the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Tropical Dry Forest Response to Nutrient Fertilization: A Model Validation and Sensitivity Analysis
Shuyue Li
Bonnie G. Waring
Jennifer S. Powers
Abstract. Soil nutrients, especially nitrogen (N) and phosphorus (P), regulate plant growth and hence influence carbon fluxes between the land surface and atmosphere. However, how forests adjust biomass partitioning to leaves, wood, and fine roots in response to N and/or P fertilization remains puzzling. Recent work in tropical forests suggests that trees increase fine root production under P fertilization, but it is unclear whether mechanistic models can reproduce this dynamic. In order to better understand mechanisms governing nutrient effects on plant allocation and improve models, we used the nutrient enabled ED2 model to simulate a fertilization experiment being conducted in a secondary tropical dry forest in Costa Rica. We evaluated how different allocation parameterizations affected model performance. These parameterizations prescribed a linear relationship between relative allocation to fine roots and soil P concentrations. The slope of the linear relationship was allowed to be positive, negative, or zero. Some parameterizations realistically simulated leaf, wood and fine root production, and these parameterizations all assumed a positive relationship between relative allocation to fine roots and soil P concentration. On a thirty-year timescale, under unfertilized conditions, our model predicted the largest aboveground biomass (AGB) accumulation when relative allocation to fine roots was positively related to soil P concentration. However, this result was mostly driven by increased water use rather than decreased nutrient limitation. On a thirty-year timescale with P fertilization, the assumption of a positive correlation between relative allocation to fine roots and soil P concentration led to over-investment to fine roots and reductions in vegetation biomass. Our study demonstrates the need of simultaneous measurements of leaf, wood, and fine root production in nutrient fertilization experiments. Models that do not accurately represent allocation to fine roots may be highly biased in their simulations of AGB, especially when simulating a range of sites with significantly different soil P concentrations.
- Preprint
(1422 KB) - Metadata XML
- BibTeX
- EndNote
Shuyue Li et al.
Status: final response (author comments only)
-
RC1: 'Comment on bg-2022-243', Anonymous Referee #1, 14 Jan 2023
General comments
This study integrated a representation of phosphorus-dependent relative allocation to root tissues into the ED2 model. The model was simulated at a site in a tropical dry forest in Costa Rica for +N, +P, and +NP fertilization treatments. Modelled results were compared to empirical observations. The model was then simulated over 30 years to examine the influence of the new process representation over a longer time scale. The comparison between a model and empirical observations of experimental manipulations of nutrient input is very useful for model development. However, it was unclear whether the process that was represented in this study (increasing allocation to fine roots with increasing soil P) is prevalent in systems outside of this site, what its underlying mechanisms are, and how it relates to other central processes (such as the relationships between allocation to fine roots and soil nitrogen or water). Furthermore, the statistics used to establish the results were unclear.
Specific comments
- The premise of this study needs to be better established. What is the mechanism underlying increasing allocation to fine roots with increasing soil P? Does this response occur in other ecosystems or are the only observations from the Costa Rica site? The introduction gave several examples of how different ecosystems and different individuals within a given ecosystem respond differently to N and P fertilization. Are there any patterns that emerge across ecosystems? If this response is specific to a single or a small number of sites, why should it be represented in TBMs? Have empirical studies indicated that this is important for larger C fluxes? This is somewhat touched on in the Discussion but its prevalence was not clear.
- How do other factors interact to determine relative allocation to roots? Water and nitrogen should play important roles as well. Is it valid to only focus on P (especially given that the results suggest that it is increased water uptake that seemed to drive AGB)? Additionally, I would assume that the role of other plant mechanisms to increase P uptake would be important as well, such as phosphatase synthesis and arbuscular mycorrhizae. These are likely intricately linked to fine root biomass in real ecosystems. While these do not necessarily need to be examined or modelled, they should be at least recognized in the experimental setup and discussion. Additionally, flexible stoichiometry could be important. How have other models approached these phenomena?
