1 Introduction
Understanding climate changes since the last glacial period is a key point in paleoclimate research, especially those impacting the Asian summer monsoon (ASM; including the East Asian summer monsoon (EASM) and Indian summer monsoon (ISM)) climate and westerlies climate, which are two pivotal circulation components of global circulation systems (
Kutzbach, 1981;
Oldfield, 1999;
Chen et al., 2016). The TP stretches over 3.08×10
6 km
2 and has an average elevation of approximately 4320 m above sea level (m a.s.l.) in mid-latitude regions (
Zhang et al., 2002,
2021); Previous researches have revealed that the TP is of great significance for the climate of Asia, as it not only influences the intensity of the EASM and the ISM by altering the land‒sea thermal contrast and the circulation coupling between both the subtropics and tropics and the lower and upper troposphere (
Liang et al., 2006;
Wu et al., 2012;
Tada et al., 2016) but also affects the intensity and location of the westerly jet in the spring (
Li and Liu, 2015). Thus, unique location and configuration enable the TP climatically sensitive and ideal for paleoclimate research (
Zhu et al., 2009;
Cheng et al., 2013).
In recent decades, a number of paleorecords from the TP have been used to reconstruct past climates, thereby improving our understanding of climate change during the Holocene and extends the spatial coverage of environmental records for the TP (
He et al., 2004;
Herzschuh et al., 2006;
Zhu et al., 2015). Additionally, some studies have tried to detect the climate evolution pattern over the TP. However, the complexity in topography and inner terrain elevation differences, as well as the interplay among the dominant circulation systems, i.e., EASM, ISM, and midlatitude westerlies, make it difficult to draw conclusions about the spatial and temporal patterns of paleoclimate evolution (
An et al., 2000,
2006;
Yang et al., 2008). According to modern observation data, moisture flowing to the western TP is relative to the Arabian Sea, as well as the ISM (
Jin, 2006).
Dong et al. (2017) confirmed that ISM low-pressure systems brought large amounts of precipitation from north and central India to the south-western TP. On the northern TP, climate change is controlled by westerlies (
Gao et al., 2014). Recently, based on three gridded precipitation databases over the ASM region from 1979 to 2009 and empirical orthogonal function analysis (EOF),
Conroy and Overpeck (2011) suggested that rainfall seasonality followed a typical monsoon pattern, which received approximately 60% of the annual total between May and October; they then divided the modern climate on the TP region into three areas with distinct precipitation variabilities, i.e., south-east, north-west, and south-west, with the driest conditions in the north-western TP and wetter conditions in the eastern TP. These different areas were likely to have different moisture sources. The precipitation in the eastern subregion was brought by the EASM from the western North Pacific, and that in the western subregion was related to the ISM. However, the precipitation on the north-western TP might be related to continental water recycling.
When considering higher Northern Hemisphere summer insolation during the Holocene (
Berger and Loutre, 1991), the existence of a spatial-temporal pattern of moisture evolution similar to the current situation has yet to gain consensus. For example,
Herzschuh et al. (2006) argued that there were east‒west regional differences in climate evolution and in the timing of optimal conditions, which may have resulted from the availability of the spatially fluctuating effective moisture.
Chen et al. (2008) compared paleolimnological data from across the TP and confirmed that effective moisture and lake levels in monsoon-influenced TP were lower during the late Holocene, while effective moisture and lake levels in mainly westerlies-influenced areas experienced low levels in the early Holocene and high levels during the middle to late Holocene.
Zhang and Mischke (2009) also supported the different spatial patterns and suggested that a possible mechanism may be the insolation-driven southward drawing of westerly winds in the early Holocene. After comparing the climate in the Nam Co area inferred from pollen data with other climates on the TP,
Li et al. (2011) summarized that the temporal succession, i.e., the termination of maximum monsoon precipitation or moisture conditions occurring earlier on the western TP than on the eastern TP, followed a hypothesized south-eastward retreat of monsoon moisture during the middle to late Holocene along a modern precipitation gradient, as concluded from a monthly 1900–2014 global gridded precipitation data set. An et al. (
2012) suggested that climate evolution on the north-eastern TP was controlled by the interplay between the westerlies and ASM during the Holocene.
