the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A model of the within-population variability of budburst in forest trees
Jianhong Lin
Daniel Berveiller
Christophe François
Heikki Hänninen
Alexandre Morfin
Gaëlle Vincent
Rui Zhang
Cyrille Rathgeber
Nicolas Delpierre
Abstract. Spring phenology is a key indicator of temperate and boreal ecosystems’ response to climate change. To date, most phenological studies have analyzed the mean date of budburst in tree populations while overlooking the large variability of budburst among individual trees. The consequences of neglecting the within-population variability (WPV) of budburst when projecting the dynamics of tree communities are unknown. Here, we develop the first model designed to simulate the WPV of budburst in tree populations. We calibrated and evaluated the model on 48,442 budburst observations collected between 2000 and 2022 in three major temperate deciduous trees, namely, hornbeam (Carpinus betulus), oak (Quercus petraea) and chestnut (Castanea sativa). The WPV model received support for all three species, with a root mean square error of 5.6 ± 0.3 days. Retrospective simulations over 1961–2022 indicated earlier budburst as a consequence of ongoing climate warming. However, simulations revealed no significant change for the duration of budburst (DurBB, i.e., the time interval from BP20 to BP80, which respectively represent the date when 20 % and 80 % of trees in a population have reached budburst), due to a lack of significant temperature increase during DurBB in the past. This work can serve as a basis for the development of models targeting intra-population variability of other functional traits, which is of increasing interest in the context of climate change.
- Preprint
(1960 KB) - Metadata XML
-
Supplement
(899 KB) - BibTeX
- EndNote
Jianhong Lin et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-1075', Marc Peaucelle, 24 Aug 2023
Summary:
In this study, Lin and co-authors developed a new phenology model to simulate the within-population variability of budburst in tree populations. Their model was calibrated and evaluated on 3 species and two French sites over 2000 -2022. Model simulations show good agreement with observations, both in simulating the average budburst date as well as the duration of budburst. The authors applied their model over 1961-2022 which simulated earlier budburst with climate warming, a result which is consistent with observations and current knowledge. Same simulations did not show significant evolution of the duration of budburst due to the lack of significant temperature increase during budburst.
Major comments:
The authors provide an interesting study and a new solution to address within population variability. Overall, the manuscript is clear and the methodology is appropriate. Results are well discussed in the light of the current knowledge. I only have some comments regarding the methodology and analyses that could strengthen the clarity and robustness of the analysis:
1) The way the model was calibrated is unclear. At L. 205, the authors mentioned a calibration of model parameters by minimizing RMSEtot. First, the optimization algorithm is not described here. I suggest the authors to describe that part (e.g. was it done with a gradient approach? Which package was used for that task? How a priori set of parameters were defined? How possible equifinality was handled?). Second, I believe there is a problem with the definition of RMSE from eq. 12 : RMSEtot = RMSEBP + RMSEDOY . The authors sum two metrics with different units (RMSEBP is in %, RMSEDOY in days) which, as it is currently described, is wrong. Apart from the unit problem, minimizing a summed RMSE does not seem to be the best approach. If I understand, the goal here is to minimize both RMSEBP and RMSEDOY. However, the same minimum RMSEtot can be achieved with multiple combinations of RMSEBP and RMSEDOY. I suggest the authors to have a multicriteria optimization instead of trying to minimize a combined metrics.
2) To continue with the methodology, I wonder why the authors considered the Jan-May temperature to compute the sensitivity? It means that they account for extra temperature for early budburst compared to late budburst, which creates biases in the analysis. I would suggest the authors to considers the preseason temperature in order to avoid this bias (for example the average temperature of the month preceding budburst).
3) The authors hypothesized that F* follows a normal distribution (L. 274, L. 329) and try to validate that hypothesis with indirect comparisons to BP (%) in Fig. 5. This is a key aspect of their work and I believe that the distribution of F* can be directly extracted from observations by computing F* for each observation. Having the “observed” distribution in F* would clearly strengthen the results and the discussion.
