the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
08 Jan 2021
08 Jan 2021
Modelled land use and land cover change emissions – A spatio-temporal comparison of different approaches
- 1Department of Geography, Ludwig Maximilians Universität, Luisenstrasse 37, 80333 Munich, Germany
- 2Max Planck Institute for Meteorology, 20146 Hamburg, Germany
- 3Max Planck Institute for Biogeochemistry, 07745 Jena, Germany
- 4Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research / Atmospheric Environmental Research, 82467 Garmisch-Partenkirchen, Germany
- 5Laboratoire des Sciences du Climat et de l’Environnement (LSCE), 91191 Gif-sur-Yvette, France
- 6Climate and Environmental Physics, Physics Institute and Oeschger Centre for Climate Change Research, University of Bern, Bern 3012, Switzerland
- 7National Center for Atmospheric Research (NCAR), Climate & Global Dynamics Lab, Boulder, USA
- 8Department of Ecological Science, Vrije Universiteit Amsterdam, 1081HV Amsterdam, the Netherlands
- 9Department of Meteorology, University of Reading, Earley Gate, Reading RG6 6BB, UK
- 10Climate Processes Section, Climate Research Division, Environment and Climate Change Canada, Victoria, BC, Canada
- 11NASA Goddard Space Flight Center, Biospheric Sciences Laboratory, Greenbelt, Maryland 20771, USA
- 12College of Life and Environmental Sciences, University of Exeter, Exeter EX4 4RJ, UK
- 13School of Forestry and Wildlife Sciences, Auburn University, 602 Ducan Drive, Auburn, AL 36849, USA
- 14Climate Change Science Institute & Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
- 15Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB, UK
- anow at: German Aerospace Center, Institute of Athmospheric Physics, 82234 Oberpfaffenhofen, Germany
- 1Department of Geography, Ludwig Maximilians Universität, Luisenstrasse 37, 80333 Munich, Germany
- 2Max Planck Institute for Meteorology, 20146 Hamburg, Germany
- 3Max Planck Institute for Biogeochemistry, 07745 Jena, Germany
- 4Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research / Atmospheric Environmental Research, 82467 Garmisch-Partenkirchen, Germany
- 5Laboratoire des Sciences du Climat et de l’Environnement (LSCE), 91191 Gif-sur-Yvette, France
- 6Climate and Environmental Physics, Physics Institute and Oeschger Centre for Climate Change Research, University of Bern, Bern 3012, Switzerland
- 7National Center for Atmospheric Research (NCAR), Climate & Global Dynamics Lab, Boulder, USA
- 8Department of Ecological Science, Vrije Universiteit Amsterdam, 1081HV Amsterdam, the Netherlands
- 9Department of Meteorology, University of Reading, Earley Gate, Reading RG6 6BB, UK
- 10Climate Processes Section, Climate Research Division, Environment and Climate Change Canada, Victoria, BC, Canada
- 11NASA Goddard Space Flight Center, Biospheric Sciences Laboratory, Greenbelt, Maryland 20771, USA
- 12College of Life and Environmental Sciences, University of Exeter, Exeter EX4 4RJ, UK
- 13School of Forestry and Wildlife Sciences, Auburn University, 602 Ducan Drive, Auburn, AL 36849, USA
- 14Climate Change Science Institute & Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
- 15Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3PB, UK
- anow at: German Aerospace Center, Institute of Athmospheric Physics, 82234 Oberpfaffenhofen, Germany
Abstract. Quantifying the net carbon flux from land use and land cover changes (fLULCC) is critical for understanding the global carbon cycle, and hence, to support climate change mitigation. However, large-scale fLULCC is not directly measurable, but has to be inferred from models instead, such as semi-empirical bookkeeping models, and process-based dynamic global vegetation models (DGVMs). By definition, fLULCC estimates are not directly comparable between these two different model types. As an example, DGVM-based fLULCC in the annual global carbon budgets is estimated under transient environmental forcing and includes the so-called Loss of Additional Sink Capacity (LASC). The LASC accounts for the impact of environmental changes on land carbon storage potential of managed land compared to potential vegetation which is not represented in bookkeeping models. In addition, fLULCC from transient DGVM simulations differs depending on the arbitrary chosen simulation time period and the historical timing of land use and land cover changes (including different accumulation periods for legacy effects). An approximation of fLULCC by DGVMs that is independent of the timing of land use and land cover changes and their legacy effects requires simulations assuming constant pre-industrial or present-day environmental forcings. Here, we analyze three DGVM-derived fLULCC estimations for twelve models within 18 regions and quantify their differences as well as climate- and CO2-induced components. The three estimations stem from the commonly performed simulation with transiently changing environmental conditions and two simulations that keep environmental conditions fixed, at pre-industrial and present-day conditions. Averaged across the models, we find a global fLULCC (under transient conditions) of 2.0 ± 0.6 PgC yr-1 for 2009–2018, of which ∼40 % are attributable to the LASC (0.8 ± 0.3 PgC yr-1). From 1850 onward, fLULCC accumulated to 189 ± 56 PgC with 40 ± 15 PgC from the LASC. Regional hotspots of high cumulative and annual LASC values are found in the USA, China, Brazil, Equatorial Africa and Southeast Asia, mainly due to deforestation for cropland. Distinct negative LASC estimates, in Europe (early reforestation) and from 2000 onward in the Ukraine (recultivation of post-Soviet abandoned agricultural land), indicate that fLULCC estimates in these regions are lower in transient DGVM- compared to bookkeeping-approaches. By unraveling spatio-temporal variability in three alternative DGVM-derived fLULCC estimates, our results call for a harmonized attribution of model-derived fLULCC. We propose an approach that bridges bookkeeping and DGVM approaches for fLULCC estimation by adopting a mean DGVM-ensemble LASC for a defined reference period.