Zhu et al. (2015) and
Hou et al. (2017) obtained lake sediments on the central TP and affirmed that the climate change during the Holocene was influenced mainly by the ISM. In contrast, general agreement in paleohydrological evolution on the TP during the Holocene was presented by
Mügler et al. (2010) after comparing wet and dry periods inferred from 6 lake records across the TP. In addition,
Wischnewski et al. (2011) applied a five-scale moisture index and average link clustering to semiquantitative moisture information to detect general patterns of moisture evolution and confirmed that there were obvious differences in moisture evolution inferred from paleoclimatic data, but it is difficult to identify coherent temporal and spatial patterns. Apart from the viewpoints outlined above,
Doberschütz et al. (2014) considered that there was temporal variability in the behavior of the moisture evolution from lake sediments, which was nearly synchronous in the middle Holocene and showed a nonuniform response to climate change in the late Holocene.
Considering the considerable discrepancy regarding the climatic significance of the proxies used in lake sediments (
Chen et al., 2016), the nature of Holocene climatic changes in the region could be concealed. Furthermore, few attempts have been made to clarify the circulation configurations on the TP. Consequently, in the present paper, we investigate the spatial and temporal characteristics of Holocene moisture evolution recorded in paleoclimatic records and detect whether the moisture pattern of regional heterogeneity on the TP truly remained over time during the Holocene as in the modern climate, as well as the possible forcing mechanisms. In this respect, paleoclimate modeling and its comparison with proxy records is a very useful strategy. Paleoclimate records provide an observational basis for verifying the precision of the model, which can further be used in developing and tuning numerical models. The models, once the accuracy in corresponding climate factors is confirmed, can be used to explore the dynamics that have driven climate variability in the past. Therefore, we used model data over the past 9.5 ka and compared the simulation results with moisture inferred from sediments to explore the driving factors for moisture changes over the TP during the Holocene.
2 Data and methods
2.1 Kiel Climate Model (KCM) description
The Kiel Climate Model (KCM), a state-of-the-art coupled ocean-sea-ice-atmosphere general circulation model, comprises the atmospheric general circulation model ECHAM5 and the Nucleus for European Modeling of the Ocean (NEMO) ocean-sea ice general circulation model, using OASIS3 as a coupler. The horizontal resolution of ECHAM5 is T31 (3.75° × 3.75°), with 19 vertical levels from 1000 hPa to 10 hPa. The horizontal resolution of NEMO, using the Mercator meshes with grid refinement in the tropical regions, attains an average resolution of 1.3°. More details on the model can be found in
Park et al. (2009).
The transient simulation started at 9.5 ka BP, which was output from an equilibrium simulation forced by Earth’s orbital parameters, including eccentricity, obliquity and precession (
Berger and Loutre, 1991). Thereafter, with a 10-fold acceleration scheme, the simulation was forced by varying the orbital parameters according to the respective period from 9.5 to 0 ka BP (
Lorenz and Lohmann, 2004). Other forcing factors except insolation forcing, such as greenhouse gas concentrations, continental ice sheets and meltwater fluxes, were set at preindustrial (AD 1850) levels. A more detailed description of the model setting can be found in
Jin et al. (2014).
KCM model data have been verified as an effective tool for paleoclimatic research.
Jin et al. (2014) employed equilibrium and transient simulation data to probe spatial-temporal patterns over ASM and monsoon marginal regions and suggested that the synthesis of multiproxy records and model experiments were in accordance. In addition,
Zhang et al. (2016) used KCM model data to investigate how summer North Atlantic sea surface temperatures (SSTs) may have influenced the ISM at centennial timescales during the Holocene (9.5−0 ka BP). Recently,
Zhang et al. (2018) compared KCM transient simulation data with proxy-based precipitation changes on the north-eastern TP and confirmed the uniformity between pollen data and KCM data.
