the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Dec 2025
Strong volcanic imprints contrast with a mild Little Ice Age: a first temperature reconstruction based on maximum latewood density from the Caucasus
Abstract. The Caucasus occupies a unique climatic region influenced by European, Mediterranean, and Asian circulation systems, yet it remains underrepresented in tree ring-based Northern Hemisphere temperature proxy networks. Here, we present the first summer temperature reconstruction for the Caucasus region based on maximum latewood density (MXD). We used X-ray micro-computed tomography of tree-ring samples from Pinus sylvestris growing at the upper tree line in the Lesser Caucasus and an ensemble nested regression approach to develop a robust 326-year-long June-September temperature reconstruction (1697–2022). The record explains – regionally unprecedented – 72 % of temperature variance during the instrumental period (1901–2022) and captures distinct interannual and multi-decadal variability including pronounced warming since the 1990s and a strong imprint of major volcanic eruptions. Temperatures in the 18th and 19th century, a period often described as the Little Ice Age, were not significantly colder in the Caucasus than in the first half of the 20th century. The reconstruction highlights the exceptional magnitude and persistence of 21st century warming in the region, which is without analogue at least in the past three centuries. Comparison with regional and large-scale temperature reconstructions reveals strong agreement within the Caucasus but negative correlations with Central Europe, indicating distinct temperature variability patterns across Europe and western Asia. Future work should focus on the climate dynamics behind this dipole and the extension of temperature-sensitive tree-ring records in the region.
- Preprint
(1415 KB) - Metadata XML
-
Supplement
(1000 KB) - BibTeX
- EndNote
Status: open (until 28 Feb 2026)
- CC1: 'Déjà vu in dendrochronology: from New to Old World', Samuli Helama, 23 Dec 2025 reply
-
RC1: 'Comment on egusphere-2025-5809', Anonymous Referee #1, 05 Jan 2026
reply
Comments on scope and interpretation of the Little Ice Age signal
Dear authors,
I find the study to be technically strong and regionally important. The MXD data are of high quality, the temperature signal is exceptionally robust, and the reconstruction clearly documents volcanic impacts and the magnitude of recent warming. These aspects are not in question.
However, I largely agree with the concerns raised in the accompanying comment regarding the interpretation of Little Ice Age (LIA) conditions. In its current form, the manuscript appears to argue for a “mild LIA” in the Caucasus, while the available data and applied standardisation do not allow this hypothesis to be tested in a rigorous way.
There are three closely linked issues that, in my view, need to be addressed explicitly.
First, temporal coverage. The reconstruction begins in the late 17th century, i.e. well within the commonly defined LIA period. As a consequence, the data do not include pre-LIA conditions (13th–14th centuries) against which the magnitude or onset of LIA cooling could be evaluated. The reconstruction can therefore only describe temperature variability during the late LIA and the subsequent transition into the 20th century, but it cannot assess whether the LIA as a whole was mild or severe relative to earlier climate states.
Second, standardisation and low-frequency variability. The use of series-wise spline detrending, even within a signal-free framework, inherently limits the preservation of multi-decadal to centennial-scale variability. This is a well-known consequence of spline-based standardisation and implies that slow, long-term cooling trends characteristic of the LIA are likely to be attenuated or removed prior to chronology development. As a result, the reconstruction is structurally much more sensitive to rapid, high-amplitude signals (e.g. volcanic cooling, 20th-century warming) than to gradual long-term changes. In this context, the absence of pronounced low-frequency cooling cannot be taken as evidence for a mild LIA.
As noted by the comment by Samuli Helema, a further prerequisite for reliably assessing centennial-scale variability using Regional Curve Standardisation (RCS) is that the regional curve represents a well-balanced distribution of different age classes. I would like to add that this condition alone is not sufficient. A robust regional curve should also be based on material that spans a sufficiently wide range of climatic conditions.
If a chronology is constructed primarily from living trees that established and grew under broadly similar climatic regimes, the resulting regional curve may be biased towards those conditions. In such cases, even an RCS-based approach may have limited ability to distinguish long-term climatic shifts from biological growth trends. This further emphasises that, for datasets lacking both pre-LIA material and strong climatic contrast across the lifespan of the trees, inferences about centennial-scale variability remain inherently constrained, irrespective of the chosen standardisation method.
Third, spatial representativeness. The conclusions are drawn from a single regional record, whereas the concept of the Little Ice Age is based primarily on large-scale, multi-site reconstructions with broad spatial coverage and millennium-long temporal extent. Comparisons between a temporally and spatially restricted regional chronology and continental-scale European networks must therefore be made with caution, particularly on centennial timescales.