- It was not made clear which PFT was being studied in these experiments. Were there multiple PFTs? Given that this is a dry tropical forest, do deciduousness and phenology play a role here? How could these results differ between tropical dry forests and tropical moist forests? Have similar experiments been conducted in tropical moist forests?
- Using different statistical analyses for leaf, wood, and root due to patterns that emerged from the observation-based data may not be the best approach. It would be a more direct comparison to use the same statistical analyses for each tissue because the biases in the empirical observations may not be present in the model outputs. Figure 4 is a central figure but it is unclear whether it shows only the control treatment or an average across treatments. Regardless, this analysis should be conducted for each fertilization treatment independently given that the premise of the study is that fertilization treatment influences relative allocation. Furthermore, are the temporal trends important here given that the same amount of fertilizer was applied each year and the experiment was only 3 years long? Given that the primary focus is the difference between tissues rather than the difference between years, it may make more sense to aggregate across years for each tissue / treatment.
- Is a 2 year spinup sufficient? Shouldn’t the spinup be run until an equilibrium is established?
Technical correction
Line 81 “While models have rarely be validated on these time scales” I would argue that models are often evaluated over the past several decades (1960s to present).
Production units should be kg m-2 yr-1.
Table 3 is challenging to interpret. Could this be transformed into a figure?
Include other parameterizations in Figure 5 (additional panels).
Include other treatments in Figure 8 (additional panels).
Figure 6: Clarify if this is averaged across treatments or if this is the control treatment only.
Citation: https://doi.org/10.5194/bg-2022-243-RC1 - AC1: 'Reply on RC1', David Medvigy, 20 Apr 2023
-
CC1: 'Comment on bg-2022-243', Katrin Fleischer, 03 Feb 2023
The premise of the study:
The authors confront a dynamic vegetation model with experimental data from a nutrient fertilization experiment in Costa Rica and assess the validity and consequences of alternative root allocation parameterization. The overarching question the authors address is how root allocation responds to changes in nutrient availability, which is a very valid question to pursue. Evaluating models and their underlying assumptions with direct field observations and experiments is necessary to advance our understanding and gain confidence in model predictions. I appreciate the effort the authors have undertaken with the demographic model ED2, and their joint effort with experimental findings and expertise to advance root dynamics and their interactions with nutrient availability. However, I currently cannot recommend publishing the manuscript due to some concerns.
Main concern:
- The ecological underpinnings of the chosen parameter sets in regard to fine root allocation are not sufficiently clear. This includes the ecological theory to justify choosing the parameter sets, the discussion of the consequences of these parameter sets, and placing the chosen parameter set and findings within the literature and previous modeling efforts.
- The study covers short-term and long-term process effects of the different parameterizations, whereby we can expect different outcomes. The experiment finds that fine root production increases with fertilization. That does not necessarily mean that higher fertile sites will be characterized by higher fine root production. The others touch on these aspects of timing in the discussion, and the significance of this difference is crucial for the study, however insufficient emphasis is placed on this in working out the premise of the study and discussing the findings. Associated with this, the experiment takes place in a young forest stand, the implications of this deserve more discussion.
- Direct nutrient acquisition via fine roots is only one of several possible mechanisms of how plants can acquire nutrients. The authors mention that the trees are associated with arbuscular mycorrhizae but the implications of this in regard to the outcome of the experiment, and how this might have affected the model performance are not addressed.
Additional comments methodology:
The chosen parameter sets reflect negative, constant, and positive relationships between fine root production and soil P. The negative one would reflect the resource-dependent parameterization, however not via internal plant demand and supply (as commonly done) but depending on external P supply, in a linear fashion. This is quite different from any of the previous model approaches, and the parameterizations are all based on a linear relationship between soluble soil P and root allocation. A discussion of the ecological underpinning of this model and the parameterization approach would be helpful. The authors touch upon the ecological theory they are addressing only in the discussion part.
Parameters a and b are coordinated to yield a similar 0.3 root-to-shoot ratio in control plots, but to different settings in fertilized plots. The reason for not testing different parameterization settings in the control plot was not clear to me, it would be helpful if the authors could elaborate on this.
