the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 May 2024
European forest cover during the Holocene reconstructed from pollen records
Abstract. Changes in tree cover influence many aspects of the Earth System. Recent regional changes in tree cover, as documented by remote-sensed observations, are insufficient to capture the response to large climate changes or to differentiate the impacts of human activities from natural drivers. Pollen records provide an opportunity to examine the causes of changes in tree cover in response to large climate changes in the past and during periods when human influence was less important than today. Here we reconstruct changes in tree cover in Europe through the Holocene using fossil pollen records, using the modelled relationship between observed modern tree cover and modern pollen samples. At a pan-European scale, tree cover is low at the beginning of the Holocene but increases rapidly during the early Holocene and is maximal at ca. 6,500 cal. BP, after which tree cover declines to present-day levels. The rapidity of the post-glacial increase in tree cover and the timing and length of maximum tree cover varies regionally, reflecting differences in climate trajectories during the early and mid-Holocene. The nature of the subsequent reduction in tree cover also varies, which may be due to differences in climate but may also reflect different degrees of human influence. The reconstructed patterns of change in tree cover are similar to those shown by previous reconstructions, but our approach is more robust and less data-demanding than previously applied methods and therefore provides a useful approach to reconstructing tree cover in regions where data limitations preclude the use of alternative methods.
- Preprint
(2911 KB) - Metadata XML
-
Supplement
(5288 KB) - BibTeX
- EndNote
Status: open (until 03 Aug 2024)
-
RC1: 'Comment on egusphere-2024-1523', Thomas Giesecke, 07 Jun 2024
reply
There is a need for assessing past forest cover change from pollen regionally and the REVEALS model requiring knowledge on pollen productivity may not be the method of choice where that information is lacking or incomplete. Thus it is useful to explore other avenues and the manuscript by Sweeney et al. does that. It also adds interesting comparisons between the European estimates produced applying the REVEALS model and the modern analogue technique. While I welcome the attempt of applying a regression, I have my doubts on the choice of predictor variables. The authors should demonstrate how % needle leave and the Shannon index improve a regression model for overall tree cover. I can see how elevation improves the model in the current situation but have my doubt that this variable will improve past reconstructions. Instead using information on over and underrepresented pollen could perhaps make this a real winner. It is also not clear to me in which way this regression model improves upon the modern analogue technique requiring the same input information and seemingly yielding a similar performance. The manuscript is not explaining how the proposed regression model reduces the bias of simply using arboreal percentage, which may be dominated by pine and birch versus elm and lime.
My second concern with the manuscript is the lack of appropriate recognition and citation of databases and initiatives that collected and curated the pollen data used here. Most of the modern and downcore pollen data used here was initially made available by the EPD/Neotoma or PANGAEA with a cc by 4 license requiring attribution and citation of this initial data release. Please see the recent discussion of the manuscript by Schild et al. (https://essd.copernicus.org/preprints/essd-2023-486/#discussion).
Specific comments
L. 57: The LRA includes local reconstructions (LOVE) which has not been applied on the European scale. Only the REVEALS model was used.
L. 58: You should cite Zanon et al. (2018) already here.
L. 63: The main focus was on reconstructing the proportion of open versus forest land cover.
L. 65: As an introductory overview this is almost too detailed while it is lacking studies to work as a good review of all that has come before: e.g. Pirzamanbein et al. (2014, Ecological Complexity), Roberts et al. (2018) Scientific Reports 8:716. Some of these appear in the discussion, but it would be good to mention them here already.
L. 80: Fall speeds are not the major issue as they can be estimated based on pollen size.
L. 86: Since you mention PFTs you may want to include Davis et al. (2015) here already not only in the discussion.
L. 115: The SPECIAL Modern Pollen Dataset (Villegas-Diaz and Harrison, 2022) compiles samples from other data sources including Neotoma and PANGAEA which also have a CC-BY-4.0 license, hence you need to cite or acknowledge the original data source not just the data compilation.