Taken together, these points do not undermine the quality or value of the reconstruction. Rather, they suggest that the interpretation of LIA conditions should be more tightly constrained to what the data can robustly support. In my view, the manuscript would be strengthened by reframing the conclusions to emphasise temperature variability during the late pre-industrial period and the exceptional nature of modern warming, while avoiding broader claims about the severity or mildness of the Little Ice Age as a whole.
I encourage the authors to clarify these limitations explicitly and to adjust the wording of the discussion and conclusions accordingly. Doing so would bring the interpretations into closer alignment with the methodological scope of the study and would substantially strengthen the paper.
Specific examples of formulations that should be revised or constrained (with line numbers):
Abstract, lines ~24–26
“Temperatures in the 18th and 19th century, a period often described as the Little Ice Age, were not significantly colder in the Caucasus than in the first half of the 20th century.”
Overgeneralises the LIA despite limited temporal coverage. Should be explicitly restricted to the late LIA and to the period covered by the reconstruction.
Introduction p. 3, l. 73-83.
The hypothesis is formulated in terms of improved reliability at both high and low frequencies and a “better confined temperature history” at local to regional scales. While this is a reasonable motivation for applying MXD, it would be helpful to clarify that the ability to assess low-frequency variability is temporally constrained by the start of the chronology. As the record begins in 1697, inferences regarding centennial-scale trends are necessarily limited to the late part of the pre-industrial period and cannot be extended to earlier phases of the Little Ice Age. Stating this explicitly here would help align the stated aims with the temporal scope of the data.
Discussion 4.1, lines ~306–307
“The minimal differences between mean raw and detrended MXD chronologies give us additional confidence that the developed chronology preserves both high- and low-frequency climate signals.”
Raw vs detrended similarity is not sufficient evidence for preservation of centennial-scale variability. This should be removed or substantially weakened.
Discussion 4.1, lines P18~387–394
The statement that the weak LIA signal is “likely the result of actual regional climate differences and not a methodological deficit” is not sufficiently supported given the temporal coverage and the applied series-wise spline standardisation. Methodological constraints affecting the retention of low-frequency variability cannot be excluded and should be explicitly acknowledged. I therefore recommend rephrasing this passage to allow for both climatic and methodological explanations, rather than excluding the latter.
Should be explicitly qualified by the start of the record within the LIA and the limitations of spline-based standardisation.
Excludes plausible methodological explanations. Should be rephrased to acknowledge both climatic and methodological factors.
Citation: https://doi.org/10.5194/egusphere-2025-5809-RC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 341 | 113 | 30 | 484 | 41 | 17 | 19 |
- HTML: 341
- PDF: 113
- XML: 30
- Total: 484
- Supplement: 41
- BibTeX: 17
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Déjà vu in dendrochronology: from New to Old World
Tree rings provide excellent proxy data about past climates. Tree-ring data benefit from annual to sub-annual resolution and exact dating to calendar years without uncertainties, meaning that they can be directly compared with meteorological records. In their contribution, Dhyani et al. (2005), referred hereafter as DH25, present tree-ring based summer temperature reconstruction for the Caucasus region. Tree-ring samples were collected from an open-canopy Pinus sylvestris trees growing on dry, south-facing slopes in a subalpine environment with rocky outcrops at the upper treeline (2200–2300 m above sea level) in the western Lesser Caucasus. X-ray micro-computed tomography was used to produce maximum latewood density data. Based on their statistical analyses, a robust 326-year-long June-September temperature reconstruction was developed over the period of past three centuries (AD 1697-2022). Over the instrumental period (AD 1901–2022), the reconstruction explains more than seventy percent of temperature variance. As the main palaeoclimatic results, DH show pronounced warming since the 1990s and a strong imprint of major volcanic eruptions. Contrast to the commonly accepted view of preindustrial climate variability, DH25 claim that temperatures during the Little Ice Age (LIA) were not significantly colder in the Caucasus than in the first half of the 20th century. In this commentary, the latter issue will be, according to the title of the original contribution, termed as the mild-LIA hypothesis.
While several statistical figures are supportive of the reconstruction, there exists a need to reconsider the mild-LIA hypothesis. The hypothesis is noteworthy, since the LIA literature is historically focussed on European conditions, and any aspects of climate variability from new regions predating the 20th century are highly relevant to palaeoclimatic research. However, such hypothesis should be based on credible information.
