the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Climate field reconstructions for the North Atlantic region of annual, seasonal and monthly resolution spanning CE 1241–1970
Abstract. The North Atlantic region is a key component of the climate system via large scale atmosphere and ocean circulation. Climate field reconstructions can provide a long-term context for ongoing climate change and contribute to our understanding of climate dynamics, impact of external forcings, and act as references for model evaluation and baseline for natural variability. There are distinct differences in North Atlantic climate variability between the seasons in terms of climate modes and amplitude of the variance. Constraining long-term climate variability in sub-annual resolution is therefore needed for a more complete understanding of the governing processes. In this study we present reconstructed climate in annual, seasonal and seasonal resolution based on a small high-quality network of proxy data combined with output from an isotope enabled climate model. Compared to earlier work we have improved the methodology to obtain better skill across a larger area and more realistic variance of the reconstructed variables which include 2m temperature (T2m), sea surface temperature (SST), sea level pressure (SLP) and precipitation amount. Here we validate the reconstructions against reanalysis data, observed SST and eight long-term records of observed temperature. The reconstructed temperature correlates with up to 0.71 for seasonal and 0.68 for annual data compared to reanalysis data, while the correlation is about 0.3 for monthly resolution. The skill for SLP shows the imprint of large-scale circulation for winter with more local pattern dominating for summer. This is also mirrored in the skill for precipitation. In addition, the reconstructed annual mean SST shows basin-wide skill for the North Atlantic, indicating relevance of the reconstruction to studies of atmosphere-ocean interaction. In summary, the results show the potential of assimilating a small high-quality network of proxy records.
- Preprint
(3993 KB) - Metadata XML
-
Supplement
(7198 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2911', Anonymous Referee #1, 09 Aug 2025
- AC1: 'Reply on RC1', Jesper Sjolte, 17 Oct 2025
-
RC2: 'Comment on egusphere-2025-2911', Anonymous Referee #2, 19 Sep 2025
Summary:
A small number of climate-sensitive proxies are used with a simulation from an isotope-enabled coupled Earth system in a field reconstruction of North Atlantic and European climate. The reconstruction spans annual, seasonal, and monthly timescales depending on the climate variable. The reconstruction is validated against instrumental products, including 20CRv3 for the atmospheric variables, and a suite of SST reconstructions for the ocean. Agreement depends on the climate field and the time period, with the best agreement in locations with abundant high-quality proxies (e.g., summer temperature in regions with latewood density records). Skill at monthly timescales, and especially for precipitation and mean-sea-level pressure, is generally poor.
There are two main issues with this study, concerning insufficient description of the method and weak validation, that need to be cleared before I can recommend publication. I elaborate on these points below and then give minor comments.
Main Points:
1) I've read the method section twice, and I would be unable to reproduce the results in this study. The method uses a weighted-analog approach from a single model simulation. The weights are determined by misfits to the proxy records, but I don't see anything about how this comparison is made when the proxy is not a climate-model variable (e.g., MXD and ring width). This suggests that these records have been inverted for temperature, but perhaps they all have (including d18O)? This is very important as it affects not only the weights, but the relevance of the validation process.
The ad hoc variation inflation method has no justification other than the reconstructed time series has more variance. Lacking any evidence, I assume that this simply adds noise and not signal; i.e., it serves no purpose.
2) The validation of the results is particularly weak. If, as I suspect, that many proxy records have been inverted for temperature, then those inversions have been calibrated on the instrumental record. If true, then the validation is in sample, and doesn't independently measure skill. If I have that wrong, then this should serve as motivation to describe this aspect very clearly.
Given how computationally cheap the reconstruction method is, another important validation approach is to leave out proxies for validation, and repeat the reconstruction process. This can take the form of a jackknife, or leaving out sets of proxies. The reconstruction can then be used to predict the proxies that were left out as a way to independently validate the results, and to test how sensitive they are to the excluded proxies. I would be particularly interested to see results when the Greenland proxies are left out, since Greenland ice core d18O correlates very weakly with European temperature (e.g., Hörhold et al., 2023).
Minor points:
1) You should account for pattern correlation in the significance calculation; I suspect the counts of significant grid cells given in the figures is not far different than random at monthly timescales
2) line 8: As I indicated above the benefit of this variance inflation needs to be shown. In particular, that the inflated variance is signal, not noise.
3) line 54: terms
4) line 83: one simulation only?
5) line 110: reconstructions
6) line 126: What does this mean when the proxy is not a model variable?
7) line 130: time series variance? over what time period? why would this be error?
8) line 163: This sounds problematic. If the “non main” modes of variability were robustly expressed in the proxy data, then larger ensembles should be better.
9) line 191 that that
10) line 193: coeval?
