the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Feb 2024
Holocene land cover change in North America: continental trends, regional drivers, and implications for vegetation-atmosphere feedbacks
Abstract. Land cover governs the biogeophysical and biogeochemical feedbacks between the land surface and atmosphere. Holocene vegetation-atmosphere interactions are of particular interest, both to understand the climate effects of intensifying human land use and as a possible explanation for the Holocene Conundrum, a widely studied mismatch between simulated and reconstructed temperatures. Progress has been limited by a lack of data-constrained, quantified, and consistently produced reconstructions of Holocene land cover change. As a contribution to the Past Global Changes (PAGES) LandCover6k Working Group, we present a new suite of land cover reconstructions with uncertainty for North America, based on a network of 1445 sedimentary pollen records and the REVEALS pollen-vegetation model coupled with a Bayesian spatial model. These spatially comprehensive land cover maps are then used to determine the pattern and magnitude of North American land cover changes at continental to regional scales. Early Holocene afforestation in North America was driven by rising temperatures and deglaciation, and this afforestation likely amplified early Holocene warming via the albedo effect. A continental-scale mid-Holocene peak in summergreen trees and shrubs (8.5 to 4 ka) is hypothesized to represent a positive and understudied feedback loop among insolation, temperature, and phenology seasonality. A last-millennium decrease in summergreen trees and shrubs with corresponding increases in open land likely was driven by a spatially varying combination of intensifying land use and neoglacial cooling. Land cover trends vary within and across regions, due to individualistic taxon-level responses to environmental change. Major species-level events, such as the mid-Holocene decline of eastern hemlock, may have altered regional climates. The substantial land-cover changes reconstructed here support the importance of biogeophysical vegetation feedbacks to Holocene climate dynamics. However, recent model experiments that invoke vegetation feedbacks to explain the Holocene Conundrum may have overestimated the land cover forcing by replacing Northern Hemisphere grasslands >30° N with forests; an ecosystem state that is not supported by these land cover reconstructions. These Holocene reconstructions for North America, along with similar LandCover6k products now available for other continents, serve the Earth system modeling community by providing better-constrained land cover scenarios and benchmarks for model evaluation, ultimately making it possible to better understand the regional- to global-scale processes driving Holocene land cover dynamics.
- Preprint
(3269 KB) - Metadata XML
-
Supplement
(5860 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on cp-2024-6', Anonymous Referee #1, 04 Apr 2024
- AC2: 'Reply on RC1', Andria Dawson, 16 May 2024
-
RC2: 'Outstanding achievement; a few points require further clarification and improved presentation', Jed Kaplan, 06 Apr 2024
Review of
Holocene land cover change in North America: continental trends, regional drivers, and implications for vegetation-atmosphere feedbacks
By Andria Dawson, et al.
This manuscript describes a study to reconstruct land cover for the Holocene over North America. As part of the LandCover6k initiative, the methodology follows a standardized procedure: first pollen records from sedimentary archives are synthesized and samples are assigned ages using up-to-date age-depth models. Then, pollen spectra are simplified and decimated to include specific taxa, and relative abundances of these taxa are passed to the REVEALS pollen-landscape model. REVEALS generates quantitative estimates of land cover for specific taxa that can be further generalized into broad groups of plant functional cover, e.g., broadleaf deciduous or needleleaf evergreen trees. These point-based data are then interpolated to a continuous 1-degree grid covering the study area. The work presented here complements similar activities undertaken for other parts of the Northern Hemisphere and ongoing work in the tropics and elsewhere. The authors present the results of the synthesis in the form of gridded maps and synthetic timeseries covering the entire North America spatial domain, and for specific regions that they analyze in further detail.
Overall, this is an excellent study that is rigorous in its methodology, interesting and in some ways novel in terms of results, and honest about shortcomings. The authors helpfully provide a roadmap for future research including on improving the land cover reconstructions and recommendations for research that could employ the maps and other datasets produced here. There are a few issues that should be clarified before publication, and ultimately this paper and the associated datasets will make a valuable contribution to the journal and support range of fields in further study.
