The mid-Holocene hydroclimates and the forcing mechanisms over arid Central Asia (ACA) are hotly debated in the context of global climate change. It is widely assumed that ACA Holocene precipitation broadly followed and/or was out-of-phase with Northern Hemisphere solar insolation. However, here we show a broadly antiphase relationship between Holocene boreal solar insolation and ACA hydroclimatic trend revealed from a well-dated peat core (at the Big Black peatland; BBP) in northwestern China, southern Altai Mountains. Multiple proxies, including peat development rate, pollen assemblages, and peat cellulose isotopic records, show wet conditions during the early and late Holocene, but drought condition during the mid-Holocene. This hydroclimatic pattern is similar to those extracted from other peatlands nearby and those inferred from sedimentary records in lakes in adjacent regions. The trend of δ18O in BBP peat cellulose is similar to that of a stalagmite in northern Xinjiang, both of which record the Holocene atmospheric precipitation δ18O trend over ACA areas and possibly suggest a changing proportion of glacier meltwater supply. We speculate that the mid-Holocene drought over ACA could be ascribed to: (1) the northward movement of the westerlies, such that when the westerlies moved northward under warm conditions, less water vapor was transported to ACA, and vice versa, and (2) increased evaporation under mid-Holocene warm conditions. The data from this study and the potential mechanisms suggest that drier conditions are expected over ACA areas under a continuous global warming expectation.

Mid-Holocene climatic changes are crucial to understanding the causes of the recent/modern global warming and to predicting future climate change (e.g., Steig, 1999; Wanner et al., 2008). It is generally deemed that the Holocene hydroclimatic changes over arid Central Asia (ACA) followed boreal solar insolation (Cheng et al., 2012), similar to those over the Asian monsoon areas (Dykoski et al., 2005). Another major viewpoint is that Holocene effective moisture over the ACA is out of phase with that over monsoonal Asia and is out of phase with the boreal solar insolation due to variable amounts and transport of water vapor modulated by North Atlantic sea-surface temperatures and high-latitude air temperatures (Chen et al., 2008). The mid-Holocene temperature increases (driven by high boreal summer insolation) over ACA areas would be larger than those of the global average because arid Asian drylands are expected to be more sensitive in comparison with the global average. For example, the observed surface warming over global drylands (1.2–1.3 °C) over the past century has been 20%–40% higher than that over humid lands (0.8–1.0 °C; Huang et al., 2017). Huang et al. (2017) even estimated that a 2 °C increase in global temperature may lead to an approximate 3.2–4 °C increase over drylands. Potential evaporation over ACA areas is generally several to ten times higher than precipitation (e.g., Ran et al., 2015). It is likely that the mid-Holocene could have been much drier over ACA areas due to the stronger evaporation driven by higher summer temperatures at that time, which would be quite different from the aforementioned hydroclimatic trends. Therefore, additional evidence is necessary to understand the mid-Holocene hydroclimatic changes and their causes over ACA.

In this study, we discuss the hydroclimatic changes and possible forcing mechanisms during the Holocene over ACA based on multiproxy indices extracted from a well-dated peat core collected at the Big Black Peatland (BBP; Fig. 1), in the southern Altai Mountains, northwestern China. We show a broadly antiphase relationship between Holocene ACA hydroclimate and boreal solar insolation and propose that the south/north movement of the westerlies and changes in evaporation could have dominated mid-Holocene hydroclimatic conditions over ACA.

A peat core (BBP-13) was collected in an intermountain basin (Item DR1 in the GSA Data Repository1) and was dated by 14C dating (Item DR2). Multiproxy indices (Fig. 2), including sedimentation rate, pollen assemblages (Item DR3), and stable oxygen and carbon isotopes (Item DR4), were determined. The 14C ages of peat cellulose indicate that the BBP peat formed at approximately 9500 yr B.P. (Item DR2), and the peat deposition rate remained relatively high from 9500 to 7600 yr B.P. The peat deposition rate began to decrease between 7600 and 6300 yr B.P. and remained very low during the interval of 6300–2800 yr B.P. The BBP peat deposition rate increased after 2800 yr B.P. and remained high during the late Holocene. Peat deposition rates at Chaiwobu (∼580 km away from the BBP peatland; Hong et al., 2014) were also high during the early and late Holocene, but very low during the mid-Holocene (see curve j in Fig. 2), similar in pattern to that of the BBP peatland.