In order to stimulate further deliberation, it is essential to consider tree rings as biological archives. Tree-ring data may reflect climate variability across scales, but when doing so they also portray trends that are related to trees’ biological age and unrelated to climate, which is why the series need to be standardized (detrended/indexed) prior to any palaeoclimatic interpretations. The importance of standardization cannot be underestimated in any dendrochronological study, nor can its significance for this field of science, in general. According to H. C. Fritts (1976): “Standardization is such a basic procedure in dendrochronology that it is considered by some to be a principle.” The idea could perhaps be continued by another idea, that hardly no one would think of standardization as a scientific discipline, if there was only one (and always correct) way to standardize the data.
Probably one of the first dendrochronologists to face the problem of standardization in terms of long-term climate variability was LaMarche (1974). He studied tree rings of bristlecone pines at the upper treeline in White Mountains, California, in order to draw comparisons with other climatic records and link anomalies in tree-ring data to circulation patters. Topical to this commentary, an interesting note on his study was later made by Fritts (1976) citing the dendrochronological analyses and tree-ring standardization trials of LaMarche (1974) and more frankly acknowledging the situation: “LaMarche recognized that if he used the standard indexing procedures, a large portion of the averaged and lagged response would be removed.” In the words of LaMarche: “if each component series is first transformed to an index series with a mean of 1.0, then as component length decreases the averaged annual values will approach unity. Thus, any evidence of trends over period greater than the length of the component series would be subdued or eliminated.“ Instead of standardizing his tree-ring series, LaMarche (1974) simply excluded the ring-width measurements near the centres of trees showing rapid growth due to young age rather than climate. Intriguingly, one of the findings based on his undetrended tree-ring data now capable of showing the long-term (low-frequency) climatic variations and trends was the cool climate prevailing during the suspected LIA interval (AD 1430-1850).
The way LaMarche (1974) excluded rings near the pith before averaging the chronology is not, in any case, possible for most of the dendrochronological data that may show age-related trends over the full segment. Instead, the series need to be standardized. In another example, this was done by Briffa et al. (1990) using “smoothing splines with a frequency-response cutoff set at two-thirds of the length of each series”. As a result, their summer temperature reconstructions of northern Fennoscandia, based on maximum densities in living tree and subfossil Pinus sylvestris tree rings, showed little evidence for the existence of any Medieval Warm Epoch and for Little Ice Age shortly between AD 1570 and 1650. The limitation of their methodology was, as Briffa et al. (1990) themselves stated, that the standardization procedure precluded “any possibility of reconstructing climate variability on timescales much of 300 years”. The extent of the limitation was estimated from the mean segment length that was 290 years. All this is very topical to DH25, who also use spline functions for standardization. The spline functions do, however, differ between the two studies.
For the sake of clarity, the three sentences describing the tree-ring standardization method used to produce the final chronology and thus the summer temperature reconstruction for the Caucasus region, DH25 can be cited as follows: “The age-dependent splines had an initial stiffness of 50 years (50% variance cut off at the wavelength of 50 years) and progressively increasing stiffness with cambial age (Cook and Peters, 1981; Melvin et al., 2007). To further enhance the capacity of the chronology to retain long-term signals, the detrending was performed within the signal-free framework (Melvin and Briffa, 2008) using the ssf() function in dplR (Bunn, 2008). This approach iteratively removes the common signal prior to detrending, allowing the detrending splines to be more accurately estimated without being influenced by strong climate signals (Melvin et al., 2007).”
This means that the limitation of preserving the long-term climate variability may be even stronger for DH25 than for Briffa et al. (1990), for it seems to be the maximum segment length that reaches ~300 years in the Caucasus, not the mean length as in the case for Briffa et al. (1990). Another reason is the spline rigidity used by DH25, starting with 50-years, not approximating 200-years as in the case of Fennoscandian Pinus sylvestris data which Briffa et al. (1990) used. That is, the long-term variations over especially the early part of the series, that could potentially overlap with the LIA, are effectively removed from the DH25 chronology. Generally, these examples demonstrate the need to consider with caution any hypothesis of missing long-term climatic trends, when tee-ring data is detrended by flexible curves fitted to each series.
