24 Feb 2021
24 Feb 2021
A novel representation of biological nitrogen fixation and competitive dynamics between nitrogen-fixing and non-fixing plants in a land model (GFDL LM4.1-BNF)
- 1Department of Ecology, Evolution and Environmental Biology, Columbia University, New York, 10027, USA
- 2Geophysical Fluid Dynamics Laboratory (GFDL), National Oceanic and Atmospheric Administration (NOAA), Princeton, 08540, USA
- 3Department of Ecology and Evolutionary Biology, Princeton University, Princeton, 08544, USA
- 4Department of Integrative Biology, The University of Texas, Austin, 78712, USA
- 1Department of Ecology, Evolution and Environmental Biology, Columbia University, New York, 10027, USA
- 2Geophysical Fluid Dynamics Laboratory (GFDL), National Oceanic and Atmospheric Administration (NOAA), Princeton, 08540, USA
- 3Department of Ecology and Evolutionary Biology, Princeton University, Princeton, 08544, USA
- 4Department of Integrative Biology, The University of Texas, Austin, 78712, USA
Abstract. Representing biological nitrogen fixation (BNF) is an important challenge in the incorporation of nitrogen (N) cycling in land models. Initial representations of BNF in land models applied simplified phenomenological relationships. More recent representations of BNF are mechanistic and include the dynamic response of BNF to N limitation of plant growth. However, they generally do not include the competitive dynamics between N-fixing and non-fixing plants, which is a key ecological mechanism that determines ecosystem-scale symbiotic BNF. Furthermore, asymbiotic BNF is generally not included in land models. Here, we present LM4.1-BNF, a novel representation of BNF (asymbiotic and symbiotic) and an updated representation of N cycling in the Geophysical Fluid Dynamics Laboratory Land Model 4.1 (LM4.1). LM4.1-BNF incorporates a mechanistic representation of asymbiotic BNF by soil microbes, a representation of the competitive dynamics between N-fixing and non-fixing plants, and distinct asymbiotic and symbiotic BNF temperature responses derived from corresponding observations. LM4.1-BNF makes reasonable estimations of major carbon (C) and N pools and fluxes and their temporal dynamics, in comparison to the previous version of LM4.1 with N cycling (LM3-SNAP) and to previous representations of BNF in land models generally (phenomenological representations and those without competitive dynamics between N-fixing and non-fixing plants and/or asymbiotic BNF). LM4.1-BNF can be applied to project the dynamic response of vegetation to N limitation of plant growth and the degree to which this will constrain the terrestrial C sink under elevated atmospheric CO2 concentration and other global change factors.
Sian Kou-Giesbrecht et al.
Status: final response (author comments only)
-
RC1: 'Comment on bg-2020-483', Anonymous Referee #1, 19 Mar 2021
General comments:
Kou-Giesbrecht et al., presented a new version of GFDL land model that considers symbiotic fixation, competitive interactions between fixer and non-fixer plants. The model development is convincing and well explained, while the following up model evaluation needs to be improved, more discussions are needed to clarify the simulated C/N dynamics. Below are my major comments and some specific suggestions.
1. Since the major development is N2 fixation, it will be good to emphasize the model evaluation more on N2 fixation. Figure 6 showed that the LM4.1-BNF capture asymbiotic fixation rate for the mature forests, however, the LM4.1-BNF performed worse than LM3-SNAP (or maybe equal) for simulated symbiotic BNF rate along with the forest successional development.
One strategy could be using a part of the data to first tune the model parameters that are related to N2 fixation processes (e.g., rNfix) and then use the rest for evaluation purposes.
A deep dive into LM4.1-BNF simulated N2 fixation rate is further needed. For example, what are the temporal dynamics of Nodule biomass, NSC, Nstress, soil microbe biomass, how they drive the changes of symbiotic and asymbiotic BNF?
2. The LM4.1-BNF simulated C cycle is convincing (e.g., biomass, GPP, NPP), however, the N cycle showed largely bias, compared with observations (e.g., Figure 7). LM4.1-BNF has too much inorganic N (both NH4 and NO3), while very low soil organic N. Does it imply that the soil nitrogen immobilization rate and soil organic matter formation rate are largely underestimated? Or the soil organic matter C:N stoichiometry was problematic?
LM4.1-BNF had reasonable NO3 leaching loss, but overestimated N2O emission, which means the system is too open/leaky. Is it part of the reason why less available inorganic nitrogen is incorporated into soil organic nitrogen pool?
3. Model structure change versus parameterization. LM4.1-BNF improve the existing model structure and showed some improvement in model performance at the site level. But 1) how much of the performance improvement could be achieved just by tuning existing parameters; 2) how much of additional model uncertainty is introduced by the model structure changes and by adding new model parameters? I would suggest running some sensitivity tests (e.g., perturb critical model parameters) for both LM4.1-BNF and LM3-SNAP to fully answer those questions.
Specific comments:
P1L13 is important for nitrogen enabled land model
P2L47 another group of land models simulates N2 fixation with nitrogenase enzymatic activity (e.g., Vmax, KM kinetic parameters), root nodulation, and temperature constraint (CABLE, Wang 2007; ELM-ECA Zhu 2019).