2.2 Paleoclimatic record selection
The objective of this study is to review Holocene moisture evolution and to investigate whether there is an obvious spatial-temporal moisture evolution pattern on the TP. To avoid the disparity of different paleoclimatic records in climate change, we only selected lake sediment and peat bog data to reconstruct the paleoclimate. However, with large discrepancies in sample resolution and age control among paleorecords, the patterns based on record-time-slide comparisons may be distorted. To ensure reliable data quality, we collected published data and selected the data according to criteria similar to previous studies (
Herzschuh, 2006;
Chen et al., 2008;
Wang et al., 2010;
Wang et al., 2017). 1) The record must cover at least 4000 years during the Holocene. In this paper, because we regarded 8 ka and 4 ka as the early/middle Holocene and middle/late Holocene time boundaries, respectively, a record covering at least 4000 is more precise to represent the general situation during the subperiods of the Holocene. 2) The proxies in the record must be indicative of past precipitation/moisture changes. In this study, lake sediment and peat bog data, including geophysical (lithological description, color index, grain size, etc.), geochemical (organic and inorganic carbon content, stable carbon and oxygen isotope ratios, and elemental composition), and pollen data were selected. Pollen data, which have been frequently used in climate reconstructions, are considered a reliable signal for climate changes at a regional scale (
Zhao et al., 2009a,
2009b).
Wang et al. (2010) used Procrustes analysis to evaluate the concordance between pollen and nonpollen and demonstrated the significant concordance between pollen and nonpollen data sets (
Peres-Neto and Jackson, 2001) and synchronous signals between pollen and nonpollen records on centennial timescales, which permits us to use them together for ordination purposes. However, in this study, carbonate oxygen isotope records (δ
18O
carb), which is prevalent in lacustrine research, were not collected. In some cases, lacustrine δ
18O
carb documents different climate evolution with other proxies, such as the δ
18O
carb data and pollen data in Qinghai Lake. According to
Zhang et al. (2018) and
Wu et al. (2022), the isotopic composition of lake water can be influenced by many factors, including temperature, evaporation, site elevation, meltwater supply, and moisture sources. Complicated influencing factors and different climate implications of δ
18O
carb make it difficult to combine with other proxies. 3) The time sequences should be based on reliable chronology, with more than four dating control points during the Holocene. 4) The sampling resolution was sufficient for this research (< 200 years for the Holocene period). All selected sites were first checked to remove any possible old carbon effect. Then, if necessary, all
14C ages were calibrated to determine ages in this paper using the calibration program Calib 8.1.0, with the newest calibration curve (IntCal 20). Calibrated ages were used when compiling moisture curves throughout the paper (expressed as cal. a BP or cal. ka BP, 1 ka = 1000 cal. a BP) (
Chen et al., 2008;
Zhang et al., 2011). 5) The old carbon/reservoir effect was mentioned and removed in the original paper. As research shows, the old carbon/reservoir effect is universal in lakes on the TP.
Hou et al. (2012) demonstrated that the old carbon/reservoir effect exhibited significant spatial and temporal variability on the TP, ranging from almost 7000 years to several years. Therein, the largest reservoir effect was in Bangong Lake (6670 years). In addition,
Mischke et al. (2013) argued that the largest reservoir effect can reach 20000
14C years. To make a valid comparison, the old carbon/reservoir effect must be removed.
To obtain moisture index inference for paleoclimatic records, multiproxy anomalies at an individual lake were first calculated and normalized. According to the normalized results, the signals of moisture evolution were coded on a four-part scale developed by
Herzschuh (2006): dry (0), moderately dry (1), moderately wet (2), and wet (3). A high wetness value was assigned to intervals with fine-grained sediments, high Artemisia/Chenopodiaceae pollen ratios, high organic content and humification, low long-chain leaf wax and so on. Each moisture degree was individualized, and there was no comparability because of their geographic locations, archives, proxies used and sensitivities. Relative moisture evolution sequences indicate how the climate around the lakes/peat evolved so that we could distinguish whether all the records were influenced under the same climate system. Every sequence was conducted for each 100-year interval. However, because there might be several proxies from one site, it was impossible to consistently translate each individual proxy into semiquantitative climate signals at all sites. Therefore, after carefully crosschecking the interpretations from original papers, we largely agreed with the conclusions drawn from the authors when classifying proxy-based climatic inferences into the moisture index. The definition of the paleorecord boundaries was based on the results of REOF using FORTRAN, which overcomes the shortcomings of EOFs in presenting the characteristics of different geographical regions and is widely used in modern climate research to investigate the spatial pattern of climate evolution (
Richman, 1986).
