Modern aridity in the Altai-Sayan mountain range derived from multiple millennial proxies

1500-year stable carbon and oxygen isotopes in larch tree-ring cellulose

The δ¹³C_cell (Fig. 1a, Fig. S2) and δ¹⁸O_cell (Fig. 1b, Fig. S3) records span 516–2016 CE, at annual resolution. The δ¹³C_cell timeseries shows mostly increasing trends during the first millennium of the Common Era (516–1120 CE), and similarly at the end of the last millennium (1720–2016 CE). The maximum δ¹³C_cell value occurs in 2016 CE (−19.6‰; + 3.2σ), while the minimum occurs in 686 CE (−24.7‰, −3.6σ) relative to the average for the period 516–2016 CE (−22.04‰) (Table S2, Fig. S2). The standard error (SE) for the whole analysed period is 0.02.

The δ¹⁸O_cell timeseries (Fig. 1b, Fig. S3) showed two positive and one negative extreme over the past 1500 years, with the minimum value (19.9‰; −6.3σ), occurring in 536 CE, and maximum values (31.9‰; + 3.8σ and 32.2‰; + 4.4σ), occurring in 1266 and 2008 CE, respectively (Table S2, Fig. S3). The SE for the whole analysed period is 0.03. The δ¹⁸O_cell data has higher standard deviation (SD) (1.15) than δ¹³C_cell (0.75).

Less than 1% of values in the δ¹⁸O_cell record are classified as extreme, with the standard deviation ≥  ± 3σ. The δ¹³C_cell and δ¹⁸O_cell records are significantly correlated (r = 0.1, p = 0.0001, n = 1500).

Local climate signals preserved in δ¹³C_cell and δ¹⁸O_cell records

We used weather observations from the local Mugur-Aksy weather station (50°N, 90°E, 1850 m asl) (Table S1) to derive quantitative paleoclimatic reconstructions from our δ¹³C_cell and δ¹⁸O_cell timeseries. A multiple linear regression analysis revealed significant correlations between δ¹³C_cell and July precipitation (r = −0.58; p < 0.0001) (Fig. S4a, Table S4), and δ¹⁸O_cell and July air temperature (r = 0.64; p < 0.0001) (Fig. S4b, Table S5), for the period from 1966 to 2015. These relationships allows us to infer July precipitation from δ¹³C_cell (r = 0.49, F = 14.79, df = 1.47, p = 0.0003), and July air temperature from δ¹⁸O_cell (r = 0.61, F = 29.76, df = 1.50, p = 0.0002). The δ¹³C_cell is also significantly (but negatively) correlated with the temperature of September in the previous year (r = −0.38; p < 0.05). The δ¹⁸O_cell is significantly positively correlated with the same variable (r = 0.36, p < 0.05), which may indicate that warm and dry climate conditions cause prolongation of vegetation season. The drought index (DRI) averaged over May–July (r = −0.52; p < 0.05) for the period from 1966 to 2016 CE is also significantly correlated with δ¹³C_cell (Fig. S4).

Local and regional precipitation reconstructions

The local July precipitation reconstruction derived from δ¹³C_cell only suggests relatively low precipitation (i.e., dry conditions) during the tenth to eleventh, thirteenth, nineteenth, and twenty-first centuries, with relatively high precipitation (i.e., wet conditions) during the seventh, twelfth, fifteenth to sixteenth, eighteenth centuries (Fig. 2a). Our reconstruction also shows pronounced decreasing July precipitation trends from the sixth to twelfth and the eighteenth to twenty-first centuries.

Precipitation extremes (dry < −2.5σ, wet >  + 2.5σ) were calculated across all individual years (Table S6) and for the historical period (Table S7). Recent dry events are superimposed on a pronounced downward precipitation trend (2000–2016 CE) (Fig. 2a, Table S6). The maximum July precipitation occurred in the seventh century, with July precipitation totals reaching double the long-term average (85.3 ± 28.8 mm) of the past 1500 years, with similarly wet years occurring during the LIA (Table S6).

