the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jul 2023
Optimizing maximum carboxylation rate for North America’s boreal forests in the Canadian Land Surface Scheme Including Biogeochemical Cycles (CLASSIC) v.1.3
Abstract. The maximum carboxylation rate (Vcmax) is an important parameter for the coupled simulation of gross primary production (GPP) and evapotranspiration (ET) in terrestrial biosphere models (TBMs) such as the Canadian Land Surface Scheme Including biogeochemical Cycles (CLASSIC). Observations of Vcmax show it to vary both spatially and temporally, but it is often prescribed as constant in time and space for plant functional types (PFTs) in TBMs, which introduces large errors over North America’s boreal biome. To reduce this uncertainty, we used a Bayesian algorithm to optimize Vcmax25 (Vcmax at 25 °C) in CLASSIC against eddy covariance observations at eight mature boreal forest stands in North America for six representative PFTs (two trees, two shrubs, and two herbs). As expected, the simulated GPP and ET using the optimized parameters generally obtained reduced root mean square deviation values compared with eddy covariance observations and corresponding stand-level estimates obtained from gridded global data products. The optimized Vcmax25 values for each PFT compared reasonably well with reported estimates derived from leaf-level gas exchange measurements. However, a large spatial variability of Vcmax25 was identified, especially for the shrub and herb PFTs. We found that the site characteristics, particularly latitude for the shrub PFTs and air temperature for evergreen needleleaf tree, explained much of the spatial variability, providing a basis to improve Vcmax25 parameterizations in TBMs at regional scales.
- Preprint
(824 KB) - Metadata XML
-
Supplement
(505 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2023-1167', Anonymous Referee #1, 22 Sep 2023
This study utilized a Bayesian algorithm to optimize Vcmax25 in the land surface model against eddy covariance observations at eight mature boreal forest stands in North America. The results showed that the Bayesian algorithm can optimize Vcmax25 and improve ET and GPP estimates. The topic is interesting, and the results look promising. However, I am not convinced that the manuscript is innovative enough to contribute to the development of physical models, and therefore, I cannot accept it for publication.
1) Data assimilation methods have been widely employed in the optimization of Vcmax25 (He et al., 2019). Data assimilation methods can improve the estimation of vegetation photosynthesis by assimilating remote sensing SIF data at regional or global scales. Although the results of this manuscript are reliable, I do not believe that site-level optimization can be extrapolated to regional or global scales. The expansion from site-specific to regional scales is crucial for the development of physical models.
2) The authors need to provide more details about the model, including meteorological data and auxiliary data. Additionally, what is the timescale of Vcmax25 optimization? Is it on a daily, monthly, or throughout the entire growing season? Which years' observational data were used for optimizing Vcmax25? And which years' observational data were used for the spin-up? The site name, vegetation type, and other key information should be listed in the manuscript.
3) This study lacks independent validation. Eddy covariance observations of ET and GPP were used to optimize Vcmax25. Subsequently, the optimized Vcmax25 estimates of ET and GPP were further compared with eddy covariance observations. Although some global gridded products have been used to assess model simulation results, these gridded products exhibit uncertainty, and their observational footprints do not align with eddy covariance observations.
4) In this study, the random forest method was employed to characterize the relative importance of various influencing factors on Vcmax25. The limited optimization of Vcmax25 values in this study may lead to overfitting or underfitting issues in the machine learning method. This will impact the credibility of the relative importance results.
He, L., Chen, J. M., Liu, J., Zheng, T., Wang, R., Joiner, J., Chou, S., Chen, B., Liu, Y., and Liu, R.: Diverse photosynthetic capacity of global ecosystems mapped by satellite chlorophyll fluorescence measurements, Remote Sens. Environ., 232, 111344, https://doi.org/10.1016/j.rse.2019.111344, 2019.
Citation: https://doi.org/10.5194/egusphere-2023-1167-RC1 -
RC2: 'Comment on egusphere-2023-1167', Anonymous Referee #2, 06 Oct 2023
Vcmax is an important parameter for TBMs. Simply setting Vcmax25 in TBMS induces the uncertainty of TBMs. In this paper, authors conducted a Bayesian algorithm to optimize Vcmax25 in CLASSIC against eddy covariance observations at eight mature boreal forest stands in North America for six representative PFTs and identified the spatial variability of Vcmax25. This paper try to explore the Vcmax change in boreal forests. However, l am major concerned about optimizing strategy used in this study .
1. How to use TPE to optimize Vcmax25 was not explained in section 2.3. Vcmax was optimized in what time scale, yearly or daily? And in single-site optimization, for example, the observations were only the GPP and ET of the whole site. Does this result in ill-fitting problems when optimizing Vcmax for multiple PFTs simultaneously?
2. The study does not include a sensitivity analysis of the parameters. And through the Farquhar equation, the relationship between GPP and Vcmax is easier to understand. Why choose ET to optimize Vcmax needs more explanation in model structure. To my knowledge, ET is sensitive to the parameters that control stomatal conductance change.
3. To avoid attributing all uncertainties of simulation results to Vcmax25, I suggest optimizing several key parameters of the carbon and water cycles together.
Minor comments
1. In section 2.2, a map of the distribution of the eight mature boreal forest stands is needed.
2.How many site-years were used in optimization needs to be described in Table S2.
3. Generally, in site optimization, the details of simulated results against observations before and after optimization should be shown as a series of figures on a daily time scale. These figures are important for audiences to understand the improvement of model performance.