The allocation process as a whole in ED2 is not sufficiently explained. The calculation of the daily leaf and root allocation, and the allometric equations to determine maximum leaf and root allocation, need to be described and included in the discussion. Similarly, the PFTs that are modeled and their parameterizations are not described.
Additional comments results:
One positive parameterization was the only one that agreed well with experimental observations over the 3-year period. Positive parameterizations led to the highest AGB over 30 years, however, not due to alleviation of nutrient limitation but water stress. It would be helpful if the reader would understand how that occurred process-wise. What was the actual allocation to fine roots? And why did that not alleviate nutrient stress? Similarly, with P fertilization, positive parameterizations led to too many fine roots over the 30-year time scale. As is described in the abstract as the main finding. Does that not indicate that the chosen parameterization works well for the short term but, importantly, not well in the long term? Is such an assumption about root allocation advisable then? It would be helpful if the authors go beyond “experiments need to measure leaf, root and wood production” in their main take-away for the paper, and rather elaborate on recommendations for model development.
Given that the experiment took place in a tropical dry forest, the anticipated interactions between soil fertility and water stress are intriguing and it would be helpful if the authors elaborate on the alleviation of water stress in the 30-year simulation that has been touched upon.
Maybe the agreement at the 3-year time scale is not sufficient to evaluate. Can the control plots not also be used to evaluate the parameterization, representing rather long-term dynamics? Should the parametrizations not be tested on the control and experimental plots, equally? The authors state that ED2 has been validated before. Did the constant parameterization work well before? Did ED2 employ a resource-dependent allocation scheme before? It would be helpful to outline which allocation approaches agreed well with observations and which did not in previous studies with ED2.
The resource limitation theory postulates plant allocation is adjusted to acquire the most limiting resource. Many ecosystem models to date, adopt a resource-dependent allocation scheme, so they would predict that more roots would be produced if soil nutrients were limited, irrespective of the return. The authors find that the experimental results contradict this hypothesis. I believe this apparent contradiction might be a time issue, so that roots grow to acquire the soluble P, while root allocation would decrease once P limitation is alleviated. It might also have to do with mycorrhizal interactions, previous outsourcing of phosphorus acquisition to mycorrhizae now becomes less beneficial with increased nutrient supply and plants switch to “do-it-yourself”. Even if we cannot be certain about any of these hypotheses, a discussion thereof would be helpful.
Line comments:
- 336 please specify: allocation parameterization sensitive to external nutrient availability
- 383- 341 what is the implication of that finding?
- 344 the model is microbial-explicit. That is an important aspect of the model. It would be helpful if this is included in the manuscript and discussed.
- 365 For future efforts, the authors can consider evaluating the model and different parameter sets at different locations in the tropical biome, and including an evaluation along a soil fertility gradient, as well as experimental changes. The combination of both would be helpful to discern short-term and long-term effects.
- 369 The experiment near Manaus found increased primary productivity in response to fertilization, indicating that production was limited by phosphorus. To my understanding, this finding is still consistent with the resource limitation hypothesis, since plants allocated carbon to roots to acquire the nutrients that they were in demand of.
- 375 As the authors note, roots are there for acquiring multiple resources at once. A discussion of these interactions would be helpful. The analogy to rain roots is interesting, the authors could elaborate here. The root production after fertilization could potentially be a similar short-term effect to acquire the limiting resource.
- 379 The paragraph on field observation is helpful, however, root stock and root production seem to be mixed up. Since the model evaluation deals with root production, the authors could elaborate on those aspects of the measurements.
- 387 It would be helpful if the authors elaborate on the interactions with mycorrhizae, here, and the implications for this study.
- 389 The “supply-limited” hypothesis, which is the basis of this study, should be introduced earlier and placed in context with the alternative hypotheses.
- 390 This section could benefit if it discussed what has been learned from this and the previous approaches.
- 398 Similarly, in this section, it would be helpful if the authors go beyond the summary of the results here and discuss the implications of these findings. See the main comments above.