General comments
While changes in ice cover were considered, it appears that sea level changes (and proglacial lakes) were ignored in this study. This is a major limitation of the spatial analyses and at the very least should be justified. It’s a bit strange because these paleogeographic changes are considered in previous, similar studies by some of the same authors (e.g., Williams, 2003; Williams et al., 2004). The early Holocene is characterized by very large proglacial lakes at the margin of the Laurentide Ice. More importantly were the postglacial isostatic adjustments that lasted throughout the Holocene. For example, the Hudson Bay Lowlands were submerged until after 5ka and low-lying areas of the Atlantic coast and Florida had significantly more land area exposed in the early Holocene. Data on sea level changes, for example from the PAGES PALSEA activity would be worth considering, and citing in an explanation of why these were not part of the current study.
In the interpolated maps, the parts of the study domain that show no data I assume are because the “confidence region” (CR) was greater than the threshold of 9, for example in much of Mexico in the early Holocene. It would be helpful to see the CR maps themselves included among the supplementary figures. Looking at Figure 1, there are only 3 or maybe 4 sites in all of Mexico, so it is hard to understand, especially given the climatic and topographic diversity of Mexico, that there is much power in the interpolations over that space.
All of the data products presented here (point-based and gridded maps) must be freely released on zenodo.org or other open-access data repository that provides a DOI upon final publication of the paper. The gridded maps should be provided in the earth system modeling-standard netCDF format.
A few notes on presentation
As “land use” is generally accepted to be an activity that is unique to humans, it is not necessary to qualify the term with “human land use” or “anthropogenic land use” in the manuscript. In the interest of conciseness, please just use “land use” alone throughout the manuscript, or maybe define it once at the beginning of the text.
I found the constant switching back and forth between scientific names and common names for taxa distracting and sometimes confusing. Use of both nomenclatures even occurs in a single sentence (e.g., lines 435-436). I ask the authors to pick one nomenclature system and stick with it throughout the entire manuscript.
Please use a thinner line thickness in all of the maps presented in the manuscript and supplement. The heavy line weight around the ice sheets and coastline distracts from the content. Perhaps the ice sheets could be plotted in a blue or brighter, contrasting color as polygons, without any outline at all.
Specific comments
Lines 48-49
It is not at all clear how changes in the abundance of hemlock could have been significant enough to have a biogeophysical feedback to climate; see further comments below.
Lines 168-169
Please explain briefly how relative abundances are calculated when some taxa are ignored? Is there an "all other taxa" bin? Or are only abundances relative to the considered taxa included? What happens when a taxon that is considered to be important in terms of land cover, even locally, is not part of those used in the REVEALS model?
Line 235
Approximately how does the grid resolution of the 1x1 degree interpolated surface compare to the 10,000km2 area represented by a REVEALS reconstruction noted on line 179? Naturally it changes by latitude, but it would be helpful to put a comparative statement here.
Lines 254-255
Here where CR is introduced, it would be good to call out supplementary figures here showing this value in map form for all periods.
Lines 260-261
I understand that the LandCover6k grid was specified as 1x1 geographic degrees, but wouldn’t it have made more sense to do the original work on an equal-area grid and then only reproject the data in a final step? At the very least it would have made interpretation of the maps more straightforward, and would be similar to earlier work (Williams, 2003; Williams et al., 2004).
Figure 1
Could you plot the 1x1 degree graticule on this map using a very thin line in an unobtrusive color? It would make interpretation of the grid resolution of the other maps easier.
Figure 1
Given the very high density of sites, it seems strange that nearly all of Minnesota is not included in any of the regional boxes. The choice to exclude this area deserves some explanation.
Figure 2
Use a thinner line weight, or no line at all for the ice sheet outline (as noted above)
Figure 2
To aid in quickly interpreting the plots and to provide better consistency with the rest of the figures, please plot the land cover fractional surfaces in the same colors as used in Fig. 3 and the other timeseries plots. That is to say, plot the first column of maps in shades of green, the second in shades of blue, and the third in shades of orange.