Based on the criteria mentioned above, we analyzed 27 lake sediments and 1 peat bog sediment on the TP. As inferred from the deposits and landforms of glaciers on the TP, climate change during the early Holocene was significantly unstable (
Owen et al., 2005;
Zhang and Mischke, 2009,
2022). For example, a climate reversal during the early Holocene probably represents the widely recognized Younger Dryas, although there are significant differences among records (
Tschudi et al., 2003). Some records cannot capture these climate events, such as those events involving Koucha and Kuhai Lakes (
Mischke et al., 2008,
2010). Evidence from other studies records the Younger Dryas with a time difference (12.5−11.7 ka BP from Lake Naleng and from 11.3 to 10.8 ka BP from Lake Qinghai) (
Shen et al., 2005;
Kramer et al., 2010a). The significant discrepancies in moisture change during the early Holocene make it difficult to perform further analysis. Therefore, to ensure the time uniformity between proxy data and model data, we studied a period spanning 9.5−0 ka BP. Herein, 21 records span the complete period between 9.5 and 0 cal. ka BP, the other 6 sediment records remained incomplete (Fig.1 and Tab.1).
3 Results
3.1 Spatial and temporal characteristics of Holocene moisture evolution on the TP
We divided the Holocene into five periods to investigate the spatial and temporal characteristics of Holocene moisture evolution on the TP. The Holocene moisture histories showed distinct spatial discrepancies (Fig.2). During the early Holocene (9.5−8 ka), the climate was clearly wet on the western, southern, central and south-eastern TP. Meanwhile, most of the sediments on the north-eastern TP experienced moderate or wet conditions. In contrast, No. 19 (Hurleg Lake) and No. 20 (Koucha Lake) recorded dry conditions. In the first half of the mid-Holocene (8−6 ka), the climate recorded in the deposits was significantly asynchronous. Most of the paleoclimatic data on the north-eastern TP recorded humid conditions, but these conditions were drier than before on the western, southern and central TP, as well as according to some paleorecords on the south-eastern TP. In contrast, No.17 (Sugan Lake) on the northern TP and No. 19 experienced dry conditions. During the last half of the mid-Holocene (6−4 ka), there were larger regional variations, with a less spatially coherent pattern compared to the early or first half of the mid-Holocene.
During the first half of the late Holocene (4−2 ka), most records showed a continuous moisture decrease except for those areas located on the south-eastern TP and northern TP and No. 19, where the climate remained under wet or relatively wet conditions. In the last 2 ka, almost all paleorecords on the western, southern, central and south-eastern TP experienced the driest conditions during the Holocene, while some records on the north-eastern and northern TP documented moderate conditions.
Then, we apply REOF, and the cumulative variance contribution of the first four modes reached 82.15%, which means that the spatial patterns of moisture evolution of the first four modes are representative of the main moisture pattern on the TP. As shown in Fig.3, in the leading-order mode, the high positive loading value centers (≥ 0.3) are located in the western-southern-central TP (boundary by ~94°E), and the north-eastern and eastern TP including Nos. 22, 23, 26, 27, and 28 (boundary by ~32°N). A low negative loading value center (≤ −0.3) is located in the north-eastern TP (No. 20). The remaining modes can be performed in the same manner.
According to the results, the moisture evolution pattern recorded in lake sediment and peat bogs on the TP can be classified into 5 subregions. They are regions A (~28°−35°N, ~78°−94°E; western-southern-central TP), B (~28°−32°N, ~94°−102°E; south-eastern TP), C (~32°−38°N, ~98°−103°E; north-eastern TP), D (~37°−39°N, ~93°−98°E; north-eastern-northern TP) and E (~33°−35°N, ~95°−98°E; north-eastern TP), and there are 16, 3, 5, 2, and 1 paleorecords in each subregion, respectively (Fig.4 and Figs. S1−S5). According to the synthesized moisture indices, moisture in Region A declined gradually from the early Holocene. In Region B, moisture remained humid during the early Holocene to early-late Holocene. In Region C, there was a wetting trend from 9.5 to approximately 6 ka BP and a persistent drying trend thereafter, with the wettest interval in the middle Holocene. In Region D, there was a wetting trend from the early Holocene to late Holocene. However, in Region E, the moisture index experienced semiwet conditions during the early Holocene and wet conditions during the middle to late Holocene.
