the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Nov 2024
Assimilating Multi-site Eddy-Covariance Data to Calibrate the CH4 Wetland Emission Module in a Terrestrial Ecosystem Model
Abstract. In this study, we use a data assimilation framework based on the Adaptive Markov Chain Monte Carlo (MCMC) algorithm to constrain process parameters in LPJ-GUESS model using CH4 eddy covariance flux observations from 14 different natural boreal, temperate and arctic wetlands. The objective is to derive a single set of calibrated parameter values. The calibrated parameter values are then used in the model to validate its CH4 flux output against independent CH4 flux observations from five different types of natural wetlands situated in different locations, assessing their generality for simulating CH4 fluxes from boreal, temperate and arctic wetlands. The results show that the MCMC framework has substantially reduced the cost function (measuring the misfit between simulated and observed CH4 fluxes) and facilitated detailed characterisation of the posterior parameter distribution. A reduction of around 95 % in the cost function and approximately 50 % in RMSE against the observations were achieved. The results of validation experiment indicate that for four out of the five validation sites the RMSE was successfully reduced, demonstrating the effectiveness of the framework for estimating CH4 emissions from wetlands not included in the assimilation experiment. For wetlands above 45° N, the total mean annual CH4 emission estimation using the optimised model resulted in 28.16 Tg C y-1, and for regions above 60° N, it resulted in 7.46 Tg C y-1.
- Preprint
(3516 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3305', Anonymous Referee #1, 07 Jan 2025
General Comments:
This manuscript presents a follow-up study to a recent paper (Kallingal et al., 2023) which showed that model parameters for the LPJ-GUESS model could be tuned using a Markov Chain Monte Carlo type approach. While Kallingal et al only tuned model parameters based on data from one site, this manuscript uses data from 14 different sites to assess whether the parameter fitting approach works on a broader scale. Study results show that the approach does indeed work, and authors use the resulting model parameters in LPJ-GUESS to estimate total mean annual CH4 emissions above 45N.
The manuscript topic is timely and valuable, and improving global estimates of wetland methane emissions is crucial to reducing uncertainty surrounding this critical greenhouse gas. The authors have clearly put significant time and effort into this work, and I appreciate how exhaustively they have presented their findings. However, in my opinion the manuscript would benefit significantly from some work to reduce it’s length. As is the paper is 29 pages, which feels much too long for the scope of work and results presented. I suggest authors spend some time trimming down the text as I think this would greatly improve the reader’s ability to follow the work. Some of the components could be moved to the appendix, and some may be best to eliminate (see specific comments on section 2.6 and 3.4.1).
Please see below for some more specific comments:
Line 22-23 – Please elaborate on what the difficulties have been, this will better justify the need for your work
Line 26-28. Again, I suggest adding more detail on what model complexities have been added over time to give the reader more context.
Lines 60-62. Is it unclear if you are focusing on just high-latitude wetlands, or are also including temperate wetlands – please clarify.
Line 111 – What is your justification for using this approach? The standard of care within the eddy covariance flux community is more advanced than this, please see Vuichard and Papale et al., 2015 (doi:10.5194/essd-7-157-2015) for details and update your analysis accordingly.
Section 2.2 – I find this section greatly lacking in necessary detail. At no point do authors even clarify what type of flux measurements they are using. Furthermore, authors do not say whether they acquired their data through established data-sharing portals (such as those from AmeriFlux or ICOS), or if data were provided directly to them via site PIs. If data are not acquired from the regional networks, which have established protocols for data QAQC, authors must specify how data were cleaned and processed. Assuming data are from regional networks, authors should provide DOI links to the data sets (in addition to the already-provided links to papers describing those data) as well as the office 5-character site names to help readers understand where their data are coming from. If authors aren’t using the standardized, gap-filled data from the FLUXNET-CH4 product, please explain why.
Section 2.6 – A lot of these presented functions are standard definitions and do not need to be written out, or at least could be saved for an appendix.
Table 3 – It would be helpful to report the # of site-years of data, rather than # of data points available. Furthermore, please specify range of data (year 1 – year x)
Table 5 – Prior values from table 5 do not match Table 1, please fix this. Values in Table 5 also do not match black dashed “Prior mean” lines in Figure 2. Furthermore, Table 5 does not contain any information that is not graphically shown in Figure 2, so should be moved to an appendix.
Figure 3. – most of these correlations are not significant, please only include correlation numbers for significant relationships.
Figure 5 – Please change the coloring so it is readable by people with color blindness.
Figure 6a – Instead of using different marker shapes to delineate different lines, I suggest instead using different line dash patterns as this will be easier to distinguish when points overlap.
Section 3.4.1 I question whether you should include a detailed assessment of how the estimated transport pathways have shifted in importance between prior and post. First of all, flux data are not able to distinguish between CH4 emitted from these various pathways, thus there are no data to check the validity of these results. Second, you acknowledge in line 442 that LPJ-GUESS has a “lack of detailed representation of ebullition”, thus further throwing into question the results presented in section 3.4.1. Without additional work to check the veracity of these estimates, I suggest eliminating this section.
Line 443 – citations for these model deficiencies?
Line 437-439: This seems like a major shortcoming, how do you justify using a methodology that requires data normality with data that may be far from normal?
Citation: https://doi.org/10.5194/egusphere-2024-3305-RC1 -
RC2: 'Comment on egusphere-2024-3305', Anonymous Referee #2, 26 Jan 2025
The manuscript by Kallingal et al. covers an interesting and timely topic appropriate for Biogeosciences. The study focuses on using data assimilation techniques to improve CH4 flux simulations with the process-based model LPJ-GUESS. This approach addresses a much-needed yet less explored area within the CH4 research community, making the study both innovative and highly relevant. The manuscript is generally well-organized, with most parts clearly written. I have a few minor comments as follows:
General comments:
Assumption on normal distribution. The assumption that assimilated fluxes follow a normal distribution needs justification. Is this a robust assumption? How does it affect the model estimation? There are some studies suggesting the flux measurements does not follow a Gaussian distribution.
Performance visualization: Figure 9 does not clearly convey performance improvements post-assimilation. Including a scatterplot to better illustrate this would be helpful.
Observation uncertainty: How is uncertainty in the observations considered in the data assimilation? Here, the daily mean values derived from half-hourly measurements are used, which the temporal coverage of half-hourly measurements can affect the derived daily mean value.
Annual budget calculation: Does the PEATMAP dataset cover all wetlands or only peatlands? If it is restricted to peatlands, the exclusion of other wetland types (e.g., mineral wetlands) makes comparisons with the GCP results problematic. This needs clarification.
Insights by wetland types: Presenting results by sites is useful but limited in scope. Can the authors share insights categorized by wetland types?
Specific comments:
Line 280: The term “contribution” is unclear. What specific aspect or variable does this refer to?
Line 307: The dominance of only two components after optimization is interesting. Are these findings supported by observations? What are the implications of this result?
Line 511: Is “underestimation” the correct word?
Citation: https://doi.org/10.5194/egusphere-2024-3305-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 131 | 43 | 12 | 186 | 6 | 9 |
- HTML: 131
- PDF: 43
- XML: 12
- Total: 186
- BibTeX: 6
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1