The trends of Holocene climate changes in northern Xinjiang, and even over the wider ACA region, are quite variable between the results of different studies. For example, Chen et al. (2008) proposed an out-of-phase relationship between ACA effective moisture and the Asian summer monsoon intensity on millennial/centennial scales based on comparison of a set of lake sediment records. However, similar trends between ACA precipitation and Asian monsoon precipitation have been inferred from the δ18O records of stalagmites. For example, the Holocene long-term precipitation trend inferred from the stalagmite δ18O record in Keshang Cave, northern Xinjiang, is similar to the Holocene monsoon precipitation trends inferred from stalagmite δ18O records in the Asian monsoon areas (such as Dongge Cave in southwestern China; Fig. 2). A third view is that ACA precipitation or effective precipitation gradually increased throughout the Holocene. For example, the Holocene precipitation trend in northern Xinjiang indicated by peat δ13C (Hong et al., 2014) shows a gradual increase. The water level of Wulungu Lake, northern Xinjiang, indicated by sedimentary grain size (Fig. 3; Liu et al., 2008), and that inferred from ostracod assemblages (Mischke and Zhang, 2011) show a broadly increasing trend during the Holocene. Sedimentary pollen records and other proxy indices from this lake also point to an increasing trend of effective moisture (Liu et al., 2008). Broadly increasing trends in effective moisture during the Holocene existed in several other lakes over northern Xinjiang (Fig. 3), like Lake Boston (Mischke and Wünnemann, 2006), Lake Sayram (Jiang et al., 2013), Lake Aibi (Wang and Feng, 2013), Lake Balikun (Tao et al., 2010), and Lake Tuolekule (Ran et al., 2015). The synthesized effective moisture over northern Xinjiang area also shows a broadly increasing trend (Wang and Feng, 2013). The effective moisture inferred from magnetic indicators in several loess profiles also exhibits a broadly increasing trend during the Holocene (Fig. 3; Chen et al., 2016).

The BBP accumulation rate and pollen assemblages (Fig. 2; Item DR3) indicate relatively wet climatic conditions during the early Holocene and much wetter conditions during the late Holocene, but drier during the mid-Holocene over northern Xinjiang. BBP peat cellulose δ13C shows a broadly similar effective moisture trend to that inferred from the Artemisia and Chenopodiaceae pollen ratio (A/C ratio; Fig. 2). This hydroclimatic pattern is similar to those over adjacent sites. For example, the Narenxia peat pollen records show wet conditions during ∼11,500– 7000 yr B.P., dry conditions during ∼7000– 4000 yr B.P., and wet conditions again after 4000 yr B.P. (Fig. 3; Feng et al., 2016). Pollen A/C ratios were much higher during the early Holocene than the mid-Holocene in a profile in Yili valley (Li et al., 2011), also indicating much drier mid-Holocene hydroclimatic conditions. BBP pollen A/C ratios are also similar in trend to those at Lake Aibi (Wang and Feng, 2013); both point to wet conditions during the early and late Holocene but dry conditions during the mid-Holocene. The A/C ratios in Lake Balikun (Tao et al., 2010), Lake Tuolekule (Ran et al., 2015), Lake Akkol, and Lake Grusha (see locations in Fig. 1; see references in Table DR1) indicate much drier mid-Holocene conditions as compared with those during the late Holocene.

The water vapor sources over ACA areas are presently hotly debated. One of the main views is that Holocene water vapor in the North Xinjiang region was mainly transported by the Asian summer monsoon (especially during the Holocene optimum). For example, Harrison et al. (1996) suggested that the Asian summer monsoon in the Holocene optimum might have gone further northwestward than the modern position. Water vapor at Lake Boston, Lake Hoton-Nur, Lake Issyk-Kul, etc. (see locations and references in Figure 1 and Table DR1), has also been suggested to have an Asian summer monsoon source during the early and mid-Holocene. The second major view is that the Asian summer monsoon has not crossed the desert/sandy area southeast of ACA; therefore, the ACA water vapor could have been transported by westerlies from the North Atlantic, Mediterranean Sea, etc., and could have been partly supplied by the recycling of Asian inland water (Aizen et al., 2006; Liu et al., 2008; Chen et al., 2008). In particular, when reaching the Altai Mountain areas in northern Xinjiang, the water vapor is largely precipitated due to the influence of topography, leading to high annual precipitation around the Altai Mountains as compared with that in surrounding areas.

Precipitation at Tianshan Mountain reconstructed from tree-ring records is significantly correlated with the North Atlantic oscillation index (Zhang et al., 2015), which implies that precipitation in the northern Xinjiang area may be transported by the westerlies. The modern rainfall δ18O observations also suggest water vapor transported by westerlies but not by Asian summer monsoon (Item DR4). In addition, the hydroclimatic pattern in northern Xinjiang reconstructed in this study is similar to that in the Alps mountain region in Europe (Fig. 3; Wirth et al., 2013), implying that even on longer time scales, the westerlies may regulate the hydroclimatic features in the regions along its route.