We also calculated a regional multi-proxy ASMR JJA precipitation reconstruction (Table S3), based on our new δ¹³C_cell record combined with Co/Inc and Rb/Sr values from the Teletskoe Lake sediment core²⁶, where each predictor explains the following amount of variance: δ¹³C_cell 43% (Fig. S5a), Co/Inc 33% and Rb/Sr 24% (Fig. S6). The mean correlation coefficient of all possible pairs of proxy datasets is 0.78. Statistical relationships between regional ASMR reconstructed JJA precipitation versus observed JJA precipitation data from the Barnaul weather station over the common period 1930–2009 CE are significant at annual resolution (r = 0.51, r² = 0.25, F = 16.11, df = 1.47, p = 0.0002, standard error of estimate 0.38) as well as for the smoothed by a 10-year average (r = 0.79; p = 0.0001). Correlations remain also stable and significant over calibration and verification periods (Fig. S5a). The first-order difference between the observed JJA precipitation from the Barnaul weather station data and reconstructed JJA precipitation were computed and passed the significance test at p = 0.00019; r = 0.46, F-criteria = 20.8, df = 1.77 (Fig. S5c). Therefore, the two series have good consistency in high-frequency changes and can be proved to be reliable.

In the ASMR JJA regional precipitation reconstruction, strong negative precipitation anomalies occur during the ninth to eleventh, eighteenth to twenty-first centuries (Fig. 2b). Comparable dry intervals occurred during both the MWP and the modern period, albeit with a more pronounced signature in the modern period. The regional reconstruction reveals an unprecedented decreasing trend in ASMR JJA precipitation by almost 49% towards the twenty-first century (1966–2016 CE) compared to the preceding millennium (535–1965 CE).

Local and regional air temperature reconstructions

The July local air temperature reconstruction derived from δ¹⁸O_cell values reveals one particularly cold year (536 CE), during which July air temperature dropped down to 10.9 °C (− 6 σ) compared to the long-term reconstructed average of 14.9 °C for the period 516–2016 CE (Fig. 3a, Table S6).

Reconstructed July air temperature reached a maximum of + 15.8 °C during the MWP (800–1300 CE), which is comparable with prevailing temperatures during the recent period (Table S7). Despite single-year anomalies in the pre-industrial that are comparable to those in the modern period (Table S6), centennial-scale trends and variability demonstrate that the warming at the beginning of twenty-first century is both more pronounced and more persistent compared to the past 1500 years. The JJA-air temperature reconstructions from other temperature-sensitive proxy records in the region (TLs, MXD and TRW, Fig. 3b)^13,15,16,27 show high synchronicity in their low-frequency variability over the period 529 to 2007 CE. Similarly, our new local July air temperature reconstruction derived from δ¹⁸O_cell shows similarities with tree-ring and lake sediment proxies when smoothed by a 101-year Hamming window. However, the shorter March-November temperature reconstruction from the Belukha ice core glacier (Fig. 3b), is less similar, and this highlights either spatial heterogeneity in local temperature, or a discrepancy in the climatic signal preserved by these different proxy archives during early first millennia and towards the third one (Fig. 3b).

In the ASMR JJA-air temperature reconstruction based on the multiple proxies, the variability is explained by the geochemical elements of Ca by 8%, Br/Sr by 8%, MXD by 26%, δ¹⁸O_cell by 6%, and TRW by 41%. Statistical relationships between observed JJA air temperature from the Barnaul weather station data and reconstructed ASMR JJA regional air temperature are significant at annual resolution (r = 0.56, r² = 0.31, F-criteria = 17.85, df = 1.40, p = 0.0003, standard error of estimate 0.71) and for the smoothed (by a 10-year window) reconstruction during the period for which we have observations i.e., 1838–2006 CE (r = 0.73, p = 0.0001, Fig. S5b). For both calibration (1970–2009, r = 0.75, p = 0.0001) and verification (1930–1969, r = 0.79, p = 0.0001) periods (Figure S5b) the correlations are significant with the standard error for reconstructed data (SE ± 0.41). The first-order difference between the observed JJA air temperature from the Barnaul weather station data and reconstructed JJA air temperature were computed and passed the significance test at p = 0.00000; r = 0.51, F-criteria = 59.28, df = 1.16 (Fig. S5d). Therefore, the two series have good consistency in high-frequency changes and can be proved to be reliable.