- 424 Here it is crucial to discuss that resource-dependent model approaches are not exactly comparable to the negative parameterization applied in this study. See the main comments above.
Citation: https://doi.org/10.5194/bg-2022-243-CC1 - AC2: 'Reply on CC1', David Medvigy, 20 Apr 2023
-
RC2: 'Comment on bg-2022-243', Anonymous Referee #2, 23 Apr 2023
Shuyue Li et al., conducted a modeling study to investigate how plant allocation in response to nutrient fertilization. They used nutrient enabled ED2 model with various different parameterizations on biomass allocation under control and fertilized conditions over tropical dry forest. Data from an fertilization experiment at Costa Rica forest was used for model comparison and validation. The paper is well organized and presentation is smooth. Below I have a few suggestions and comments.
- introduction, first paragraph needs to be improved, Nutrient availability could affect plant activity in many different ways. The most relevant (to this paper) way is through mediation C/N/P allocation and biomass construction. However, the first paragraph try to explain how nutrient availability could affect plant response to CO2 enrichment, which is not much relevant here.
- introduction, paragraph 2 and 3 provide a nice summary of many fertilization experiments for tropical trees. However, each fertilization experiment was discussed individually. I would suggest adding some discussion about why and how experiment results differ from one another to improve the coherence of the summary.
- introduction, paragraph 5 and 6 highlight the need to investigate and improve the allocation scheme under long-term fertilization for current generation CNP models. In this case, a survey of allocation schemes used by current generation models are necessary, for example some models assume constant allocation, some assume multiple resource coordination, some are based on carbon cost …Besides the seven CNP models mentioned in this section, two more recent global CNP models are:FUN-CNP: Braghiere, R.K., Fisher, J.B., Allen, K., Brzostek, E., Shi, M., Yang, X., Ricciuto, D.M., Fisher, R.A., Zhu, Q. and Phillips, R.P., 2022. Modeling global carbon costs of plant nitrogen and phosphorus acquisition. Journal of Advances in modeling earth systems, 14(8), 2022MS003204. ELM-CNP: 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), 2238-2258.
- section 2.3, r2l is a function of soil P concentration (psol), I wonder mathematically will this equation lead to huge variability of r2l parameters especially at the time when fertilizers were applied. Maybe showing a figure of r2l during the 3 years of fertilization experiment will help to clarify this.
- section 2.4.1. Vegetation and soil are both initialized with in situ observations, rather than being determined by long-term spinup. Such approach often time will result in an dis-equilibrate vegetation and soil processes. Therefore, after initialization the vegetation and soil states will quickly changes towards quasi-equilibrate conditions, which could be largely different from the initialized conditions. I wonder if the re-equilibration also occur in ED2, how long does it re-equilibrate, and how that affect fertilization results?
- section 2.4.1. It was mentioned that fine root production was evaluated with linear regression, however, it also mentioned linear regression was not appropriate because there existed only three years of data. Here needs more clarification.
- Figure 2, most of the simulated variability of NO3, NH4 already exist in control run (solid blue lines), it doesn’t look like there were sudden increase of NH4 or NO3 right after the N fertilization. Also, it will be helpful, if the fertilization date could be marked on the x-axis.
- Section 4.1. It’s still not clear to me which parametrization is the best. It was stated that “only one of the 13 parameterizations that we tested was able to simultaneously simulate leaf, wood and fine root (missing word) production consistent with the observations”. Here, the screensful parameterization needs to be highlight. Also, in Figure 4, it doesn’t look like any parameterization was significantly superior to others.
Citation: https://doi.org/10.5194/bg-2022-243-RC2 -
AC3: 'Reply on RC2', David Medvigy, 25 May 2023
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2022-243/bg-2022-243-AC3-supplement.pdf
Shuyue Li et al.
Shuyue Li et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 340 | 125 | 19 | 484 | 4 | 5 |
- HTML: 340
- PDF: 125
- XML: 19
- Total: 484
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1