To support the mild-LIA hypothesis, DH25 cite another previously published dendrochronological dataset from the Caucasus. With these regards, DH25 maintain as follows: “the strong agreement with the Greater Caucasus summer temperature reconstruction of Dolgova (2016), which also shows no pronounced multidecadal cooling, supports the view that the weak LIA signal in our record is likely the result of actual regional climate differences and not a methodological deficit.” However, it is quite essential to reread the cited paper to investigate the tree-ring standardization methods exploited in that study. The cited paper provides a short statement of the way the series of healthy old Pinus sylvestris trees were detrended: “Indices of the chronology were calculated as residuals from a 50-year smoothing spline.” That is, the method of standardization Dolgova (2016) applied seems to be rather similar than that of SH25, especially over the earlier centuries that overlap with the suspected LIA conditions. Therefore, it appears likely that similar standardization issues hold for both DH25 and Dolgova (2016). In other words, the question of long-term climate variability remains unresolved due to the tree-ring standardization that was carried out using the splines, and the study of Dolgova (2016) cannot for this reason be cited in support of the mild-LIA hypothesis.
In support of the final chronology, DH25 also state: “The minimal differences between mean raw and detrended MXD chronologies give us additional confidence that the developed chronology preserves both high- and low-frequency climate signals.” Even this argument is not free of doubts, however. If the temperature-sensitive trees covering the early part of the chronology were, contrary to the hypothesis of mild-LIA, growing under relatively cold conditions, then the higher growth values due to young age over this part of the chronology may have been at least partly nullified by such a climatic factor. In such a case, the raw tree-ring series would display only minimal age-related trends, for which reason the raw and detrended mean records may mimic each other better than otherwise expected. Tree-ring series with no age-related trends were recently found by Zhang et al. (2022) studying open-canopy tree-ring data of trees growing at the treelines on the Tibetan Plateau. Age structure to help the interpretations is commonly provided with age trend which, for the Caucasus data seems to indicate a weak decline in latewood density due to ageing, in comparison to Eurasian Pinus sylvestris values shown in Figure 2 of Briffa et al. (2001).
In fact, the foregoing limitations set by the standardization methods were described in detail by Cook et al. (1995). They also coined the term Segment Length Curse to explain the hardships of retaining tree growth variations at timescales exceeding the segment length when standardization curves are fitted to each series. Further, they suggested that the Segment Length Curse may be exorcized by an alternative standardization method, Regional Curve Standardization (RCS). Their statement was likely encouraged by the work of Briffa et al. (1992) who had already used the RCS method for producing a renewed Fennoscandian tree-ring based summer temperature reconstructions, now with improved evidence for the Medieval Warm Period and, topically, for a long cold period from the late sixteenth to the middle eighteenth century hence coinciding with the LIA evidence. As stated by Cook et al. (1995): “The premise behind RCS is that there is a single common age and size-related biological growth curve for a given species and site that can be applied to all series regardless of when the trees were growing. Because this standardization curve is assumed to be an intrinsic, time-invariant feature of tree growth for the species and site under study, RCS allows for the possibility that the overall level of actual tree growth during any particular time period may be systematically over- or underestimated by the regional curve due to changing climatic/environmental conditions. In this sense, RCS allows for long-term change in the mean due to climate, while at the same time removing trend that is believed to be mostly biological in origin.” In the context of LaMarche (1974), the over- and underestimation by the regional curve means that the RCS indices of each tree are no longer locked to a mean of 1.0.
Over the past 30 years the RCS method has been studied more intensively. As a result, the RCS concept shows a diversification with several types of variants and modifications; in any case, the RCS-type methods remain as essential tools for retaining information on low-frequency tree-growth forcing (Helama et al. 2017). Even so, it may be that the tree-ring based LIA temperatures are not fully solved without RCS indices from both living and dead trees found for examples as standing or downed boles from the dry mountainous terrains, as such extension would markedly increase the replication and balance the age structure of the Caucasus chronology over the earlier centuries. In any case, there is no reason to suggest that DH25 data ought to be reevaluated this way, for the reconstructions based on other standardization methods has their own merits. However, the mild-LIA hypothesis, or any other hypothesis of missing long-term climatic trends, is hardly plausible based on tee-ring data detrended by curves fitted to each series.
In addition to the obvious benefits, the RCS method brings its own limitations that need to be also considered. Particularly suitable in the case of tree-ring chronologies built merely using living trees, Klesse and Frank (2013) provide a following guideline: “To achieve an unbiased RCS chronology, a multi-site strategy and the use of pith-offset estimates are of major importance. In addition to the common knowledge that a high sample replication is required for RCS, we found it is also helpful that these samples are distributed among several sites with different age structures. This sampling design allows for the systematic testing of biases. The use of pith-offset estimates greatly increases the accuracy of the RC.” The statement may be a simplification summarizing the troubles but nevertheless serves as an instruction for the fieldwork. Also, these thoughts concur with those of Linderholm et al. (2010) who considered the many aspects of RCS method and, to mitigate problems associated with the method, concluded: “all age-classes should be represented throughout a chronology. If this is not possible, the use of “non-RCS” standardization is recommended, although this method results in a loss of low-frequency variability.”