P2L63 downregulation of N fixation under weak N limitation must be discussed together with upregulation of N fixation under strong P limitation. Otherwise, the model will simulate an extremely low N fixation rate over tropical ecosystems that has abundant N.
P3 L70-77. It will be good to add some discussion about the importance of asymbiotic N2 fixation so that the existing land model should include this flux. For example, how much of global N2 fixation is from asymbiotic fixation, how efficient can plants get nitrogen from asymbiotic pathway.
P4L101. It will make the model evaluation more concrete if the model could run at least two sites (tropical site with abundant soil N versus the temperate site with relatively strong N limitation).
P4L108-117 belongs to section 2. model description
P4L122 Does LM4.1-BNF have to run at fixed spatial resolution (roughly 1degree by 1 degree)? Or it could run at any resolution? How about the temporal resolution?
P5L138-140. What are fine root, sapwood, heartwood, seed C:N ratios, how about soil organic pools C:N ratios?
L6 Eq1,2. How to parameterize rNfix? What is the functional shape of f(T)?
P7L180 One microbe pool for all soil microbial activities? Please justify.
P7L185 How to parameterize rNfixasymb? What is the functional shape of f(T)?
P11L312 Did LM4.1-BNF-NPP and LM4.1-BNF-ET restart from steady-state of LM4.1-BNF spinup simulation? Will LM4.1-BNF spinup steady-state differ dramatically from LM4.1-BNF-NPP(/ET) steady-state?
P13L365 Here need more discussion about the model bias in dbh growth rate. For example, is that because N stress is too weak?
P16Figure3 The y-axis scale is misleading (suggest not using log-scale). The model vs data difference is actually big. Again, discuss why and how it occur in LM4.1 model. How is the LM4.1-SNAP simulated density, compared with FIA data?
P25L480 and Figure 10. What’s the implication of zero BNF rate during the late succession when the model initialized with mixed Acer and Robinnia? The whole ecosystem will rely on soil mineralization generated inorganic N? Will all plants gradually die?
Reference
Wang, Y.P., Houlton, B.Z. and Field, C.B., 2007. A model of biogeochemical cycles of carbon, nitrogen, and phosphorus including symbiotic nitrogen fixation and phosphatase production. Global Biogeochemical Cycles, 21(1).
Zhu, Q., Riley, W.J., Tang, J., Collier, N., Hoffman, F.M., Yang, X. and Bisht, G., 2019. Representing nitrogen, phosphorus, and carbon interactions in the E3SM land model: Development and global benchmarking. Journal of Advances in Modeling Earth Systems, 11(7), pp.2238-2258.
-
RC2: 'Comment on bg-2020-483', Anonymous Referee #2, 29 Mar 2021
In this study, Kou-Giesbrecht reported the GFDL LM4.1-BNF model with a new representation of biological nitrogen fixation and evaluated the impact of competition between nitrogen-fixing and non-fixing plants on simulated carbon, nitrogen and demographic dynamics in a temperate forest site. They showed the LM4.1-BNF did a fair job in simulating the many lumped variables reported in the US forest inventory and analysis database for a temperate forest site at Coweeta Hydrologic Laboratory in North Carolina. Particularly, they showed that the competition between N-fixing and non-N fixing plants is an important factor in interpreting the dynamics of carbon accumulation. Overall, the paper is clearly written, but there are some issues need to be resolved before the model is able to be considered as doing sufficiently well.
More details are listed below
The abstract is generally OK; however, it lacks details on the model performance. The authors may consider to add more content from their model evaluation against the observational data.
Technical description is long but written well.
For results, my major complain is the authors have yet to demonstrate the model LM4.1-BNF compares well with high frequency temporal data, such as eddy flux measurement of carbon and water fluxes. The comparison with lumped data is fair, but not great. This is especially important to evaluate the effect of new temperature response function. Perhaps the authors should consider applying the model for a site with eddy flux measurements as well?
Further the discussion is a little bit detached from results. Authors may consider move some of the analysis into discussion to better explain the significance of updated processes.
Code availability: some of my colleagues say “upon request” is a bad exercise. Authors should at least provide whom to send such a request, or provide a web link to send such a request.
Other comments
Table 2. Missing group separation between 2nd and 3rd sets of analyses?
Figure 2. Why does the model under predict the low dbh growth rates?
Figure 4. Use stronger color contrast between two FIA data? Or maybe even different symbols? Currently, it is not easy to differentiate them.
Figure 6. What is the uncertainty of the data points? Also, why does LM3-SNAP predict more evident oscillations?
Figure 9a: what happened to LM4.1-BNFNPP? Why its time series is much shorter?
Equation (A1), why is there no adsorption effect considered? The behavior of NH4 and NO3 are quite different in soil.
Sian Kou-Giesbrecht et al.
Sian Kou-Giesbrecht et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
235 | 50 | 9 | 294 | 2 | 5 |
- HTML: 235
- PDF: 50
- XML: 9
- Total: 294
- BibTeX: 2
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1