In northern Xinjiang, snowfall occurs mainly in winter and early spring, and the snow δ18O value (about –20‰ to ∼–30‰) is much lower than rainfall δ18O values in summer (e.g., Wang et al., 2016). The water vapor recycling recharged by snowmelt in warm seasons may lead to negative δ18O values in rainfall water. The ratio of recycling water in Urumqi, Xinjiang, is ∼16%, as simulated by Wang et al. (2016). Aizen et al. (2006) suggested that the local water vapor recycling might account for a larger proportion of the regional precipitation. It is likely that on long-term time scales (e.g., multidecadal, centennial, and millennial scales), the ratio of snowmelt recycling supplied by the glaciers could be larger during warm intervals, and the regional precipitation δ18O could be more negative; alternatively, during the cold periods, the ratio of snowmelt recycling could be smaller, and the regional precipitation δ18O values would be larger (less negative). Therefore, we propose that the negative atmospheric precipitation δ18O values during the early to mid-Holocene over ACA areas, as inferred from both the BBP peat cellulose δ18O values and the Keshang Cave stalagmite δ18O values (Fig. 2), may be related to the relatively higher proportion of glacier meltwater recycling during this period. During the late Holocene, both the BBP peat cellulose δ18O and the Keshang Cave stalagmite δ18O values gradually increased (Fig. 2), implying that the proportion of glacier meltwater involved in the recycling gradually decreased, which is consistent with the observed shrinking of the glacier cover in the ACA region.

During the winter and early spring, the study area is ice-covered, and precipitation (including snow) would be stored in the catchment because evaporation is very low (Item DR1) due to the low temperature and high snow/ice albedo. When temperature increases in the mid-late spring, snow cover and the lakes/wetlands melt around the study area, and at the same time evaporation increases sharply (Item DR1). Annual evaporation is much higher than precipitation (E >> P) over the ACA areas (Item DR1); the precipitation during the cold seasons is lower than the evaporation of the next spring seasons. Therefore, precipitation during the cold seasons could have no or minor influence on the summer hydroclimatic conditions over the study areas. What mainly influences the summer hydroclimatic conditions are summer water vapor transportation, glacier meltwater (if there are any glaciers in a specific catchment), and evaporation.

The variations of water vapor transport in ACA are related to both the intensity of the westerlies and the location of the westerly belt. Generally, the intensity of westerly wind is relatively strong in winter, and the westerly belt is shifted southward in winter compared to that in summer (Folland et al., 2009). This suggests that during longer cold periods, the westerly intensity may be increased, and the westerly zone could move south, resulting in more water vapor transported from the North Atlantic, Mediterranean, Caspian Sea, etc., to the ACA region. On the contrary, the flux of water vapor to the ACA region would decrease during the relatively warm periods. Solar activity/insolation, as an external heat source for Earth, may directly or indirectly affect the regional temperature and the location of the westerlies. Stronger (weaker) solar activity/insolation corresponds to higher (lower) surface temperature, and northward (southward) movement of the westerlies, resulting in less (more) water supply to ACA regions. Therefore, the hydroclimate over ACA areas is expected to act in a “warm/dry–cold/wet” pattern, which is also supported by a large number of paleoclimate records. For example, recent work shows that it was much wetter during the cold periods over the eastern ACA areas during the past 4000 yr (Lan et al., 2018). Therefore, we propose that the dry and wet changes in ACA may be partly ascribed to the variation of water vapor supply due to the north-south movement of the westerly belt, where the high mid-Holocene boreal summer insolation (Berger, 1978) may have led to high boreal summer temperatures and northward migration of the westerly belt, and eventually led to a decrease in the water vapor flux to the ACA region.

The high mid-Holocene boreal summer temperature (Fig. 3) may also have led to increases in summer surface evaporation. Although the low boreal mid-Holocene winter insolation (Berger, 1978) may have resulted in low winter temperatures, this would not have significantly influenced the annual evaporation because the winter evaporation is very small compared with the annual value (Item DR1). Since annual evaporation is much higher than annual precipitation over ACA areas, the increases in evaporation (ΔE) could also be much higher than changes in annual precipitation (ΔP), which may also partly contribute to the observed mid-Holocene drought over ACA areas.

Our multiproxy records show wetter early and late Holocene and drier mid-Holocene conditions in the Altai area of northern Xinjiang, which is similar in pattern to those recorded in adjacent peatlands, lakes, and loess profiles. We contend that the mid-Holocene drought in the northern Xinjiang region, and even in ACA, could be ascribed to the northward advance of the westerly belt and the increased evaporation caused by higher summer temperatures during this period. Providing these mechanisms stand, the ACA areas are expected to be even drier under a continuous global warming scenario.

We thank the editor and anonymous reviewers for valuable comments and suggestions. This work was funded by National Natural Science Foundation of China (NSFC) research programs no. 41761144070 and no. 41672169, and by the External Cooperation Program of the Bureau of International Cooperation, Chinese Academy of Sciences (no. 132B61KYSB20130003).

1GSA Data Repository item 2019092, details of modern climates, sampling, dating, analysis, and climatic significance of the multi- proxy indices, is available online at, or on request from
Gold Open Access: This paper is published under the terms of the CC-BY license.