Citation: https://doi.org/10.5194/egusphere-2023-1167-RC2 -
AC1: 'Author responses to anonymous referees #1 and #2', Bo Qu, 01 Feb 2024
We thank the two anonymous referees for their time and efforts in reviewing our manuscript and providing constructive comments that have helped us improve it. We have carefully
considered the comments and made the revisions to our manuscript. Please see the attached document for our detailed responses and revisions.
Status: closed
-
RC1: 'Comment on egusphere-2023-1167', Anonymous Referee #1, 22 Sep 2023
This study utilized a Bayesian algorithm to optimize Vcmax25 in the land surface model against eddy covariance observations at eight mature boreal forest stands in North America. The results showed that the Bayesian algorithm can optimize Vcmax25 and improve ET and GPP estimates. The topic is interesting, and the results look promising. However, I am not convinced that the manuscript is innovative enough to contribute to the development of physical models, and therefore, I cannot accept it for publication.
1) Data assimilation methods have been widely employed in the optimization of Vcmax25 (He et al., 2019). Data assimilation methods can improve the estimation of vegetation photosynthesis by assimilating remote sensing SIF data at regional or global scales. Although the results of this manuscript are reliable, I do not believe that site-level optimization can be extrapolated to regional or global scales. The expansion from site-specific to regional scales is crucial for the development of physical models.
2) The authors need to provide more details about the model, including meteorological data and auxiliary data. Additionally, what is the timescale of Vcmax25 optimization? Is it on a daily, monthly, or throughout the entire growing season? Which years' observational data were used for optimizing Vcmax25? And which years' observational data were used for the spin-up? The site name, vegetation type, and other key information should be listed in the manuscript.
3) This study lacks independent validation. Eddy covariance observations of ET and GPP were used to optimize Vcmax25. Subsequently, the optimized Vcmax25 estimates of ET and GPP were further compared with eddy covariance observations. Although some global gridded products have been used to assess model simulation results, these gridded products exhibit uncertainty, and their observational footprints do not align with eddy covariance observations.
4) In this study, the random forest method was employed to characterize the relative importance of various influencing factors on Vcmax25. The limited optimization of Vcmax25 values in this study may lead to overfitting or underfitting issues in the machine learning method. This will impact the credibility of the relative importance results.
He, L., Chen, J. M., Liu, J., Zheng, T., Wang, R., Joiner, J., Chou, S., Chen, B., Liu, Y., and Liu, R.: Diverse photosynthetic capacity of global ecosystems mapped by satellite chlorophyll fluorescence measurements, Remote Sens. Environ., 232, 111344, https://doi.org/10.1016/j.rse.2019.111344, 2019.
Citation: https://doi.org/10.5194/egusphere-2023-1167-RC1 -
RC2: 'Comment on egusphere-2023-1167', Anonymous Referee #2, 06 Oct 2023
Vcmax is an important parameter for TBMs. Simply setting Vcmax25 in TBMS induces the uncertainty of TBMs. In this paper, authors conducted a Bayesian algorithm to optimize Vcmax25 in CLASSIC against eddy covariance observations at eight mature boreal forest stands in North America for six representative PFTs and identified the spatial variability of Vcmax25. This paper try to explore the Vcmax change in boreal forests. However, l am major concerned about optimizing strategy used in this study .
1. How to use TPE to optimize Vcmax25 was not explained in section 2.3. Vcmax was optimized in what time scale, yearly or daily? And in single-site optimization, for example, the observations were only the GPP and ET of the whole site. Does this result in ill-fitting problems when optimizing Vcmax for multiple PFTs simultaneously?
2. The study does not include a sensitivity analysis of the parameters. And through the Farquhar equation, the relationship between GPP and Vcmax is easier to understand. Why choose ET to optimize Vcmax needs more explanation in model structure. To my knowledge, ET is sensitive to the parameters that control stomatal conductance change.
3. To avoid attributing all uncertainties of simulation results to Vcmax25, I suggest optimizing several key parameters of the carbon and water cycles together.
Minor comments
1. In section 2.2, a map of the distribution of the eight mature boreal forest stands is needed.
2.How many site-years were used in optimization needs to be described in Table S2.
3. Generally, in site optimization, the details of simulated results against observations before and after optimization should be shown as a series of figures on a daily time scale. These figures are important for audiences to understand the improvement of model performance.
Citation: https://doi.org/10.5194/egusphere-2023-1167-RC2 -
AC1: 'Author responses to anonymous referees #1 and #2', Bo Qu, 01 Feb 2024
We thank the two anonymous referees for their time and efforts in reviewing our manuscript and providing constructive comments that have helped us improve it. We have carefully
considered the comments and made the revisions to our manuscript. Please see the attached document for our detailed responses and revisions.
Data sets
A boreal forest model benchmarking dataset for North America: a case study with the Canadian Land Surface Scheme including Biogeochemical Cycles (CLASSIC) Bo Qu, Oliver Sonnentag, Alexandre Roy, Joe R. Melton, T. Andrew Black, Brian Amiro, Eugénie S. Euskirchen, Masahito Ueyama, Hideki Kobayashi, Christopher Schulze, Gabriel Hould Gosselin, and Alex J. Cannon https://doi.org/10.5281/zenodo.7266010
Model code and software
PFT-Vcmax25 optimization in CLASSIC (v.1.3) Bo Qu, Roy, Alexandre Roy, Joe R. Melton, Jennifer L. Baltzer, Youngryel Ryu, Matteo Detto, and Oliver Sonnentag https://doi.org/10.5281/zenodo.8136578
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 355 | 121 | 37 | 513 | 52 | 15 | 23 |
- HTML: 355
- PDF: 121
- XML: 37
- Total: 513
- Supplement: 52
- BibTeX: 15
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1