Figure 2
As noted in the supplement the three interpolated surfaces sum across to 100% in each row, and there is no “missing” fraction that represents bare ground. In the Arctic and in desert areas, the landscape is not 100% vegetated. This information should not be buried in the supplement, and needs to be clearly noted when the main figures are presented in the figure caption and body text. It should further be noted as a limitation and explained why this is not the case in the main manuscript text.
Line 311
Given that there are only 3 sites in Mexico, is the spatial domain of the study justified? Wouldn't a maximum distance buffer around nearest site be better - e.g., up to 100 km apart (corresponding to the REVEALS indicative catchment area)? As noted above there is a distance filter on the grid based on the CR value, but it would be interesting to see how this translates into distance from a site. Some statistics, such as the max distance from any site in the interpolation, would be helpful, even if only in the supplementary materials.
Line 320
The number of gridcells contributing to the curves presented in Figure 3 changes based on ice area, apparently not sea level, but also CR value. Can we see an additional curve on this figure showing the total area in the spatial domain contributing to the cover estimate?
Lines 435-436
In this sentence, and others, please just choose one form of plant nomenclature or the other, and stick with it.
Lines 541-543
“… desert, steppe, and other open-land arid ecosystems are likely to be underrepresented in these reconstructions, due to a scarcity of dryland sites” yet the interpolated maps and timeseries curves imply continuous vegetation cover (without bare ground), if I understand correctly. This limitation of the methodology should be further described and justified.
Line 566
Is there really nothing to say here about sea level dynamics over the period?
Line 625
I suggest a small rewording of this sentence to: “During the late Holocene, the growth of Indigenous populations and intensification of land use in the Americas had increasing effects on land cover. Understanding the interactions among…”
Line 632
Evidence for dense populations and land use in North America are dismissed here, yet a number of examples of this are provided in the following paragraph. This sentence could be reworded to better tie to what is coming next.
Paragraph starting on line 660
What is the purpose of this paragraph? Can it be tied back to the data presented in the current study?
Section starting on line 668 (4.1.1)
This section needs to be tied back more clearly to the findings in the current study, at least speculatively. The section reads like a review paper now and there is nothing new in here.
Line 675-676
The full name of the “TEMPO” acronym could be removed here and just put in the bibliography.
Paragraph starting on line 707
This paragraph does a very good job of explaining how the data synthesized in the current study ties back to previous work. It should be a model for how section 4.1.1 could be improved.
Line 715
“Great Plains”
Line 728-731
It is not clear from this section or from the maps or timeseries how large, in absolute terms, the coverage of T. Canadensis could have ever been. The paragraph seems to insinuate that it could have been abundant enough to make a majority proportion of forest cover, therefore having a strong influence on, e.g., albedo. But… (see next comment)
Lines 731-732
Am I missing something because I don't see a shift in the dominance in Fig. 4, which is always more than 50% summergreen trees and shrubs, with evergreen less than 30% cover fraction throughout the Holocene. Are you arguing that ETS forests were conifer-dominated? Otherwise, the albedo changes would have been very subtle, especially since T. Canadensis can persist in the understory for a century or longer and so while it is there and producing pollen, it will have no influence on summer albedo and relatively little on winter.
Lines 741-743
Looking at the summary figures, these changes must have all be very subtle. If not, then some further quantitative information should be highlighted here.
Line 747
Broadleaf summergreen trees have greater maximum evapotranspiration rates than needleleaf evergreens and this effect should also be mentioned here as it is probably more important than the summertime albedo differences.
Line 753
I am not convinced that there is anything more than “relatively subtle shifts in the proportions of summergreen and evergreen trees and shrubs” shown in the data presented here.
Line 775-776
If “… REVEALS estimates are sensitive to parameter choices…” then why didn't you not just explore a larger parameter space and make a range of reconstructions? Instead of just one? Seems like it would be an easy change and could lead to the preparation of a range of maps or uncertainty fields.
Line 784-785
The sentence mentions that “…this approach does not mechanistically represent the underlying processes that link pollen to vegetation”. The GMRF method also does not account for soil, slope, aspect, and other edaphic controls on vegetation cover. This should be mentioned.