L. 116: SMPDS needs to be introduced. It is not clear from the above that this refers to the surface sample data.
L. 119: Particularly where core tops were used thus assumption is daring.
L. 122: Give a brief motivation not just a reference.
L. 122-124: Here you are referring to surface samples, core tops or Holocene records?
L. 126: So you include small bogs but exclude large bogs? I cannot find this constraint discussed in Githumbi et al. (2022).
L. 135: It would be useful to mention what is included in shrub pollen: Are you including dwarf shrubs like Calluna or rather taller perennial woody plants like Corylus and Juniperus?
L. 139: How did you deal with situations where alien tree plantations make up most forest cover: e.g. Eucalyptus. Also plantations of Pseudotsuga (0.83 million ha in Europe) may be a potential problem.
L. 140: Large proportions of Cyperaceae and Polypodiales are limited to bogs, excluding them would reduce the biases from including bog samples.
L. 145: It would be good if you mentioned here the range of resulting source areas considered.
L. 153: What do you mean by “non-natural vegetation” here?
L. 154: How many from bogs?
L. 156: The same problem of attribution applies to the SPECIAL-EPD. Please cite and acknowledge the EPD. See https://www.neotomadb.org/data/data-use-and-embargo-policy
L. 167ff: I like the idea, but am skeptical about the predictors used. Rather than using % needleleaf, it would have been better to classify the pollen types according to high mid and low pollen producing plants. Needleleaf trees include the high pollen producing Pines and low producing Larix (or Pseudotsuga). I am not sure elevation is a good predictor when thinking about the past as vegetation belts moved up and down the mountains during the Holocene. I would perhaps rather limit the inclusion of modern and fossil sites to below 500 m. I don’t understand the need of including the Shannon Index, particularly I don’t understand the provided motivation.
L. 233ff: We know that % tree pollen is a strong predictor of forest cover without any transformation so it would be useful to compare the model performance to the performance of a simple regression model of % tree and shrub pollen (depending on what is in the shrubs) versus forest cover.
L. 233: The negative correlation between %needleleaf and tree cover is interesting and unexpected. Could that be due to frequent Pine pollen in generally open areas. Picea pollen should however correlate with high tree cover.
L. 258: The overestimation of tree cover in northern Scandinavia is interesting and expected as pollen productivity is lower. This is also the case for higher elevations, which is why elevation is a good covariable for the present, but this relationship may not hold true in the past where temperature changes resulted in changing pollen productivities in the mountains.
L. 330: The difference in tree cover between the reconstructions for the last 1000 years and the early Holocene is intriguing. As Zanon et al (2018) and Serge et al (2023) use completely different methodologies, but show the same trend, my initial response would be to trust them more, even if the absolute modern cover is off for both. Here it would be interesting to explore the reasons for the deviations of the current study. Could one reason be the separation of shrub pollen from tree pollen?
L. 341: Please see the recent manuscript by Schild et al. (https://essd.copernicus.org/preprints/essd-2023-486/#discussion ) who argue that the REVEALS method underestimates the forest cover. If that would be true then your new method would perform worse as it scores below the REVEALS estimates. If you argue that forest cover was generally lower then it would be useful to find supporting evidence and make that a point of discussion.
L. 421: The main deforestation of Northwestern Europe took place during the Bronze Age and Medieval period leading to an all-time low around 1800 (see e.g. Bradshaw and Sykes 2014 Ecosystem Dynamics, Wiley).Citation: https://doi.org/10.5194/egusphere-2024-1523-RC1
Data sets
European forest cover during the Holocene reconstructed from pollen records: Species classification schema, updates to SMPDSv1 and site level reconstructions L. Sweeney and S. P. Harrison https://doi.org/10.5281/zenodo.11220915
Model code and software
European forest cover during the Holocene reconstructed from pollen records: R code L. Sweeney https://doi.org/10.5281/zenodo.11220915
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 184 | 36 | 17 | 237 | 19 | 13 | 10 |
- HTML: 184
- PDF: 36
- XML: 17
- Total: 237
- Supplement: 19
- BibTeX: 13
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1