Wolfgang A. Obermeier et al.
Status: closed
-
RC1: 'Comment on esd-2020-93', Anonymous Referee #1, 18 Jan 2021
In this study, the authors analyze three DGVM-derived fLULCC estimations for twelve models within 18 regions and quantify their differences as well as climate- and CO2-induced components. Results showed a global fLULCC of 2.0 ± 0.6 PgC per year for 2009–2018, of which ∼40% are attributable to the LASC. Regional hotspots of high cumulative and annual LASC values are found in the USA, China, Brazil, Equatorial Africa and Southeast Asia, mainly due to deforestation for cropland. Distinct negative LASC estimates, in Europe (early reforestation) and from 2000 onward in the Ukraine (recultivation of post-Soviet abandoned agricultural land), indicate that fLULCC estimates in these regions are lower in transient DGVM- compared to bookkeeping-approaches. By unraveling spatio-temporal variability in three alternative DGVM-derived fLULCC estimates, our results call for a harmonized attribution of model-derived fLULCC. This study proposes an approach that bridges bookkeeping and DGVM approaches for fLULCC estimation by adopting a mean DGVM-ensemble LASC for a defined reference period. I would recommend this work for publication with few minor modifications.
Specific comment:
Line 130: More introduction about” gridded output” is needed. For example, the resolution of these data. Monthly data or Annual data?Line140: All abbreviations must be explained. For example, HYDE and FAO.
-Line 154: ’the amount of precipitation in the Poyang Lake Basin’ was not consistent with the caption.
-Line 166: Descriptions of three alternative fLULCC are not clear in the current version.Line 385: I have some serious concern about the assumption that the last 100 years due to climate change –clarify it?
Eq 1,2,3: I really had difficulty in understanding these equations. I suggest the authors made them easy to follow in the revised manuscript.
-Figure 1 box4 presents fLULCC differences, but no information about different line.
-
AC1: 'Reply on RC1', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Wolfgang Obermeier, 20 Mar 2021
-
RC2: 'Comment on esd-2020-93', Wei Li, 28 Jan 2021
The difference in Eluc estimates between DGVMs and bookkeeping models has been reported in many previous studies but never been quantified by decomposing into more specific fluxes like LASC. The authors filled this research gap using fractional simulations by various DGVMs. The analysis is very detailed and comprehensive with precise definitions of different LULCC fluxes and components and corresponding quantifications. It is a significant step in disentangling the Eluc components on top of Pongratz et al. (2014) and Gasser & Ciais (2013) and has important implications on the definition of Eluc in the global carbon budget and implementing climate mitigation measures. The manuscript is well written with clear description and detailed supplementary materials. I see no major flaw in this manuscript and thus suggest this work for publication with few minor revisions.
I have some concerns that may need some clarification and discussion.
Although the LASC explains the Eluc difference between DGVMs and bookkeeping models, it would be better to address whether LASC exists in the real fluxes of carbon emission and sink, i.e. whether can be observed. If I understood correctly, positive LASC represents the loss of potential carbon sink and thus didn’t physically exist, i.e. this part of CO2 wasn’t released into the atmosphere. For the negative LASC like in reforestation, this part should be physically stored in the biomass or soil C pools, right?
As the authors stated, the timing of LULCC matters. Therefore, from my understanding, estimation of the accurate LASC for a specific LULCC event (a deforestation event in 1950 for an example) needs simulations similar to the S0 and S4 but using the climate and CO2 status when the LULCC event occurs (i.e. 1950) instead of the pre-industrial climate and CO2 status. Although the authors came to this point somewhere in the text, it would be better to emphasize this point explicitly.
The attribution to climate and CO2 in Sect. 2.2.2 is rather uncertain. Is there any observation data (e.g. FACE + warming experiment field data) that can be used to validate this attribution method?
