the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Global Fire Emissions Database burned-area dataset into Community Land Model version 5.0 – Biogeochemistry: Impacts on carbon and water fluxes at high latitudes
Hocheol Seo
Abstract. Wildfires influence not only ecosystems but also carbon and water fluxes on Earth. Yet, the fire processes are still limitedly represented in land surface models (LSMs), thus simulating the occurrence and consequences of fires. Especially, the performance of LSMs in estimating burned areas across high northern latitudes is poor. In this study, we employed the daily burned areas from the satellite-based global fire emission database (version 4) (GFED4) into the community land model (version 5.0) with a biogeochemistry module (CLM5-BGC) to identify the effects of accurate fire simulation on carbon and water fluxes over Alaska and Eastern Siberia. The results showed that the simulated carbon emissions with the burned areas from GFED4 (i.e., experimental run) were significantly improved in comparison to the open-loop run (i.e., default run), which resulted in opposite trends of the net ecosystem exchange for 2004, 2005, and 2009 over Alaska between the open-loop and experimental runs. Also, we identified carbon emissions were more sensitive to the wildfires in Alaska than in Eastern Siberia, which could be explained by the vegetation distribution (i.e., tree cover ratio). In terms of water fluxes, canopy transpiration in Eastern Siberia was relatively insensitive to the size of burned area due to the interaction between leaf size and soil moisture. This study uses CLM5-BGC to improve our understanding of the role of burned areas in eco-hydrological processes at high latitudes. Furthermore, we suggest that the improved approach will be required for better predicting future carbon fluxes and climate change.
Hocheol Seo and Yeonjoo Kim
Status: final response (author comments only)
-
CC1: 'Comment on gmd-2022-294', Sarah Gallup, 09 Jan 2023
This study focuses on a decidedly useful topic. The design is a reasonable approach to learning about the limitations and potential to improve CLM’s fire simulations. Phrasing and copyediting are substandard. The analysis of the results needs more meat in terms of using details of the two runs to better understand how and why they differ. The model runs would support a substantially tighter and more coherent assessment. What can the authors show or even speculate about why CLM fire matches not only rather poorly to the datasets, but also differently in the two continents? Ssaying CLM Fire is “limited”, pointing out that it is imperfect, is less useful than helping the community think about reasons and specifics. Several of the speculations about real-world reasons the CLM fire algorithm is imperfect are insightful and useful.
Some notes about uncertainty in the benchmark datasets seem warranted. As an obvious example, GFED emissions too are a model. While it is reasonable to assume the comparison data is more accurate than an ESM simulation of fire, what considerations about the inevitably imperfect inventories should a reader keep in mind? How similar are the two inventories’ derivation algorithms and data sources? Making the comments specifically relevant to the patterns the study finds would be most helpful. As only an example, what is the correlation of GFED and AKFED emissions for the study area and period?
Any information about the relevance of peat fire would make this paper a substantially stronger tool for improving fire in CLM. As examples, what is the relative abundance of peat in the study area compared to the rest of Siberia? What portion if any of the “open loop” Siberian burned area and emissions were generated from the peat fire algorithm within CLM fire?
Thank you for tackling this study.
general - Please either use a consistent number of significant digits, or justify why not.
line 41 - 'Human-caused' is conflated with human-ignited. Warmer climate, too, is human-caused.
102 - Equations 1 & 2 should be cited, including with equation numbers from Lawrence19
106 - By “leaf size” do you mean LAI? The terms are not interchangeable.
153 - What data source do you use for woodfuel burning estimates?
166 - Pls explain why you chose the specific area within Siberia, and what relevant ways it is similar to or different from the rest of Siberia.
201 - an egregious example of the need for copyediting. Ditto l. 230-234.
217 - While the general point is very well taken, “no human impacts” is an overstatement. See p.29 of https://fire.ak.blm.gov/content/aicc/Alaska%20Statewide%20Master%20Agreement/3.%20Alaska%20Interagency%20Wildland%20Fire%20Managment%20Plan%20(AIWFMP)/2022%20AIWFMP%20Final%20Signed%202022-02-28.pdf. Responding agencies will “conduct site protection as warranted.”
237 - OK, but there now exists information about differences between GFED3 and GFED4. To what extent is Veraverbeke’s explanation that you reiterate perhaps now addressed - or not?
239 - Rather than speculate, pls look up the numbers and compare them at least to each other and ideally also to additional references.
256 - rates of change, or changes?
279 - 281 needs replacing. Line 279 is an overstatement; Line 280 was known before the study started simply because all models are imperfect; Line 281 is not a logical conclusion based on the prior two statements. Writing a stronger analysis as requested in the general notes above will provide better material to summarize in this paragraph.