11) line 220: which is
12) line 270-271: I do not understand this
---
referencesHörhold, M., Münch, T., Weißbach, S. et al. Modern temperatures in central–north Greenland warmest in past millennium. Nature 613, 503–507 (2023). https://doi.org/10.1038/s41586-022-05517-z
Citation: https://doi.org/10.5194/egusphere-2025-2911-RC2 - AC2: 'Reply on RC2', Jesper Sjolte, 17 Oct 2025
Data sets
Climate field reconstructions for the North Atlantic region of annual, seasonal and monthly resolution spanning CE 1241-1970 Jesper Sjolte and Qin Tao https://zenodo.org/records/15746008
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,131 | 51 | 25 | 1,207 | 29 | 44 | 48 |
- HTML: 1,131
- PDF: 51
- XML: 25
- Total: 1,207
- Supplement: 29
- BibTeX: 44
- EndNote: 48
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of the study "Climate field reconstructions for the North Atlantic region of
annual, seasonal and monthly resolution spanning CE 1241-1970" by Jesper Sjolte and Qin Tao
This study presents new climate field reconstructions for the North Atlantic region spanning from 1241 to 1970 CE at annual, seasonal, and monthly resolutions, using a small network of proxy data combined with isotope-enabled climate model simulations. The authors reconstructed four key climate variables: 2-meter temperature, sea surface temperature, sea level pressure, and precipitation. Validation against reanalysis data and long-term temperature observations shows strong correlations of up to 0.7 for seasonal and annual temperature data, though monthly resolution correlations were weak. The reconstructions successfully captured large-scale winter circulation patterns in sea level pressure and demonstrated basin-wide skill for North Atlantic sea surface temperatures, providing valuable long-term climate context for understanding natural variability.
Scientific Significance:
This manuscript makes a meaningful contribution to paleoclimate reconstruction by extending North Atlantic climate records back to 1241 CE with multi-resolution temporal coverage. While the geographic focus and general reconstruction approach build on established methods, the improved methodology, proxy data and sub-annual resolution data represent valuable advances.
Scientific Quality:
The scientific approach appears methodologically sound. The validation strategy is comprehensive, employing multiple independent datasets including reanalysis data, observed SST, and instrumental temperature records. However, at least for the monthly reconstruction I would suggest a validation with instrumental data sets for the 20th century.
Presentation Quality:
While the abstract provides a clear overview and the introduction is very well written, some parts of the main paper could be improved. For instance, proxy selection criteria could be better explained. Results and discussion could be separated better, e.g. ModE-RA results only appear in the discussion instead of the results.
Specific Comments:
1. I'm really skeptical about the monthly reconstruction in this study. ModE-RA winter temperatures at least in Europe should be well constrained by historical information and early instrumental measurements. How would Fig. 7 look for 20CR? Maybe that would be a figure for the supplement. Please also check the monthly reconstruction with gridded instrumental data sets for the 20th century. Maybe rather remove the monthly reconstruction from the paper and "monthly" from the title or discuss it more carefully.
2. How can you deal with 1 year dating uncertainty (line 80)? And how may wrongly dated proxies influence your entire study and data set? Did you consider comparing neighboring proxies with different lags to check for potential dating problems?
3. Why do you "use JJA for representing the growing season and the extended winter season (Nov-Apr) to represent the winter preceding the growing season"? In this setup May does not play any role. Why not also an extended summer growing season starting in May?
4. With this analog method you disturb the temporal evolution of the model simulations. Are the transitions from one year to the next year smooth in the ocean with its memory/autocorrelation? Especially in the monthly reconstruction I would imagine that this could be an issue.
5. In the results you ask the question "how many model analogues to use in the ensemble reconstruction". But in the methods section above you wrote that you "calculated the ensemble mean using a logarithmic weighting function". I understood that all members are included. Maybe, I just misunderstood something but please explain this more clearly.
6. Fig. 2: the JJA SLP reconstruction has negative correlations if you just move slightly out of the region covered by the proxy data. What could be the reason and would you consider shrinking the reconstruction region to the area covered by proxy data?
Small remarks:
Abstract:
Line 6: "seasonal and seasonal" should probably be "seasonal and monthly"
Introduction:
Line 19: "... data and while ..."should be without "and"
Line 27: "the main mode of wither variability" should be "winter variability"
Data Section:
Line 110: "reconstrcutions" should be "reconstructions"
Line 176: "emsemble members" should be "ensemble members"
Unclear English:
Introduction:
Line 51-52: "Well-dated, long-term seasonally resolved proxy data" maybe better "well-dated, seasonally resolved, long-term proxy data"?
Line 65: "creates biases the representation" should be "creates biases in the representation"?
Results Section:
Line 193: "This is coeval with results" - "coeval" is uncommon; "consistent with" would be clearer
Line 208: "skill of the reconstructing" should be "skill of the reconstruction"
Discussion:
Line 248: "Compared to the SEA18/20 reconstructions our new reconstruction" missing comma after "reconstructions"
Line 267-269: "as shown in SEA18/20 even a low number" missing comma after "SEA18/20"