3.2 Model results
The time series of summer precipitation evaluated using model grid points that correspond as closely as possible to subregions are compared with the synthesized moisture indices (Fig.5). The KCM-simulated summer precipitation on the western-southern-central TP, averaged over the region (28°−35°N, 79°−94°E), decreased gradually from the early Holocene to the late Holocene. The model result agreed well with the synthesized moisture index in Region A, with the climatic optimum taking place during the early Holocene (Fig.2(a) and Fig.5(a)). On the south-eastern TP, modeling data (28°−32°N, 94°−105°E) show that the summer precipitation is high between approximately 6 and 9.5 ka. During the late middle to early late Holocene, summer precipitation is stable and remains at a relatively high level, which resembles the synthesized moisture index in Region B, with maximal summer precipitation occurring in the early and middle Holocene and a relatively moist climate during the late Holocene (Fig.5(b)). In addition, summer precipitation on the north-eastern TP (35°−43°N, 98°−105°E) shows a wetting trend from approximately 9.5 to ~6.2 ka BP, and summer precipitation declines thereafter, consistent with the synthesized moisture index in Region C (Fig.5(c)). However, the moisture index in Region D and Region E is inconsistent with the precipitation inferred from nearby model grid points (39°−43°N, 94°−98°E and 32°−35°N, 94°−98°E, respectively) (Fig.5(d) and Fig.5(e)).
4 Discussion
4.1 Moisture evolution pattern on the TP during the Holocene
The first-order mode emphasizes not only the east‒west asymmetry in moisture variation but also the south‒north asymmetry on the eastern TP. The south‒north asymmetry on the TP, where the separatrix is at approximately 33°N, has been reported by
Flohn (1957), who pointed out that the Tibetan Highland, with its rolling, rather flat steppe or barren land, acts as an elevated heat source in summer that results in the temperature gradient and baric gradient overturning south of 35°N. In addition, based on the stable isotopes of precipitation of short station records,
Tian et al. (2001) insisted that the monsoonal moisture boundary lies at ~35°N.
Zhang et al. (2018) confirmed that during the Holocene, the precipitation changes on the north-eastern TP were caused by the interplay between westerlies and the EASM. During the early Holocene, strong mid-latitude westerlies reduced summer precipitation on the north-eastern TP. As a result, the amount of summer precipitation reached its maximum in the mid-Holocene, which was different from paleorecords in other regions over the TP.
In addition, an east‒west asymmetry in the monsoon rainfall on the TP remained throughout different periods.
Conroy and Overpeck (2011) utilized gridded precipitation data from 1979 to 2009 and confirmed that the south-east region on the TP may not respond to the same monsoon subsystem in the same way as the south-west region.
Chen et al. (2015) synthesized the most up-to-date and comprehensive proxy moisture/precipitation records during the past 1000 years in China and showed the opposite phase on the TP in terms of rainfall during the Medieval Climate Anomaly and the Little Ice Age. On the millennial scale, Hudson and Quade (
Hudson and Quade, 2013), who used the area surrounded by high shorelines of early Holocene paleolakes to reconstruct paleorainfall patterns during the early Holocene, also confirmed that paleolake areas expanded by approximately 4-fold in the western plateau compared to approximately twofold expansion in the eastern region.