Fig. 5-7 show results from each DGVM, but not reported in the text. Could add one or two sentences to say which models always give high e.g. LASC in Fig.5 or all models are similar?
L29-30: “high-latitudes” usually refers to boreal region. I think much early agricultural expansion occurred in the temperate regions. May rephrase.
L248: bookkeping -> bookkeeping
L253, L256: What are the numbers in the brackets?
L274: What are the pulses of the purple line in Fig. 2c?
L312-313: Why reforestation increased fLULCC_pd?
-
AC2: 'Reply on RC2', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Wolfgang Obermeier, 20 Mar 2021
-
RC3: 'Comment on esd-2020-93', Anonymous Referee #3, 08 Feb 2021
Review of esd-2020-93:
Modelled land use and land cover change emissions - A spatio-temporal comparison of different approaches
Summary
The authors present a detailed analysis of multi-dynamic-ecosystem-model estimates of LULCC emissions under different initial and environmental conditions, with some comparison with recent bookkeeping approaches. A main focus is the contribution of the “Loss of Additional Sink Capacity” (LASC) to emissions, which can be best estimated by dynamic ecosystem models with transient environmental conditions. The authors conclude that LASC is considerable, and has some regional variation. Furthermore, they show that CO2 effects dominate the contributions to LASC over climate effects. They also raise the question of whether LASC is appropriate for a harmonized attribution of LULCC emissions.
Overall review
This is an important contribution to LULCC and carbon/climate science and presents interesting results from the trendy model ensemble. There are a few main issues that need to be addressed, however, prior to publication. These are summarized here, with further detail following:
1) There seems to be a better way to calculate the CO2 vs climate estimates. You have simulations that you can difference directly, rather then using the ratio method that has limitations.
2) The results and discussion of regional LASC are confusing and not to the point. These can be better constrained and made more clear with main points in mind.
3) The suggested “harmonized attribution” of LULCC emissions is not consistent with the paper findings and doesn’t differentiate between science and policy. The real issue here is full scientific estimates of emissions, as the authors clearly show require the inclusion of LASC, vs attribution of LULCC emissions for policy purposes, which is a political issue. This paper isn’t about harmonizing different model approaches, but rather clarifying and understanding their differences. This provides scientific bases for accurate estimation of emissions, and provides info for political decisions regarding accounting for attribution policy, but doesn’t show an answer to the “correct” accounting for policy or whether/how the different approaches should be “harmonized.”
Specific comments and suggestions
Abstract
The abstract focuses more on the argument for LASC and DGVMs, than on the results of the study. The argument is better suited for the discussion, with a brief statement in the abstract stating that you use DGVMs to estimate LASC. Then you can add more findings/explanation to the abstract, such as temporal transitions in PTD that support the DGVM vs bookkeeping difference statements.
line 8:
“arbitrary chosen” is unnecessary.
lines 10-11:
Constant environmental forcing does not generate a condition independent of timing and legacy of lulcc. It generates a condition independent of changing environmental conditions, which is how this should be described.
Ensuring that the lulcc magnitude and trajectory are identical across simulations, along with constant environmental forcing, creates a condition almost independent of the timing and legacy of lulcc, as long as the transient dependence isn’t the question. Full independence in this case requires that the biogeochemical dynamics are also identical across simulations, such as may be possible when a single model is used. But this is not possible across different models because each model will respond differently to the same exact lulcc (which is also very difficult to achieve across models), even if the environmental forcing is identical.
I think you mean this in the context of a single-model experiment, with the exact same lulcc and environmental forcing, and not focused on how the LULCC trajectory determines emissions (i.e., path dependence), but this needs to be qualified here as it sounds more general.
lines 21-23:
I am not sure that your analysis calls for this, nor that your approach bridges the two. The issue is a full scientific accounting of emissions vs a partial accounting for policy purposes. While you have evidence for a full scientific accounting, you still have little basis for a partial-LASC approach.
Introduction
lines 28-30:
awkward sentence. delete “e.g.” and use three clauses: deforestation…high lats, tropical deforestation, and recent forest expansion in high lats.
line 31:
“…contributed approximately one-third of global anthropogenic…”
line 35:
delete “in line.” then: flulcc may also gain…
line 42:
“…but also change in…”
line 50:
delete “allow to”
lines 69-72:
The language here starts to confuse the definition and example of LASC. It sounds like just having the potential vegetation or the foregone sink is part of LASC, in addition to the difference of environmental effects on the managed vs unmanaged vegetation. These two sentences should be reworded to clarify that “assuming potential vegetation” and “capture the foregone sinks a given LULCC event destroys” do not obfuscate the definition of LASC.
lines 71-72:
this statement is not correct. there is no guarantee that the change in environmental conditions is increasing LASC at the time of change. so “same direction” does not make sense. I think you are still referring to the example here, but this is a more general concluding statement.
it is more correct to state that these emissions “can” accumulate in the absence of further LULCC, depending on the environmental conditions. which you then give examples of in the next paragraph.
lines 98-99:
I don’t agree that constant conditions create a condition independent of lulcc timing. regardless of the forcing, the time since disturbance makes a difference, especially in the context of DGVMs where ecosystems are not static. A subsequent LULCC has a different emission depending on the period between it and the previous LULCC. And an event early in time on ‘pristine’ land may affect a different biomass then one later in time, even if the conditions are identical. Even with a spun-up model, the biomass may not remain constant during a simulation, and depending on the resolution and application of LULCC, a later LULCC may or may not be applied consistently to ‘pristine’ or ‘managed’ area (and corresponding biomass), or some combination of the two.
