulate a global denitrification N loss of 73 ± 31 Tg N yr⁻¹, which is almost double the isotope-benchmarking estimate of 38 ± 11 Tg N yr⁻¹. This represents a 92% overestimation.
  • Leaching Underestimation: Conversely, CMIP6 ESMs underestimate leaching N loss by 62%.
  • Spatial Patterns: The isotope-benchmarking derived fdenit shows a latitudinal gradient, decreasing from tropics to high latitudes, consistent with empirical knowledge. Many CMIP6 ESMs, however, simulate a more uniform fdenit globally, often close to 1, which contradicts the known dependence of denitrification on environmental factors like temperature and soil moisture.
  • Impact on Plant Growth: In boreal regions, a negative correlation was found between the sensitivity of Net Primary Production (NPP) to CO2 (BNPP) and simulated fdenit. ESMs with higher fdenit values tended to underestimate BNPP, suggesting an exaggerated N limitation on plant growth in response to elevated CO2.

Limitations and Future Work

  • Uncertainty in Isotope Effects: While the study used a conservative estimate for the isotope effect of denitrification (Edenit), variations in this parameter can influence the derived fdenit. The authors explored sensitivity to Edenit and found that higher values would lead to even greater overestimation by ESMs.
  • Steady-State Assumption: The isotope-benchmarking was primarily derived under a steady-state assumption (N inputs = N outputs). While the authors showed that terrestrial N sinks have a minor impact (<1%) on fdenit, this assumption might still introduce some bias.
  • Model Complexity: The study highlights that denitrification and related N cycle processes are still poorly represented in ESMs. Future work should focus on improving these representations by incorporating recent advances in understanding microbial processes, soil conditions, and plant-soil interactions.
  • Refined N Loss Partitioning: The authors suggest their isotope-benchmarking approach could be extended to better partition gaseous N losses into specific components (N2O, NO, N2), leading to more refined ESM assessments.
  • Cross-Constraint: The study emphasizes the need to cross-constrain ESMs using both δ15N data and observations of N fluxes in stream and river discharges.

Summary updated

Abstract

Denitrification and leaching nitrogen (N) losses are poorly constrained in Earth System Models (ESMs). Here, we produce a global map of natural soil <sup>15</sup>N abundance and quantify soil denitrification N loss for global natural ecosystems using an isotope-benchmarking method. We show an overestimation of denitrification by almost two times in the 13 ESMs of the Sixth Phase Coupled Model Intercomparison Project (CMIP6, 73 ± 31 Tg N yr<sup>-1</sup>), compared with our estimate of 38 ± 11 Tg N yr<sup>-1</sup>, which is rooted in isotope mass balance. Moreover, we find a negative correlation between the sensitivity of plant production to rising carbon dioxide (CO<sub>2</sub>) concentration and denitrification in boreal regions, revealing that overestimated denitrification in ESMs would translate to an exaggeration of N limitation on the responses of plant growth to elevated CO<sub>2</sub>. Our study highlights the need of improving the representation of the denitrification in ESMs and better assessing the effects of terrestrial ecosystems on CO<sub>2</sub> mitigation.
BibTeX
@inproceedings{feng2023overestimated,
  author = {Maoyuan Feng and Shushi Peng and Yilong Wang and Philippe Ciais and Daniel S. Goll and Jinfeng Chang and Yunting Fang and Benjamin Z. Houlton and Gang Liu and Yan Sun and Yi Xi},
  title = {{Overestimated nitrogen loss from denitrification for natural terrestrial ecosystems in CMIP6 Earth System Models}},
  booktitle = {Nature Communications},
  year = {2023},
  doi = {10.1038/s41467-023-38803-z},
}

Loading related content...

Figure