According to our REOF results, the climate change on the northern and north-eastern TP could be divided into 3 subregions: Region C, Region D, and Region E. The amounts of summer precipitation on the north-eastern TP (Region C) were under the control of the mix of EASM and westerlies during the Holocene, which was verified by
Zhang et al. (2018). In Region D, Hurleg Lake lies in an arid low-lying basin in the Qaidam Basin and out of the area influenced by the ASM in the modern climate; it showed dry and variable conditions from 9.5 to 5.5 ka and relatively wet and stable conditions after 5.5 ka (
Zhao et al., 2007). Additionally, warm and dry conditions with high evaporation/precipitation ratios could be inferred from Sugan Lake from 7.5 to 5.8 ka. Then, the climate became colder with higher lake levels from 5.8 ka on, followed by colder and drier conditions after 3.5 ka BP (
Zhang et al., 2022). Both of these lakes recorded humid climate conditions during the middle Holocene and relatively humid conditions during the late Holocene. However, the KCM model data cannot reflect the moisture change, which might be attributed to the fact that in the mid-latitude westerly dominant areas, the dry conditions in the early Holocene seemed to result mostly from changes in winter rather than summer climate (
Jin et al., 2014). After comparing the moisture evolution index in Region D with the synthesized Holocene mean moisture index in arid Central Asia, namely, a dry early Holocene, a wetter early to mid-Holocene, and a moderately wet late Holocene (
Chen et al., 2008), we infer that this area may have been affected by mainly westerly circulation during the late Holocene under the background of the strengthening westerlies. In Region E, there is only one lake sediment (Koucha Lake). According to the original result, the early Holocene around Koucha Lake was characterized by warmer and drier conditions than today. From approximately 6.6 ka onward, wet and cold conditions dominated in this region, which is contradictory to other paleoclimate records from monsoon-influenced Asia (
Herzschuh et al., 2009). In addition, model data cannot capture this pattern. Therefore, strong regionalization of paleoclimate change might be the only reliable explanation.
Zhao et al. (2007) also attributed the different moisture evolution to the local terrain after comparing Hurleg sediment with other sediment records.
4.2 Possible forcing mechanism responsible for the discrepancy in moisture inferred from proxy data and model data
4.2.1 Moisture evolution discrepancies induced by circulation configurations during the early Holocene
To determine whether the regional heterogeneity of moisture variability recorded in lake/peat systems on the TP is the consequence of the influence of different climate systems similar to the modern climate, we calculated the East Asian summer monsoon index (EASMI) and Indian summer monsoon index (ISMI) based on the KCM model. The ISMI, which is significantly positively correlated (r = 0.98, n = 95) with summer precipitation in nearby Region A (Fig.6(a)) and significantly positively correlated (r = 0.81, n = 95) with the rainfall amount nearby Region B (Fig.6(b)), was defined by boreal summer (JJA) meridional wind anomalies at 850 hPa and 200 hPa over the ISM region (10°−30°N, 71.25°−108.75°E), which confirmed that the ISM was the main control factor for the moisture change in Region A and influenced the moisture change in Region B. The EASMI, which is negatively correlated (r = − 0.62) with summer precipitation in Region B (Fig.6(c)), was determined by the average boreal summer (JJA) horizontal wind speed at 850 hPa over this area (27.83°−31.54°N, 97.5°−105°E), which suggests that the climate was mainly controlled by the interplay between the EASM and westerlies and the ISM.
Therefore, we conclude that under the different circulation configurations and superposing the terrain influence, the lake/peat records showed a spatial-temporal moisture evolution during the Holocene.
As discussed above, different circulation configurations resulted in regional moisture discrepancies. Here, we analyze the climatologic mean during the Holocene using KCM model data to understand the general circulation pattern and sea level pressure (SLP) pattern (Fig.7(a) and Fig.7(b)). Westerlies prevailed at 500 hPa in most regions of the TP, similar to the modern climate. At 850 hPa, westerlies still controlled the climate on the northern TP. In addition, strong anticyclonic activity over the Pacific transports water vapor from the Pacific Ocean (EASM circulation) to the eastern TP. On the south-eastern TP, wind vectors from the ISM, which brought water vapor from the Indian Ocean, met the wet EASM. In addition to the influence of westerlies on the western-southern-central TP, ISM circulation dominated.
Comparing the early (9.5−8.0 ka)/middle Holocene (8.0−6.0 ka) with the middle/late Holocene (2.0−0 ka), the circulation and SLP patterns are similar (Fig.7(c)−Fig.7(f)). The high-pressure center is on the north-western Pacific, which indicates that the EASM circulation declined gradually from the early Holocene to late Holocene. Meanwhile, water vapor brought forth by a strengthening ISM from the Bay of Bengal and Arabian Sea advanced toward the western, southern, and south-eastern TP, which also indicates a decreasing trend in the ISM circulation. As a result, moisture change inferred from paleoclimatic records in Region A, where the ISM dominated, went through relatively wet conditions in the early Holocene and declined thereafter for the gradually decreasing trend of ISM. In Region B, ISM, EASM, and westerlies influenced the moisture change. Additionally, moisture evolution in Region C and Region E was caused by the interplay between the EASM circulation and the mid-latitude westerlies. In Region D, the climate was under the control of the westerlies. Therefore, the moisture evolution inferred from paleorecords and simulated precipitation was significant and under the control of different circulation and SLP patterns during the Holocene.
