the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Aug 2024
Precipitation-fire-functional interactions control biomass stocks and carbon exchanges across the world’s largest savanna
Abstract. Southern African woodlands (SAW) are the world’s largest savanna, covering ~3 M km2, but their carbon balance, and its interactions with climate and disturbance are poorly understood. Here we address three issues that hinder regional efforts to address international climate agreements: producing a state-of-the-art C budget of SAW region; diagnosing C cycle functional variation and interactions with climate and fire across SAW; and evaluating SAW C cycle representation in land surface models (LSMs). Using 1506 independent 0.5° pixel model calibrations, each constrained with local earth observation time series of woody carbon stocks (Cwood) and leaf area, we produce a regional SAW C analysis (2006–2017). The regional net biome production is neutral, 0.0 Mg C ha-1 yr-1 (95 % Confidence Interval –1.7 – 1.6), with fire emissions contributing ~1.0 Mg C ha-1 yr-1 (95 % CI 0.4–2.5). Fire-related mortality driving fluxes from total coarse wood carbon (Cwood) to dead organic matter likely exceeds both fire-related emissions from Cwood to atmosphere and non-fire Cwood mortality. The emergent spatial variation in biogenic fluxes and C pools is strongly correlated with mean annual precipitation and burned area. But there are multiple, potentially confounding, causal pathways through which variation in environmental drivers impacts spatial distribution of C stocks and fluxes, mediated by spatial variations in functional parameters like allocation, wood lifespan and fire resilience. Greater Cwood in wetter areas is caused by positive precipitation effects on net primary production and on parameters for wood lifespan, but is damped by a negative effect with rising precipitation increasing fire-related mortality. Compared to this analysis, LSMs showed marked differences in spatial distributions and magnitudes of C stocks and fire emissions. The current generation of LSMs represent savanna as a single plant functional type, missing important spatial functional variations identified here. Patterns of biomass and C cycling across the region are the outcome of climate controls on production, and vegetation-fire interactions which determine residence times, linked to spatial variations in key ecosystem functional characteristics.
- Preprint
(2771 KB) - Metadata XML
-
Supplement
(11522 KB) - BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2024-2497', Anonymous Referee #1, 28 Oct 2024
reply
The carbon cycle in the Southern African woodlands (SAW) is significantly influenced by climate change and fire disturbances. The lack of Eddy flux observations and spatial variation in ecosystem functional traits, such as lifespan and fire resilience, complicate regional carbon budget assessments. This study innovatively employs data assimilation to generate individual simulations for 1,506 pixels in SAW, addressing the spatial variability associated with using a single plant functional type (PFT) within land surface models (LSMs). It develops a comprehensive regional carbon budget and investigates the causal relationships between carbon flux, climate factors, and fire disturbances, highlighting the importance of spatial variation in ecosystem functional characteristics.
Overall, this manuscript is in good shape for publication, and the methodology is innovative with rich contents. However, several specific issues need to be addressed, including citation errors and disconnections between the research findings and the narrative, which may hinder clarity. Please find them below:
Please provide more details about the method and materials, such as an accessible code of the model and data. Also at Page 13, Lines 13-16: There are no materials to support these results.
Page 13, Lines 17-18: Figure S2 does not include Carbon Tracker Europe.
Page 15, Lines 15-17: I cannot calculate the three numbers (19%, 81%, and 59%) based on the text and Figure 3. Could you clarify the calculations?
Page15, Lines 19-20: T The numbers do not match those in Figure 3, which may cause confusion. For instance, the mean value of NEE calculated as Ra + Rh - GPP = 7.67 + 7.04 - 15.95 = -1.24 does not correspond with -1.04 in Figure 3. The authors should verify all results carefully. Additionally, the numerical representation in Figure 3 retains two decimal places, while other sections, like the abstract, use one decimal place. This inconsistency can be confusing; I recommend a unified format.
Page16. Lines 9-10: There appears to be a reference error; it may be Figure S4 instead of Figure 4. Additionally, how are the uncertainties measured from Figure S4? I cannot conclude that uncertainties on ΔCDOM being four times higher than for ΔCveg.
Figure 4: Additional Y-axis titles in the second row need to be removed.
Page 20, Lines 5-6: The text states that Twood,fire and Twood,other are spatially variable, but there is no spatial pattern for Twood,fire in Figure 8.
Page 20, Line 2: A right bracket is missing.