Line 828-830
Here it is admitted that the changes in “… continental-scale fractional forest cover were
broadly stable.” This statement does not seem to support the idea that biogeophysical feedbacks between land and atmosphere would have been very important, in contrast to what is insinuated earlier in the manuscript. Some further explanation would be helpful here.
References
Williams, J. W. (2003). Variations in tree cover in North America since the last glacial maximum. Global and Planetary Change, 35(1-2), 1-23. doi:10.1016/S0921-8181(02)00088-7
Williams, J. W., Shuman, B. N., Webb, T., Bartlein, P. J., & Leduc, P. L. (2004). Late-Quaternary Vegetation Dynamics in North America: Scaling from Taxa to Biomes. Ecological Monographs, 74(2), 309-334. doi:10.1890/02-4045
Citation: https://doi.org/10.5194/cp-2024-6-RC2 - AC1: 'Reply on RC2', Andria Dawson, 16 May 2024
Status: closed
-
RC1: 'Comment on cp-2024-6', Anonymous Referee #1, 04 Apr 2024
- AC2: 'Reply on RC1', Andria Dawson, 16 May 2024
-
RC2: 'Outstanding achievement; a few points require further clarification and improved presentation', Jed Kaplan, 06 Apr 2024
Review of
Holocene land cover change in North America: continental trends, regional drivers, and implications for vegetation-atmosphere feedbacks
By Andria Dawson, et al.
This manuscript describes a study to reconstruct land cover for the Holocene over North America. As part of the LandCover6k initiative, the methodology follows a standardized procedure: first pollen records from sedimentary archives are synthesized and samples are assigned ages using up-to-date age-depth models. Then, pollen spectra are simplified and decimated to include specific taxa, and relative abundances of these taxa are passed to the REVEALS pollen-landscape model. REVEALS generates quantitative estimates of land cover for specific taxa that can be further generalized into broad groups of plant functional cover, e.g., broadleaf deciduous or needleleaf evergreen trees. These point-based data are then interpolated to a continuous 1-degree grid covering the study area. The work presented here complements similar activities undertaken for other parts of the Northern Hemisphere and ongoing work in the tropics and elsewhere. The authors present the results of the synthesis in the form of gridded maps and synthetic timeseries covering the entire North America spatial domain, and for specific regions that they analyze in further detail.
Overall, this is an excellent study that is rigorous in its methodology, interesting and in some ways novel in terms of results, and honest about shortcomings. The authors helpfully provide a roadmap for future research including on improving the land cover reconstructions and recommendations for research that could employ the maps and other datasets produced here. There are a few issues that should be clarified before publication, and ultimately this paper and the associated datasets will make a valuable contribution to the journal and support range of fields in further study.
General comments
While changes in ice cover were considered, it appears that sea level changes (and proglacial lakes) were ignored in this study. This is a major limitation of the spatial analyses and at the very least should be justified. It’s a bit strange because these paleogeographic changes are considered in previous, similar studies by some of the same authors (e.g., Williams, 2003; Williams et al., 2004). The early Holocene is characterized by very large proglacial lakes at the margin of the Laurentide Ice. More importantly were the postglacial isostatic adjustments that lasted throughout the Holocene. For example, the Hudson Bay Lowlands were submerged until after 5ka and low-lying areas of the Atlantic coast and Florida had significantly more land area exposed in the early Holocene. Data on sea level changes, for example from the PAGES PALSEA activity would be worth considering, and citing in an explanation of why these were not part of the current study.
In the interpolated maps, the parts of the study domain that show no data I assume are because the “confidence region” (CR) was greater than the threshold of 9, for example in much of Mexico in the early Holocene. It would be helpful to see the CR maps themselves included among the supplementary figures. Looking at Figure 1, there are only 3 or maybe 4 sites in all of Mexico, so it is hard to understand, especially given the climatic and topographic diversity of Mexico, that there is much power in the interpolations over that space.
All of the data products presented here (point-based and gridded maps) must be freely released on zenodo.org or other open-access data repository that provides a DOI upon final publication of the paper. The gridded maps should be provided in the earth system modeling-standard netCDF format.