I think this description should be changed to “independent of environmental trends” or something like this.
line 108:
delete “in line”
lines 109-110:
unclear: “and consequently into the natural land c sink.” how is this related to ‘included or excluded?’
Data and methods
lines171-172 and 178-179 and 194-195:
These really seem like a single equation each.
line 188:
again, this isn’t true. timing and legacy of LULCC are inherently critical and determining elements in any transient LULCC simulation. these simulations are indifferent to environmental trends or changing environmental conditions.
line 219:
“…C stock changes…”
lines 220-240:
Why did you do it this way, particularly for climate effects? You can get co2+ndep-related emissions directly by subtracting S0 and S1, which would be interesting to compare with your estimate of co2 only emissions using this ratio method.
And you can get the climate-related emissions directly by subtracting S1 and S2, including interactions with CO2, without using this indirect ratio method. and your assumption regarding zero interactions between climate and co2 is not sound because S2 includes both at the same time, and S1 includes only co2. I suggest you use the actual nbp difference of these 2 sims to estimate the climate effects.
Also, you don’t report on the co2+no2 effects at all, so why is this simulation listed in the methods?
Results and discussion
lines 246-247:
unclear: “…to consistently use the same models for the flux and bias estimates on a spatio-temporal level…”
line 254:
did you use the trendyv9 version of sdgvm throughout? if so, you need to state this in the methods.
lines 256-257:
not necessarily. the single year pre-ind estimate is closer to bk estimates, and is equally close to gasser as the present-day for the decade.
line 259:
you haven’t shown the cumulative yet
line 260:
awkward. try “…in the 1950’s and again at the end of the simulation…”
line 272:
“…comes into play…”
lines 272-273:
awkward first half of sentence.
line 277:
during this period (first half of 20th century)
line 279:
“these peaks”
lines 281-284:
maybe because the LULCC occurrence is the same between simulations. but post-peak carbon stocks might have been lowered enough differentially between the sims to reduce the post-peak difference because carbon reductions are generally implemented as fractions, in which case the same LULCC would reduce more carbon in the pd sim than in the pi sim because of higher carbon stocks in the pd sim. and the eed peak is actually after the other peaks and stretched out. so is this really independent of the lulcc timing, as there is a peak in lulcc area that coincides with this emission peak.
line 293:
post 1950
lines 296-320:
This section is difficult to follow, and it isn’t entirely clear what the main point is, other than that LULCC dominates the EED pattern. But more importantly, LULCC seems to drive the fLULCC pattern also, since the EED and the fLULCC patterns also match. temporally, which is does not need a regional breakdown to be shown. And it would appear that the difference in biomass between Pi and Pd is a main driver of the corresponding difference in fLULCC. This boils down to biomass+LULCC controlling fLULCC estimates, with some regional variation in how this relationship contributes to the global estimates.
If the EED pattern is the main point, then take out the other regional plots (5 and 6), as figs 3 and 4 and 7 and 8 show this. The regional plots are difficult to interpret, and so the discussion is helpful, but it should be more clear how the regional patterns contribute to the global patterns, as figure 8 can show (and 7 can show magnitudes better is the scales are matched).
Plots 9-11 are relevant, but they are not used to explain why the EED pattern dominates fLULCC, but they can support the basic relationship between biomass, LULCC trajectory, and fLULCC.
There are too many plots that are too difficult to read, and their referencing is difficult to follow also.
lines 302-303:
The regional comparison plots (figures 5 - 7 and the supplemental) are difficult to compare because they are all on different scales. while it may make it more difficult to see some of the individual lines, putting them on the same scale is the best way to show the differences, which I think is the point.