The LIA (1400–1900 CE) is strongly marked in the ASMR region in the local δ¹⁸O summer temperature regional reconstruction (Fig. 3a,b). The extremely cold summer of the year of 536 CE is captured by annually resolved δ¹⁸O_cell, showing the lowest value over the past 1500 years in comparison with other stratospheric volcanic eruptions 1259 CE, 1641 CE, 1815 CE, 1991 CE. This signal is also represented in the regional JJA temperature reconstruction (Fig. 3c), however, with a less pronounced reduction of summer air temperature.

According to our δ¹⁸O_cell-based local July air temperature reconstruction, July 2008 CE was the warmest month in the reconstructed period (+ 15.9 °C), with similarly high temperatures occurring only during the MWP. Our new multi-proxy regional JJA temperature reconstruction reveals unprecedented air temperature increase towards the twenty-first century. This can be compared with the observed modern global temperature increase relative to the last millennia, which has a similar shape, but less strong increase² (Fig. 3c).

Our new local and regional summer temperature and precipitation reconstructions demonstrate that modern dryness in the ASMR is unprecedented in the context of the past 1500 years. This leads to higher risk of forest fires, evidence for which is recorded by the ASMR paleo archives^6,26–28.

Although some single-year temperature extremes (Table S6) were observed in the past of a similar magnitude to the modern period, the frequency of these events was not as high as that of recent drought. Recent aridity was reported for eastern China based on instrumental observations and reconstructed rainfall²⁹, which is in line with our finding for the ASMR region. High drought frequency is also observed in Mongolia³⁰ and captured in Palmer Drought Severity Index (PDSI) reconstruction for Europe³¹ and in the Swiss Alps³², again suggesting that the drought conditions over the recent decades in many parts of Eurasia are unusual in a long-term context.

The September air temperature of the previous hydrological year is negatively correlated with δ¹³C_cell, but positively correlated with δ¹⁸O_cell. This relationship is in line with observations during unusually warm and dry conditions during recent decades; such conditions may limit carbon assimilation and therefore late-season conditions in the previous year are important for accumulating carbon as starch for storage¹⁹. A hydrologic link between δ¹⁸O in precipitation and temperature (Figure S4) is well known¹⁸ and therefore can also be expected for tree-ring δ¹⁸O in cellulose as it was reported for Chinese monsoon region²⁹. For δ¹³C_cell, an influence of precipitation is expected via the stomatal conductance, which is known to respond to dry conditions¹⁹.

In the multi-proxy regional ASMR summer precipitation reconstruction, the dry LIA period is most pronounced during 1820–1870 CE. This is in line with paleo- precipitation reconstructions from European sites, e.g., high-elevated sites at the Swiss Alps, where LIA was recorded by the stable carbon isotope proxy indicating a drought³². However, LIA is heterogeneously recorded in hydroclimatic conditions world-wide³³ and specifically for Asian regions, where monsoons affect climate patterns significantly^29,34,35.

Our local July temperature reconstruction indicates that the coldest year of the past 1500 years occurred in 536 CE (−6σ), and that this summer has no analogues over this past period (Fig. 3a). This corresponds with a major “unknown” volcanic eruption that occurred most likely in 536 CE in the tropics³⁶, which is confirmed not only regionally^10,11,17 but also globally⁸ and possibly resulted in the Northern Hemisphere LALIA¹⁰.