References
Briffa, K.R., Bartholin, T.S., Eckstein, D., Jones, P.D., Karlén, W., Schweingruber, F.H., and Zetterberg, P.: A 1400-year tree-ring record of summer temperatures in Fennoscandia, Nature 346, 434-439, https://doi.org/10.1038/346434a0, 1990.
Briffa, K.R., Jones, P.D., Bartholin, T.S., Eckstein, D., Schweingruber, F.H., Karlén, W., Zetterberg, P., and Eronen, M.: Fennoscandian summers from AD 500: Temperature changes on short and long timescales, Climate Dynamics, 7, 111-119, https://doi.org/10.1007/BF00211153, 1992.
Briffa, K. R., Osborn, T. J., Schweingruber, F. H., Harris, I. C., Jones, P. D., Shiyatov, S. G., and Vaganov, E. A.: Low-frequency temperature variations from a northern tree ring density network, Journal of Geophysical Research, 106, 2929–2941, https://doi.org/10.1029/2000JD900617, 2001.
Bunn, A. G.: A dendrochronology program library in R (dplR), Dendrochronologia, 26, 115–124, https://doi.org/10.1016/j.dendro.2008.01.002, 2008.
Cook, E.R., and Peters, K.: The smoothing spline: a new approach to standardizing forest interior tree-ring series for dendroclimatic studies, Tree-Ring Bulletin, 41, 45-53, 1981.
Cook, E. R., Briffa, K. R., Meko, D. M., Graybill, D. A., and Funkhouser, G.: The "segment length curse" in long tree-ring chronology development for palaeoclimatic studies, The Holocene, 5, 229-237, https://doi.org/10.1177/095968369500500211, 1995.
Dhyani, R., Martin-Benito, D., Verschuren, K., Matskovsky, V., Van den Bulcke, J., Doğan, M., Kvaratskhelia, R., Köse, N., Güner, H. T., and Schneider, L.: Strong volcanic imprints contrast with a mild Little Ice Age: a first temperature reconstruction based on maximum latewood density from the Caucasus, EGUsphere, https://doi.org/10.5194/egusphere-2025-5809, 2025.
Dolgova,E.: June–September temperature reconstruction in the Northern Caucasus based on blue intensity data, Dendrochronologia, 39, 17-23. https://doi.org/10.1016/j.dendro.2016.03.002, 2016.
Fritts, H.C.: Tree Rings and Climate, Academic Press, 567 pp., 1976.
Helama, S., Melvin, T.M., and Briffa, K.R.: Regional curve standardization: State of the art, The Holocene, 27, 172-177, https://doi.org/10.1177/0959683616652709, 2017.
Klesse, S. and Frank, D.: The application of regional curve standardization on living tree datasets, Tree Rings in Archaeology, Climatology and Ecology, 11, 47-55, https://doi.org/10.2312/GFZ.b103-13058, 2013.
LaMarche, V.C. Jr.: Paleoclimatic Inferences from Long Tree-Ring Records, Science, 183, 1043-1048, https://doi.org/10.1126/science.183.4129.1043, 1974.
Linderholm, H.W., Gunnarson, B. E., and Liu, Y.: Comparing Scots pine tree-ring proxies and detrending methods among sites in Jämtland, west-central Scandinavia, Dendrochronologia, 28, 239-249, https://doi.org/10.1016/j.dendro.2010.01.001, 2010.
Melvin, T. M. and Briffa, K. R.: A “signal-free” approach to dendroclimatic standardisation, Dendrochronologia, 26, 71-86, https://doi.org/10.1016/j.dendro.2007.12.001, 2008.
Melvin, T. M., Briffa, K. R., Nicolussi, K., and Grabner, M.: Time-varying-response smoothing, Dendrochronologia, 25, 65-69, https://doi.org/10.1016/j.dendro.2007.01.004, 2007.
Zheng, J., Lyu, L., and Zhang, Q.: Isolated Trees in Two Tibetan Plateau Treelines Reveal Growth Plasticity to Harsh Conditions of the Little Ice Age, Forests, 13, 1371, https://doi.org/10.3390/f13091371, 2022.