You can split these plots into multiple figures also because they are so small, which also makes them difficult to read. If that is too many figures, then maybe all the regional plots should be supplemental.
lines 321-352:
This section is difficult to follow, and it isn’t clear what the point is of the comparison between LASC and EED. LASC is a component of fLULCC estimation, and EED is a difference between two estimates of fLULCC that don’t include LASC. In general, the dynamics discussed here are still driven by biomass+LULLCC, since the forests are the high biomass areas and where LASC will have the most pronounced influence. While the regional breakdown and LASC discussion is somewhat informative, as the LASC dynamics are interesting and a key part of this study and since there is some regional variation, the relationships between LASC and EED seem to still be similar across regions, in that LASC tends to grow over time due to LULCC timing, while EED is driven mainly by biomass differences and the fixed LULCC trajectory. So is the point here to just discuss LASC dynamics, or to support the hypothesis that LASC is critical to fLULCC estimates because static biomass conditions do not include temporal effects of changing conditions (and drive errors associated with these biomass conditions)? I don’t think you need the EED stuff here, as your larger point about estimation methods is discussed elsewhere (section 4?).
lines 353-377:
This discussion is a bit more clear as it focuses on LASC dynamics and doesn’t mix in references to other calculations. I suggest combining this section and the previous section and focusing just on LASC dynamics. And the last sentence seems less a concluding sentence and more an evidence sentence to support the discussion.
line 383-384:
What about LULCC effects due to abandonment of agricultural land? This also contributes to higher biomass in some places, and isn’t trivial. You show some evidence of this in figure 12, where it can outweigh the climatic effects in even the long term, but your net results over the whole period do not show how regrowth has contributed more recently to increased biomass in places where this doesn’t outweigh previous emissions.
lines 416-445:
This section seems to be referring to a political question for emission attribution to regions for policy, rather than a scientific question of what are LULCC emissions.
So you should separate these two questions/recommendations. Is your scientific recommendation to discard some LULCC emissions? It doesn’t seem to be so as you clearly show that LASC is an important component in “Accurate quantification of the net carbon flux from land use and land cover changes…” (line 447). Whether this makes sense for regional carbon policy is a different question.
From a policy standpoint, the question here is then why include only part of the total emissions from LULCC (i.e. LASC from second half of 21st century) while including some estimate of the directly effects of changes all the way back to 1850? If it isn’t ok to include LASC back to 1850, why should any effects of pre-1950 LULCC be included? Why not just start full attribution of emissions at 1950?
liens 467-472:
Here you clearly state that full accounting includes LASC. Then you vaguely backtrack and mention the derived LASC without context, which is the political context of attribution for policy. The “arbitrary length of their simulation period” is a given fact for all modeling and is handled by explicit definition of the period in question. For example, climate change is defined in reference to the “pre-industrial” period (usually pre-1850 equilibrium).
Figures
Your figures and supplemental figures are out of order with respect to references in the text.
Consider using the same scale across the regional comparison plots, both in the paper and in the supplemental.
Appendix A1
lines 484-485:
This statement appears incorrect based on the discussion that follows.
It appears that the variation may arise from models with different ecological and productivity responses to the cycled environmental conditions.
-
AC3: 'Reply on RC3', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Wolfgang Obermeier, 20 Mar 2021
Status: closed
-
RC1: 'Comment on esd-2020-93', Anonymous Referee #1, 18 Jan 2021
In this study, the authors analyze three DGVM-derived fLULCC estimations for twelve models within 18 regions and quantify their differences as well as climate- and CO2-induced components. Results showed a global fLULCC of 2.0 ± 0.6 PgC per year for 2009–2018, of which ∼40% are attributable to the LASC. Regional hotspots of high cumulative and annual LASC values are found in the USA, China, Brazil, Equatorial Africa and Southeast Asia, mainly due to deforestation for cropland. Distinct negative LASC estimates, in Europe (early reforestation) and from 2000 onward in the Ukraine (recultivation of post-Soviet abandoned agricultural land), indicate that fLULCC estimates in these regions are lower in transient DGVM- compared to bookkeeping-approaches. By unraveling spatio-temporal variability in three alternative DGVM-derived fLULCC estimates, our results call for a harmonized attribution of model-derived fLULCC. This study proposes an approach that bridges bookkeeping and DGVM approaches for fLULCC estimation by adopting a mean DGVM-ensemble LASC for a defined reference period. I would recommend this work for publication with few minor modifications.
Specific comment:
Line 130: More introduction about” gridded output” is needed. For example, the resolution of these data. Monthly data or Annual data?Line140: All abbreviations must be explained. For example, HYDE and FAO.
-Line 154: ’the amount of precipitation in the Poyang Lake Basin’ was not consistent with the caption.
-Line 166: Descriptions of three alternative fLULCC are not clear in the current version.Line 385: I have some serious concern about the assumption that the last 100 years due to climate change –clarify it?
Eq 1,2,3: I really had difficulty in understanding these equations. I suggest the authors made them easy to follow in the revised manuscript.
-Figure 1 box4 presents fLULCC differences, but no information about different line.