Our ASMR regional JJA-temperature reconstruction indicates that the average temperature during the recent period (2000–2016 CE) is 2 °C higher than the average for the preceding 1490 years, which is in line with the global-mean temperature reconstruction¹. Maximum temperature anomalies showed up to 4 °C summer air temperature increase for the single years during the period 2000–2016 CE. Moreover, the ASMR JJA-temperature reconstruction shows similar patterns over the recent decades as the PAGES 2k global mean surface temperature (GMST) reconstruction², emphasizing an unusual recent warming trend towards 2000s. However, the MWP and LIA in the regional ASMR JJA-temperature reconstruction show differences in magnitude and duration compared with the same intervals in the PAGES 2K GMST reconstruction²_.. Causes of these differences may be related to external factors, e.g., volcanic eruptions, permafrost and solar variability, and atmospheric patterns³⁷.

Specifically, some of the differences between the GMST reconstruction and our regional ASMR JJA air temperature reconstruction can be explained by seasonal effect of different proxies, similar as it was shown by PAGES2k Consortium². The results make it clear that some paleoclimatic archives are sensitive to July, or averaged June–July air temperature, but not or less sensitive to annual air temperature. Despite that, the majority of tree-ring records included in the PAGES 2K global air temperature reconstruction² are from the Northern Hemisphere, there are some proxies with the low spatial resolution, which are included from both Northern and Southern Hemisphere or proxies, which cannot be precisely calibrated in time. It is well known, that the uncertainties for most composites are due to the paucity of records (combination of corals, marine sediments), which show a cooling trend through most the Common Era². The PAGES 2K Consortium² explained this by the low resolution of marine sediment records and local oceanographic factors over the past millennium, as well as the process of bioturbation of the sediment archive. These factors may suppress the ‘true’ signature of climatic changes occurring over years and decades³⁸, including the most recent warming.

Simultaneous temperature increase and moisture deficit enhance the drought stress of trees to recent climate change, which is expressed by an increased δ¹⁸O_cell signal towards the twenty-first century. The amplifying effect of vapor pressure deficit on drought stress of trees has been shown globally for many species and sites^39,40,41,42.

Discrepancies between high- and low-temporal resolution proxies are unsurprising and can be explained by different integration times, different driving factors and by recording different seasonal variability (at the beginning of the season early June–until late middle/end of August) or mixed temperature and moisture signals (e.g., thawed permafrost water). Specifically, variability in δ¹⁸O_cell does not only represent local temperature variability, but also information about the VPD and evaporation as hydrological changes^40,43.

The main discrepancy in our local δ¹⁸O_cell—derived July temperature reconstruction, compared with earlier published JJA temperature reconstructions inferred from MXD¹⁰ and TRW¹² only, is that it shows a significant temperature increase only over the past few decades. The onset of the recent temperature increase was observed earlier in the TRW and the MXD proxies; this may be because the TRW and MXD proxies contain information about June temperatures, whereas δ¹⁸O_cell is most strongly correlated with July temperature²⁰. Furthermore, VPD increases as temperature increases, which is reflected in the δ¹⁸O_cell values as well. Enhanced evapotranspiration and permafrost thaw depth under elevated CO₂ and temperature increase can impact δ¹⁸O_cell and, therefore, can explain an offset in response to temperature changes towards the third millennia. In contrast, for TRW and MXD summer temperature alone is the dominant driver. The δ¹⁸O mainly captured the July air temperature signal modulated by drastic permafrost degradation over the recent decades. Information about permafrost thaw soil depth is recorded in the δ¹⁸O in tree-ring cellulose only in contrast to other tree-ring proxies, which mainly record only air temperature signal (TRW–JJ, and MXD–JA). These differences in seasonal window can average the summer temperature signal, dampen uncertainties and therefore has advantages over single-parameter reconstructions. Offsets in temperature and precipitation reconstructions between individual chronologies can be explained by the use of different weather stations for the calibration period; one weather station each is at low- (180 m a.s.l.) and high-elevation (1850 m a.s.l.). Therefore, using a combination of tree-ring parameters and other paleoclimatic archives provides valuable information about paleo-temperature change, that cannot be obtained from single-proxy reconstructions alone.

Source: Ecology - nature.com