Page 20, Lines 6-8: Is this referring to Figure S7? However, Figure S7 appears to relate to Cfoliage rather than Cwood.
Citation: https://doi.org/10.5194/egusphere-2024-2497-RC1 -
RC2: 'Comment on egusphere-2024-2497', Anonymous Referee #2, 18 Nov 2024
reply
Williams et al. describe how an established, intermediate-complexity biophysical model has been constrained using biomass and leaf area observations, along with a soil carbon product, to estimate a carbon (C) budget for the extensive southern African savanna. This parameterization was then employed to investigate potential spatial variations in ecosystem functional relationships and causal links between C stocks, fluxes, climate, and fire. The results were contrasted with the typical representation of the region in a suite of land surface models that use more prognostic approaches to simulating fire in the C cycle.
The study is ambitious in scope, and its results and discussion aim to address three key research questions. However, these findings should be contextualized by examining the effectiveness of the model-data fusion approach and the efficiency of the data-constrained outputs. Clearly linking the emergent spatial patterns of C stocks and fluxes to the underlying parameterizations poses challenges, particularly due to the brevity of the model description in Section 3.2.1 and the optimization processes in Section 3.2.2, and how the calibration and validation results are discussed in Section 4.1. These issues are especially pertinent to how fire is integrated into the model.
Fire is identified as a crucial control on ecosystem dynamics, and interpreting the overall results heavily depends on how fire's impact is represented. Given this importance, additional details are necessary to fully explain how fire and associated emissions are modeled within the CARDAMOM framework.
The description of fire emission calculations requires further elaboration and should clarify the following points without assuming in-depth familiarity with Exbrayat et al. (2018):
- How is "burnt area" used as a driver? Is there a specific step that converts burnt area into biomass or "vegetation pools in the burned area"?
- On Page 10, Line 7, it is stated that emissions are calculated by multiplying burnt area by a combustion fraction parameter from Exbrayat et al. (2018). How does this differ from the "specific combustion parameters" applied to each C pool?
- The same resilience factor is applied to all C pools (except SOM). How is this assumption justified?
- Without a diagram, it is difficult to understand the flow of C from pool to emission and between pools, as described on Page 10, Lines 9–12 (although Figure 3 in the results section is somewhat helpful).
Section 3.2.2 would benefit from an expanded explanation of how the fire and combustion parameters are constrained. Specific clarifications needed include:
- Is the only observational data available to constrain these parameters a rapid change in aboveground biomass/LAI, coinciding with a time step where significant burnt area is observed in the forcing?
- Within the EDC framework, is there a mechanism to explicitly link biomass changes to fire occurrence, thereby impacting fire and combustion parameters? Alternatively, does the optimization infer fire, disturbance, and turnover parameters solely from biomass changes?
- What happens if/when observations of rapid biomass changes and burnt area forcing do not align in space and time?
- The simulated change in aboveground biomass appears to depend on both combustion completeness for wood and biomass resilience. What are the implications for estimating these parameters, which exhibit significant equifinality when constrained only by biomass change observations? How does this affect the fire-related Mort_wood flux and the subsequent C_som dynamics, leading to the large E_som flux?
Section 4.1 provides limited information about how parameters, including fire and combustion parameters, are constrained or how model performance is validated. To address this, the following questions should be clarified:
- Why and how was the validation data in Figure S2 combined, and why is this not discussed in Section 3.2.3?
- On Page 13, Lines 20–21, it is stated that "fire emissions fell at the lower end of the range of fire emissions products," yet this appears to contradict the lower panel of Figure S2.
- Page 13, Lines 21–22, mentions that "uncertainties were much larger than the products' range." Why is this the case, and what are the implications for interpreting emergent functional and causal relationships?
Given the challenges in constraining fire and combustion parameters and the resulting large uncertainties in fire emissions, which hinder interpretation of ecosystem-level model outputs, it would be informative to compare CARDAMOM emissions with emissions from the prognostic fire models in the TRENDY LSM dataset. On Page 25, Line 30, it is suggested that "inconsistencies in C emissions from respiration and fire" account for mismatches in NBP, and this claim should be substantiated with more information on TRENDY model emissions in Section 4.5.
Citation: https://doi.org/10.5194/egusphere-2024-2497-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 336 | 80 | 16 | 432 | 36 | 7 | 13 |
- HTML: 336
- PDF: 80
- XML: 16
- Total: 432
- Supplement: 36
- BibTeX: 7
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1