-
AC1: 'Reply on RC1', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Wolfgang Obermeier, 20 Mar 2021
-
RC2: 'Comment on esd-2020-93', Wei Li, 28 Jan 2021
The difference in Eluc estimates between DGVMs and bookkeeping models has been reported in many previous studies but never been quantified by decomposing into more specific fluxes like LASC. The authors filled this research gap using fractional simulations by various DGVMs. The analysis is very detailed and comprehensive with precise definitions of different LULCC fluxes and components and corresponding quantifications. It is a significant step in disentangling the Eluc components on top of Pongratz et al. (2014) and Gasser & Ciais (2013) and has important implications on the definition of Eluc in the global carbon budget and implementing climate mitigation measures. The manuscript is well written with clear description and detailed supplementary materials. I see no major flaw in this manuscript and thus suggest this work for publication with few minor revisions.
I have some concerns that may need some clarification and discussion.
Although the LASC explains the Eluc difference between DGVMs and bookkeeping models, it would be better to address whether LASC exists in the real fluxes of carbon emission and sink, i.e. whether can be observed. If I understood correctly, positive LASC represents the loss of potential carbon sink and thus didn’t physically exist, i.e. this part of CO2 wasn’t released into the atmosphere. For the negative LASC like in reforestation, this part should be physically stored in the biomass or soil C pools, right?
As the authors stated, the timing of LULCC matters. Therefore, from my understanding, estimation of the accurate LASC for a specific LULCC event (a deforestation event in 1950 for an example) needs simulations similar to the S0 and S4 but using the climate and CO2 status when the LULCC event occurs (i.e. 1950) instead of the pre-industrial climate and CO2 status. Although the authors came to this point somewhere in the text, it would be better to emphasize this point explicitly.
The attribution to climate and CO2 in Sect. 2.2.2 is rather uncertain. Is there any observation data (e.g. FACE + warming experiment field data) that can be used to validate this attribution method?
Fig. 5-7 show results from each DGVM, but not reported in the text. Could add one or two sentences to say which models always give high e.g. LASC in Fig.5 or all models are similar?
L29-30: “high-latitudes” usually refers to boreal region. I think much early agricultural expansion occurred in the temperate regions. May rephrase.
L248: bookkeping -> bookkeeping
L253, L256: What are the numbers in the brackets?
L274: What are the pulses of the purple line in Fig. 2c?
L312-313: Why reforestation increased fLULCC_pd?
-
AC2: 'Reply on RC2', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Wolfgang Obermeier, 20 Mar 2021
-
RC3: 'Comment on esd-2020-93', Anonymous Referee #3, 08 Feb 2021
Review of esd-2020-93:
Modelled land use and land cover change emissions - A spatio-temporal comparison of different approaches
Summary
The authors present a detailed analysis of multi-dynamic-ecosystem-model estimates of LULCC emissions under different initial and environmental conditions, with some comparison with recent bookkeeping approaches. A main focus is the contribution of the “Loss of Additional Sink Capacity” (LASC) to emissions, which can be best estimated by dynamic ecosystem models with transient environmental conditions. The authors conclude that LASC is considerable, and has some regional variation. Furthermore, they show that CO2 effects dominate the contributions to LASC over climate effects. They also raise the question of whether LASC is appropriate for a harmonized attribution of LULCC emissions.
Overall review
This is an important contribution to LULCC and carbon/climate science and presents interesting results from the trendy model ensemble. There are a few main issues that need to be addressed, however, prior to publication. These are summarized here, with further detail following:
1) There seems to be a better way to calculate the CO2 vs climate estimates. You have simulations that you can difference directly, rather then using the ratio method that has limitations.
2) The results and discussion of regional LASC are confusing and not to the point. These can be better constrained and made more clear with main points in mind.
3) The suggested “harmonized attribution” of LULCC emissions is not consistent with the paper findings and doesn’t differentiate between science and policy. The real issue here is full scientific estimates of emissions, as the authors clearly show require the inclusion of LASC, vs attribution of LULCC emissions for policy purposes, which is a political issue. This paper isn’t about harmonizing different model approaches, but rather clarifying and understanding their differences. This provides scientific bases for accurate estimation of emissions, and provides info for political decisions regarding accounting for attribution policy, but doesn’t show an answer to the “correct” accounting for policy or whether/how the different approaches should be “harmonized.”
Specific comments and suggestions
Abstract
The abstract focuses more on the argument for LASC and DGVMs, than on the results of the study. The argument is better suited for the discussion, with a brief statement in the abstract stating that you use DGVMs to estimate LASC. Then you can add more findings/explanation to the abstract, such as temporal transitions in PTD that support the DGVM vs bookkeeping difference statements.
line 8:
“arbitrary chosen” is unnecessary.
lines 10-11:
Constant environmental forcing does not generate a condition independent of timing and legacy of lulcc. It generates a condition independent of changing environmental conditions, which is how this should be described.