However, there are questions regarding why moisture evolution on the TP was asynchronous and why the south-eastern TP remained steady and smooth when the ASM circulations were stronger during the early Holocene than during the middle Holocene, as inferred from paleorecords and simulated precipitation. An alternative explanation, excluding the terrain, may be the natural difference between ISM precipitation and EASM precipitation. ISM precipitation is directly influenced by the intertropical convergence zone, associated with tropical heating and ultimately related to solar radiation (
Chao, 2000;
Fleitmann et al., 2007). As a result, even though the precipitation in Region A was also influenced by westerlies during the Holocene, the moisture evolution still tracked the ISM circulation. In contrast, EASM precipitation was mainly generated by a frontal system, which formed once the warm and moist air flows met the cold and dry flows (
Ding and Chan, 2005). During the early Holocene, the high latitude forcing, i.e., high northern latitude ice volume, weakened land surface temperature and zonal sea‒land thermal contrast, and eventually weakened EASM precipitation. This external forcing is accepted by many studies (
Li and Xu, 2016;
Lu et al., 2019). However, our simulated data did not involve such an external forcing. Therefore, internal feedback is an additional mechanism for consideration. After removing the insolation forcing during the Holocene, the simulated summer precipitation was negatively associated with 850-hPa zonal winds over central-eastern Asia and positively associated with these factors in northern China and the adjacent region (Fig.8(a)). In addition, simulated summer precipitation on the south-eastern TP was negatively correlated with 850-hPa meridional winds over western central Asia and the north-western Pacific and positively correlated with that over East Asia during the Holocene (Fig.8(b)). This correlation pattern was similar to that on the north-eastern TP, which resembled a wave-like anomaly pattern in the middle latitudes (
Zhang et al., 2018), as shown by the composite difference in the summer wind fields at 850 hPa and 500 hPa between the middle and late Holocene (Fig.7(e) and Fig.7(f)). This wave-like pattern was beneficial for the southward displacement of cold air and the westward transport of warm water vapor from the north-western Pacific. As a result, anomalous precipitation was generated over the south-eastern TP after this frontal system formation. However, no significant ridge or trough was induced during the early Holocene compared with the middle Holocene (Fig.7(c) and Fig.7(d)). This circulation pattern reduced frontal rainfall on the south-eastern TP, which was affected by the interplay between the EASM and westerlies during the early Holocene, despite the enhanced westward transport of warm water vapor from the north-western Pacific, as well as the strong EASM circulation.
Coupling the weakening in ISM circulation and the strengthening in EASM precipitation on the south-eastern TP from the early Holocene to middle Holocene, the moisture evolution on the TP was asynchronous, and on the south-eastern TP, it remained steady and smooth, as inferred from paleorecords and simulated precipitation. In general, with different circulation configurations, especially during the early Holocene and middle Holocene, the moisture evolution inferred from paleorecords and model data showed a spatial discrepancy.
4.2.2 Discrepancies in moisture evolution induced by ENSO during the late Holocene
After experiencing relatively weak EASM precipitation during the early Holocene, EASM precipitation reached its maximum at ~6 ka BP and declined thereafter. Meanwhile, the ISM precipitation also decreased gradually in response to the insolation change. However, there has still been a nonuniform climate evolution inferred from paleorecords and simulated precipitation during the late Holocene.
Apart from the increasing intensity of anthropogenic impact during the late Holocene (
Zhang et al., 2018), here, we argue that El Niño/Southern Oscillation (ENSO) variability in the equatorial eastern Pacific changes may affect the spatial and temporal variability in precipitation on the TP during the late Holocene. The ENSO index calculated using wintertime (DJF) SST anomalies in Niño 3.4 and averaged every century from KCM data are quite consistent with the Holocene ENSO frequency (Fig.9(a) and Fig.9(b)) inferred from Laguna Pallcacocha sediment color changes (
Moy et al., 2002). The variability, which was relatively weak during the early Holocene and strong during the late Holocene, is also similar to other modeling data. Here, we utilized a composite analysis in the simulated JJA precipitation and 850-hPa wind vectors (m/s) (arrows) between high Niño 3.4 SST years and the climatological mean during the late Holocene (4−0 ka).