A few notes on presentation
As “land use” is generally accepted to be an activity that is unique to humans, it is not necessary to qualify the term with “human land use” or “anthropogenic land use” in the manuscript. In the interest of conciseness, please just use “land use” alone throughout the manuscript, or maybe define it once at the beginning of the text.
I found the constant switching back and forth between scientific names and common names for taxa distracting and sometimes confusing. Use of both nomenclatures even occurs in a single sentence (e.g., lines 435-436). I ask the authors to pick one nomenclature system and stick with it throughout the entire manuscript.
Please use a thinner line thickness in all of the maps presented in the manuscript and supplement. The heavy line weight around the ice sheets and coastline distracts from the content. Perhaps the ice sheets could be plotted in a blue or brighter, contrasting color as polygons, without any outline at all.
Specific comments
Lines 48-49
It is not at all clear how changes in the abundance of hemlock could have been significant enough to have a biogeophysical feedback to climate; see further comments below.
Lines 168-169
Please explain briefly how relative abundances are calculated when some taxa are ignored? Is there an "all other taxa" bin? Or are only abundances relative to the considered taxa included? What happens when a taxon that is considered to be important in terms of land cover, even locally, is not part of those used in the REVEALS model?
Line 235
Approximately how does the grid resolution of the 1x1 degree interpolated surface compare to the 10,000km2 area represented by a REVEALS reconstruction noted on line 179? Naturally it changes by latitude, but it would be helpful to put a comparative statement here.
Lines 254-255
Here where CR is introduced, it would be good to call out supplementary figures here showing this value in map form for all periods.
Lines 260-261
I understand that the LandCover6k grid was specified as 1x1 geographic degrees, but wouldn’t it have made more sense to do the original work on an equal-area grid and then only reproject the data in a final step? At the very least it would have made interpretation of the maps more straightforward, and would be similar to earlier work (Williams, 2003; Williams et al., 2004).
Figure 1
Could you plot the 1x1 degree graticule on this map using a very thin line in an unobtrusive color? It would make interpretation of the grid resolution of the other maps easier.
Figure 1
Given the very high density of sites, it seems strange that nearly all of Minnesota is not included in any of the regional boxes. The choice to exclude this area deserves some explanation.
Figure 2
Use a thinner line weight, or no line at all for the ice sheet outline (as noted above)
Figure 2
To aid in quickly interpreting the plots and to provide better consistency with the rest of the figures, please plot the land cover fractional surfaces in the same colors as used in Fig. 3 and the other timeseries plots. That is to say, plot the first column of maps in shades of green, the second in shades of blue, and the third in shades of orange.
Figure 2
As noted in the supplement the three interpolated surfaces sum across to 100% in each row, and there is no “missing” fraction that represents bare ground. In the Arctic and in desert areas, the landscape is not 100% vegetated. This information should not be buried in the supplement, and needs to be clearly noted when the main figures are presented in the figure caption and body text. It should further be noted as a limitation and explained why this is not the case in the main manuscript text.
Line 311
Given that there are only 3 sites in Mexico, is the spatial domain of the study justified? Wouldn't a maximum distance buffer around nearest site be better - e.g., up to 100 km apart (corresponding to the REVEALS indicative catchment area)? As noted above there is a distance filter on the grid based on the CR value, but it would be interesting to see how this translates into distance from a site. Some statistics, such as the max distance from any site in the interpolation, would be helpful, even if only in the supplementary materials.
Line 320
The number of gridcells contributing to the curves presented in Figure 3 changes based on ice area, apparently not sea level, but also CR value. Can we see an additional curve on this figure showing the total area in the spatial domain contributing to the cover estimate?
Lines 435-436
In this sentence, and others, please just choose one form of plant nomenclature or the other, and stick with it.
Lines 541-543
“… desert, steppe, and other open-land arid ecosystems are likely to be underrepresented in these reconstructions, due to a scarcity of dryland sites” yet the interpolated maps and timeseries curves imply continuous vegetation cover (without bare ground), if I understand correctly. This limitation of the methodology should be further described and justified.
Line 566
Is there really nothing to say here about sea level dynamics over the period?