Ensuring that the lulcc magnitude and trajectory are identical across simulations, along with constant environmental forcing, creates a condition almost independent of the timing and legacy of lulcc, as long as the transient dependence isn’t the question. Full independence in this case requires that the biogeochemical dynamics are also identical across simulations, such as may be possible when a single model is used. But this is not possible across different models because each model will respond differently to the same exact lulcc (which is also very difficult to achieve across models), even if the environmental forcing is identical.
I think you mean this in the context of a single-model experiment, with the exact same lulcc and environmental forcing, and not focused on how the LULCC trajectory determines emissions (i.e., path dependence), but this needs to be qualified here as it sounds more general.
lines 21-23:
I am not sure that your analysis calls for this, nor that your approach bridges the two. The issue is a full scientific accounting of emissions vs a partial accounting for policy purposes. While you have evidence for a full scientific accounting, you still have little basis for a partial-LASC approach.
Introduction
lines 28-30:
awkward sentence. delete “e.g.” and use three clauses: deforestation…high lats, tropical deforestation, and recent forest expansion in high lats.
line 31:
“…contributed approximately one-third of global anthropogenic…”
line 35:
delete “in line.” then: flulcc may also gain…
line 42:
“…but also change in…”
line 50:
delete “allow to”
lines 69-72:
The language here starts to confuse the definition and example of LASC. It sounds like just having the potential vegetation or the foregone sink is part of LASC, in addition to the difference of environmental effects on the managed vs unmanaged vegetation. These two sentences should be reworded to clarify that “assuming potential vegetation” and “capture the foregone sinks a given LULCC event destroys” do not obfuscate the definition of LASC.
lines 71-72:
this statement is not correct. there is no guarantee that the change in environmental conditions is increasing LASC at the time of change. so “same direction” does not make sense. I think you are still referring to the example here, but this is a more general concluding statement.
it is more correct to state that these emissions “can” accumulate in the absence of further LULCC, depending on the environmental conditions. which you then give examples of in the next paragraph.
lines 98-99:
I don’t agree that constant conditions create a condition independent of lulcc timing. regardless of the forcing, the time since disturbance makes a difference, especially in the context of DGVMs where ecosystems are not static. A subsequent LULCC has a different emission depending on the period between it and the previous LULCC. And an event early in time on ‘pristine’ land may affect a different biomass then one later in time, even if the conditions are identical. Even with a spun-up model, the biomass may not remain constant during a simulation, and depending on the resolution and application of LULCC, a later LULCC may or may not be applied consistently to ‘pristine’ or ‘managed’ area (and corresponding biomass), or some combination of the two.
I think this description should be changed to “independent of environmental trends” or something like this.
line 108:
delete “in line”
lines 109-110:
unclear: “and consequently into the natural land c sink.” how is this related to ‘included or excluded?’
Data and methods
lines171-172 and 178-179 and 194-195:
These really seem like a single equation each.
line 188:
again, this isn’t true. timing and legacy of LULCC are inherently critical and determining elements in any transient LULCC simulation. these simulations are indifferent to environmental trends or changing environmental conditions.
line 219:
“…C stock changes…”
lines 220-240:
Why did you do it this way, particularly for climate effects? You can get co2+ndep-related emissions directly by subtracting S0 and S1, which would be interesting to compare with your estimate of co2 only emissions using this ratio method.
And you can get the climate-related emissions directly by subtracting S1 and S2, including interactions with CO2, without using this indirect ratio method. and your assumption regarding zero interactions between climate and co2 is not sound because S2 includes both at the same time, and S1 includes only co2. I suggest you use the actual nbp difference of these 2 sims to estimate the climate effects.
Also, you don’t report on the co2+no2 effects at all, so why is this simulation listed in the methods?
Results and discussion
lines 246-247:
unclear: “…to consistently use the same models for the flux and bias estimates on a spatio-temporal level…”
line 254:
did you use the trendyv9 version of sdgvm throughout? if so, you need to state this in the methods.
lines 256-257:
not necessarily. the single year pre-ind estimate is closer to bk estimates, and is equally close to gasser as the present-day for the decade.
line 259:
you haven’t shown the cumulative yet
line 260:
awkward. try “…in the 1950’s and again at the end of the simulation…”
line 272:
“…comes into play…”
lines 272-273:
awkward first half of sentence.
line 277:
during this period (first half of 20th century)
line 279:
“these peaks”
lines 281-284:
maybe because the LULCC occurrence is the same between simulations. but post-peak carbon stocks might have been lowered enough differentially between the sims to reduce the post-peak difference because carbon reductions are generally implemented as fractions, in which case the same LULCC would reduce more carbon in the pd sim than in the pi sim because of higher carbon stocks in the pd sim. and the eed peak is actually after the other peaks and stretched out. so is this really independent of the lulcc timing, as there is a peak in lulcc area that coincides with this emission peak.
line 293:
post 1950
lines 296-320:
This section is difficult to follow, and it isn’t entirely clear what the main point is, other than that LULCC dominates the EED pattern. But more importantly, LULCC seems to drive the fLULCC pattern also, since the EED and the fLULCC patterns also match. temporally, which is does not need a regional breakdown to be shown. And it would appear that the difference in biomass between Pi and Pd is a main driver of the corresponding difference in fLULCC. This boils down to biomass+LULCC controlling fLULCC estimates, with some regional variation in how this relationship contributes to the global estimates.