When the tropical Pacific Ocean is in a state of El Niño conditions (high Niño 3.4 SST years), the amount of precipitation on the western-southern-central TP decreases with the wind anomalies compared to the climatological mean during the late Holocene. However, regional differences in precipitation are obvious, where the increasing area is on the eastern TP and its adjacent region (Fig.10).
In fact, the regional differences in summer precipitation in the East Asian region during the Holocene induced by ENSO variability have been mentioned by many published studies (
Jin et al., 2014;
Lu et al., 2019). However, because of the complex circulation configurations on the TP discussed above, the precipitation patterns associated with different circulation configurations are more complicated on the TP. First, ENSO-related SST changes can affect the intensity of the western North Pacific (WNP) subtropical high, which is closely related to the EASM strength as well as the position of the EASM rain belt. When El Niño conditions are dominant, with complementary heating in the eastern North Pacific and cooling in the WNP, the WNP subtropical high is enhanced. Therefore, under the background of the enhancement and westward expansion of the WNP subtropical high, more water vapor is transported from the South China Sea toward the middle and lower Yangtze River Valley along the north-west flank of the WNP subtropical high. The enhanced WNP subtropical high during the late Holocene is also supported by
Wang et al. (2015), who calculated the position of the West Ridge Point (WRP) using KCM data and suggested that there is a westward displacement of the WNP subtropical high during the late Holocene. Consequently, precipitation on the eastern TP has increased. Meanwhile, the anomalously strong WNP subtropical high accompanied by low-level south-westerly airflow has weakened the normal south-east monsoon wind over central and northern China, resulting in less precipitation over semiarid northern China (
Yang and Lau, 2004;
Lu et al., 2019). As a result, precipitation, inferred from paleorecords and simulation data, has become wetter on the eastern TP in response to the El Niño conditions of the late Holocene. These anti-phased responses to ENSO in the EASM area at the millennial-centennial time scales are also similar to those on interannual change and in the precession scale EASM variability from a long-term (284 ka) transient simulation using the fully coupled fast ocean-atmosphere model (FOAM) (
Shi et al., 2012).
Second, the millennial-scale ISM tracks the changes in the zonal structure of tropical Pacific SSTs that resemble the spatial patterns of ENSO (
Bird et al., 2014), which are also roughly consistent with the modern ISM response to ENSO. However, the influence of ENSO on hydroclimatic variations in the ISM area is also known to be temporally stable (Fig.11), which reflects the complicated nature of the monsoon dynamics in response to the variability in ENSO on different time scales (
Wilson et al., 2010;
Xu et al., 2013). Given the anti-phase pattern of summer precipitation in the EASM area under ENSO conditions and the unstable relationship between ISM and ENSO, combined with strong variability during the late Holocene, it is reasonable that the ENSO variability contributes to the spatial-temporal variability in precipitation on the TP.
5 Conclusions
After the critical assessment of the original data and semiquantitative reconstruction, 27 paleoclimate proxy records covering the past 9500 years were selected for detecting the moisture evolution patterns in the last 9.5 ka on the TP. The REOF analysis mainly revealed that the moisture evolution patterns over the TP can be divided into 5 subregions, which are summarized as the south‒north asymmetry and east‒west asymmetry moisture evolutions.
After comparing the synthesized moisture indices for each subregion with the simulated nearby summer precipitation from KCM data, we confirmed that KCM can be used to detect the possible forcing mechanisms for the regional discrepancies in moisture evolution, and under the different circulation configurations and the superposing influence of terrain, the moisture evolution experienced spatial discrepancies, especially during the early Holocene and middle Holocene. In addition, given the anti-phase in the EASM area under ENSO conditions and the unstable relationship between ISM and ENSO, it is reasonable that the relatively strong ENSO variability during the late Holocene contributed to the spatial-temporal pattern of moisture evolution on the TP.
PDF
(35749KB)