4) To continue with F*, the authors discuss L. 336-339, that the variability in the timing of budburst can result from the genetic diversity and microclimate. Following my previous suggestion regarding the computation of “observed” F* for each observation, looking at the evolution of F* for each individual could help in discussing that aspect. If the data allow it, showing if the same trees are always early/late within the population, or if it changes from year to year along with F* could help in quantifying that genetic vs. microclimate variability and provide some answers here.
5) It seems that the model is less performant for Castanea (Fig 5c and 7b) compared to other species. Is it linked to that null g parameter? DurBB seems to be well captured for the two other species and not Castanea, what could explain the variability in durBB then? Is it only linked to differences in F* distribution? or extreme temperature? Also the authors discussed the distribution in F*, could we imagine the same distribution for CCrit and what would it imply? Discussing these points could clarify the key message of the manuscript.
6) Finally, it would be very helpful to provide the scripts that were used to process the data, generate the analysis and figures for reproducibility of the results. The script currently archived on Zenodo provides the key functions to run the model but needs substantial effort to reproduce the results.
I hope these comments will help the authors in improving their nice manuscript.
Best regards,
Marc Peaucelle
Citation: https://doi.org/10.5194/egusphere-2023-1075-RC1 -
RC2: 'Comment on egusphere-2023-1075', Yongshuo H. Fu, 18 Sep 2023
General comments
The spring budburst in plants exhibits sensitivity to climate change. In this manuscript, the authors built the within-population variability (WPV) model based on a two-phase parallel model framework, to simulate the budburst in tree populations. Model performance of the WPV model has improved, rendering this research highly significant. I do like this study and it would be a nice contribution to the phenology modeling, and I have some concerns regarding the structure of the manuscript. Firstly, the introduction section need to be improved by adding the mechanisms of the dormancy release process and the introduction of the original model. Secondly, it is recommended to make appropriate deletions and adjustments to make the logic clearer for the whole manuscript, especially the content related to WPV in the introduction section. Additionally, the discussion section suggests adding species-specific effects on budburst and better explaining the reasons for considering intra-population variability to predict budburst. Therefore, a major revision is recommended.
Specific comments
- L28-30: Moisture (precipitation or air humidity) is also an important factor affecting the timing of leaf phenology in spring.
- L55-67: These two paragraphs could be merged.
- L62-67: ‘the increasing warming rate during the budburst period’? Note that in the cited reference it is: ‘slower warming rates during the budburst period’. Please carefully review and correct the citations throughout the manuscript to ensure they are accurately cited.
- L137-140: Why is ‘budburst percent’ in italics? Also, 'erf' should be in italics in the formula.
- L155-156: What does T(t) mean? Symbols that appear for the first time should be interpreted clearly.
- L172-209: For the performance of the MPV model, please add R and p values to further verify the model accuracy. Also, what does ‘i-th’ mean in the L209.
- L251-258: Please add the values of R and p to the model performance and explain them.
- L378-386: Add some discussion about the effects of changes in BP20, BP80, and DurBB on the temporal niche, coexistence mechanism, and carbon sink capacity of tree populations.
- L413-414: Other environmental factors affecting budburst should also include the interaction between environmental factors.
- Table 1: Please correct ‘48491° N’.
- 1: ‘Bud burst’ in the title of the ordinate? Budburst.
- 2 and Fig. 3: Please add R and p values.
- 6: What does each label represent? Please specify.
- 7: The legend is missing. Also, what does each label represent? Please specify.
Citation: https://doi.org/10.5194/egusphere-2023-1075-RC2
Jianhong Lin et al.
Data sets
the data for WPV model for budburst Eric Dufrene, Nicolas Delpierre, Gaëlle Vincent, Alexandre Morfin, Daniel Berveiller, and Jianhong Lin https://doi.org/10.5281/zenodo.7962840
Model code and software
code for wpv model (model simulating the within-population of budburst in tree populations) Lin Jianhong https://doi.org/10.5281/zenodo.7957944
Jianhong Lin et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 126 | 29 | 12 | 167 | 23 | 7 | 6 |
- HTML: 126
- PDF: 29
- XML: 12
- Total: 167
- Supplement: 23
- BibTeX: 7
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1