If the EED pattern is the main point, then take out the other regional plots (5 and 6), as figs 3 and 4 and 7 and 8 show this. The regional plots are difficult to interpret, and so the discussion is helpful, but it should be more clear how the regional patterns contribute to the global patterns, as figure 8 can show (and 7 can show magnitudes better is the scales are matched).
Plots 9-11 are relevant, but they are not used to explain why the EED pattern dominates fLULCC, but they can support the basic relationship between biomass, LULCC trajectory, and fLULCC.
There are too many plots that are too difficult to read, and their referencing is difficult to follow also.
lines 302-303:
The regional comparison plots (figures 5 - 7 and the supplemental) are difficult to compare because they are all on different scales. while it may make it more difficult to see some of the individual lines, putting them on the same scale is the best way to show the differences, which I think is the point.
You can split these plots into multiple figures also because they are so small, which also makes them difficult to read. If that is too many figures, then maybe all the regional plots should be supplemental.
lines 321-352:
This section is difficult to follow, and it isn’t clear what the point is of the comparison between LASC and EED. LASC is a component of fLULCC estimation, and EED is a difference between two estimates of fLULCC that don’t include LASC. In general, the dynamics discussed here are still driven by biomass+LULLCC, since the forests are the high biomass areas and where LASC will have the most pronounced influence. While the regional breakdown and LASC discussion is somewhat informative, as the LASC dynamics are interesting and a key part of this study and since there is some regional variation, the relationships between LASC and EED seem to still be similar across regions, in that LASC tends to grow over time due to LULCC timing, while EED is driven mainly by biomass differences and the fixed LULCC trajectory. So is the point here to just discuss LASC dynamics, or to support the hypothesis that LASC is critical to fLULCC estimates because static biomass conditions do not include temporal effects of changing conditions (and drive errors associated with these biomass conditions)? I don’t think you need the EED stuff here, as your larger point about estimation methods is discussed elsewhere (section 4?).
lines 353-377:
This discussion is a bit more clear as it focuses on LASC dynamics and doesn’t mix in references to other calculations. I suggest combining this section and the previous section and focusing just on LASC dynamics. And the last sentence seems less a concluding sentence and more an evidence sentence to support the discussion.
line 383-384:
What about LULCC effects due to abandonment of agricultural land? This also contributes to higher biomass in some places, and isn’t trivial. You show some evidence of this in figure 12, where it can outweigh the climatic effects in even the long term, but your net results over the whole period do not show how regrowth has contributed more recently to increased biomass in places where this doesn’t outweigh previous emissions.
lines 416-445:
This section seems to be referring to a political question for emission attribution to regions for policy, rather than a scientific question of what are LULCC emissions.
So you should separate these two questions/recommendations. Is your scientific recommendation to discard some LULCC emissions? It doesn’t seem to be so as you clearly show that LASC is an important component in “Accurate quantification of the net carbon flux from land use and land cover changes…” (line 447). Whether this makes sense for regional carbon policy is a different question.
From a policy standpoint, the question here is then why include only part of the total emissions from LULCC (i.e. LASC from second half of 21st century) while including some estimate of the directly effects of changes all the way back to 1850? If it isn’t ok to include LASC back to 1850, why should any effects of pre-1950 LULCC be included? Why not just start full attribution of emissions at 1950?
liens 467-472:
Here you clearly state that full accounting includes LASC. Then you vaguely backtrack and mention the derived LASC without context, which is the political context of attribution for policy. The “arbitrary length of their simulation period” is a given fact for all modeling and is handled by explicit definition of the period in question. For example, climate change is defined in reference to the “pre-industrial” period (usually pre-1850 equilibrium).
Figures
Your figures and supplemental figures are out of order with respect to references in the text.
Consider using the same scale across the regional comparison plots, both in the paper and in the supplemental.
Appendix A1
lines 484-485:
This statement appears incorrect based on the discussion that follows.
It appears that the variation may arise from models with different ecological and productivity responses to the cycled environmental conditions.
-
AC3: 'Reply on RC3', Wolfgang Obermeier, 20 Mar 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2020-93/esd-2020-93-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Wolfgang Obermeier, 20 Mar 2021
Wolfgang A. Obermeier et al.
Wolfgang A. Obermeier et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 637 | 179 | 26 | 842 | 9 | 11 |
- HTML: 637
- PDF: 179
- XML: 26
- Total: 842
- BibTeX: 9
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1