Line 625
I suggest a small rewording of this sentence to: “During the late Holocene, the growth of Indigenous populations and intensification of land use in the Americas had increasing effects on land cover. Understanding the interactions among…”
Line 632
Evidence for dense populations and land use in North America are dismissed here, yet a number of examples of this are provided in the following paragraph. This sentence could be reworded to better tie to what is coming next.
Paragraph starting on line 660
What is the purpose of this paragraph? Can it be tied back to the data presented in the current study?
Section starting on line 668 (4.1.1)
This section needs to be tied back more clearly to the findings in the current study, at least speculatively. The section reads like a review paper now and there is nothing new in here.
Line 675-676
The full name of the “TEMPO” acronym could be removed here and just put in the bibliography.
Paragraph starting on line 707
This paragraph does a very good job of explaining how the data synthesized in the current study ties back to previous work. It should be a model for how section 4.1.1 could be improved.
Line 715
“Great Plains”
Line 728-731
It is not clear from this section or from the maps or timeseries how large, in absolute terms, the coverage of T. Canadensis could have ever been. The paragraph seems to insinuate that it could have been abundant enough to make a majority proportion of forest cover, therefore having a strong influence on, e.g., albedo. But… (see next comment)
Lines 731-732
Am I missing something because I don't see a shift in the dominance in Fig. 4, which is always more than 50% summergreen trees and shrubs, with evergreen less than 30% cover fraction throughout the Holocene. Are you arguing that ETS forests were conifer-dominated? Otherwise, the albedo changes would have been very subtle, especially since T. Canadensis can persist in the understory for a century or longer and so while it is there and producing pollen, it will have no influence on summer albedo and relatively little on winter.
Lines 741-743
Looking at the summary figures, these changes must have all be very subtle. If not, then some further quantitative information should be highlighted here.
Line 747
Broadleaf summergreen trees have greater maximum evapotranspiration rates than needleleaf evergreens and this effect should also be mentioned here as it is probably more important than the summertime albedo differences.
Line 753
I am not convinced that there is anything more than “relatively subtle shifts in the proportions of summergreen and evergreen trees and shrubs” shown in the data presented here.
Line 775-776
If “… REVEALS estimates are sensitive to parameter choices…” then why didn't you not just explore a larger parameter space and make a range of reconstructions? Instead of just one? Seems like it would be an easy change and could lead to the preparation of a range of maps or uncertainty fields.
Line 784-785
The sentence mentions that “…this approach does not mechanistically represent the underlying processes that link pollen to vegetation”. The GMRF method also does not account for soil, slope, aspect, and other edaphic controls on vegetation cover. This should be mentioned.
Line 828-830
Here it is admitted that the changes in “… continental-scale fractional forest cover were
broadly stable.” This statement does not seem to support the idea that biogeophysical feedbacks between land and atmosphere would have been very important, in contrast to what is insinuated earlier in the manuscript. Some further explanation would be helpful here.
References
Williams, J. W. (2003). Variations in tree cover in North America since the last glacial maximum. Global and Planetary Change, 35(1-2), 1-23. doi:10.1016/S0921-8181(02)00088-7
Williams, J. W., Shuman, B. N., Webb, T., Bartlein, P. J., & Leduc, P. L. (2004). Late-Quaternary Vegetation Dynamics in North America: Scaling from Taxa to Biomes. Ecological Monographs, 74(2), 309-334. doi:10.1890/02-4045
Citation: https://doi.org/10.5194/cp-2024-6-RC2 - AC1: 'Reply on RC2', Andria Dawson, 16 May 2024
Supplement
Model code and software
Project workflow in R Andria Dawson et al. https://github.com/andydawson/reveals-na
Neotoma Lake Size Workflow Simon Goring et al. https://github.com/NeotomaDB/neotoma_lakes
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 711 | 279 | 47 | 1,037 | 40 | 50 | 37 |
- HTML: 711
- PDF: 279
- XML: 47
- Total: 1,037
- Supplement: 40
- BibTeX: 50
- EndNote: 37
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 401 | 38 |
| China | 2 | 152 | 14 |
| Germany | 3 | 83 | 8 |
| Canada | 4 | 55 | 5 |
| Sweden | 5 | 41 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 401