Citation: https://doi.org/10.5194/gmd-2022-294-CC1 - AC1: 'Response to RC1, RC2, and CC1', Yeonjoo Kim, 02 Apr 2023
-
RC1: 'Comment on gmd-2022-294', Anonymous Referee #1, 06 Feb 2023
The authors were trying to evaluate the impacts of using prognostic and diagnostic wildfire schemes on ecosystem carbon and hydrological cycles simulated by CLM5-BGC. They found the default CLM5-BGC overestimated/underestimated the burned area in Eastern Siberia/Alaska, causing the overestimating/underestimation of wildfire carbon emissions. In contrast, the CLM5-BGC prescribed with observational burned area showed evident improvement in simulating wildfire carbon emissions for both regions. They further compared the two simulations with different wildfire schemes, in terms of major carbon and water fluxes, and showed larger influences on NEE than those for other variables. The modeling idea was unique, in terms of prescribing remote-sensing burned area within land surface model, especially across the high latitudes regions. However, the model evaluations and intercomparisons still need to be improved. For example, the authors may compare these two model results for more variables (e.g., LAI, GPP, ET, Soil Moisture) using different sources of observations or observation-based products. The authors could also add more thoughts/analyses on thow to quantify and reduce wildfire related biases for CLM5-BGC, in terms of drivers, processes and parameters. Additionally, sentences between lines 231 and 234 read confusing; and color scheme in Fig. 3 needs to be reversed?
Citation: https://doi.org/10.5194/gmd-2022-294-RC1 - AC1: 'Response to RC1, RC2, and CC1', Yeonjoo Kim, 02 Apr 2023
-
RC2: 'Comment on gmd-2022-294', Anonymous Referee #2, 14 Feb 2023
Hocheol Seo and Yeonjoo Kim integrated GFED burned area dataset into CLM5-BGC model and investigated how fire activity affect ecosystem carbon and water fluxes over Alaska and Easter Siberia. They found that using GFED observed burned area, CLM5-BGC performed better in capturing fire emissions. Moreover, the carbon emissions over Alaska was sensitive to wildfire, while transpiration over Easter Siberia was insensitive to burn. The paper is well constructed. Below are major concerns:
- The design of the model experiment needs to be improved to fully account the impact of historical fires and for a fair comparison between OL vs EXP-GFED4 runs.
This study used BGC version of CLM5, that have carbon, nutrient, water cycles. The 200 year spin up might be enough to stabilize soil temperature and moisture, but is too short to stabilize soil carbon pool and ensure a quasi-steady state condition. Suggestion: plot out total ecosystem carbon for the last 10 or 20 years of spinup period, the changes of total ecosystem carbon should be trivial. Also, long-term average of net ecosystem exchange (NEE) should be near zero in the end of spinup. If not, tried a longer spinup period.
After the spinup, a long-term transient simulation (starting from year 1850 or 1901) is necessary to ensure the land use, warming, and CO2 enrichment signals are all appropriately picked up by the CLM5-BGC model. For example, in this period, we often see initial decline of vegetation carbon due to land use, and then enhanced vegetation growth in response to warming and higher CO2 concentration will overwhelm. Such historical changes of vegetation activities, soil moisture conditions will affect fuel availability and combustibility for simulations from year 2001-2012 (period of focus for analysis). Otherwise, without appropriate spinup and transient simulation, the comparison of fire emissions and transpiration fluxes and others might not be convincing.
- CLM5-BGC baseline model performance and improvement
In order to understand how much of improvement was due to the integration of GFED burned area dynamics, it will be necessary to first showed CLM5-BGC baseline model performance over the area of interest. It will be good to compare OL and EXP-GFED4 simulations against observations (Figure 6-10). Suggested datasets are e.g., FLUXCOM GPP/NEP, MODIS LAI, GLEAM ET, GEOCARBON vegetation biomass.
- Area of focus
The box area in Figure 1 seems arbitrary. It will be better to use geographic boundary (e.g., state of Alaska) instead of a random box.
- Plant functional type differences
It is not clear, how each different plant functional type (PFT) handled when GFED data is integrated to CLM5-BGC at gridcell level. At each gridcell, CLM has multiple PFT that have different level of fuel conditions (e.g., arctic grass vs boreal tree). When the observed burned area gets integrated, how to reasonably assign the burn to each PFT? For example, the observed burn may occur over forest, while in CLM the observed burn is imposed on the whole gridcell that have both trees and grasses.
- Temporal scale issue
GFED is a monthly product, while CLM5-BGC runs at a much finer temporal resolution (e.g., 30min). How to assigned monthly burned area to each individual time step in CLM5-BGC? It will make a big difference to apply the burned area to the beginning versus to the end of each month?
Citation: https://doi.org/10.5194/gmd-2022-294-RC2 - AC1: 'Response to RC1, RC2, and CC1', Yeonjoo Kim, 02 Apr 2023
- AC1: 'Response to RC1, RC2, and CC1', Yeonjoo Kim, 02 Apr 2023
Hocheol Seo and Yeonjoo Kim
Hocheol Seo and Yeonjoo Kim
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 531 | 137 | 18 | 686 | 4 | 5 |
- HTML: 531
- PDF: 137
- XML: 18
- Total: 686
- BibTeX: 4
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1