中国西南地区近千年湖泊水位变化

张旭龙，宋云平，张贵林，徐 海

天津大学 地球系统科学学院，表层地球系统科学研究院，天津 300072

Variations in lake levels reflect the water mass balance between precipitation and evaporation in a catchment area and can provide first-order hydrological information, as well as related climate information such as precipitation (or effective precipitation) (Xue, 2001; Goldsmith et al, 2017; Xu et al, 2019; Xu et al, 2020). Lake level variations have an important influence on human life, industrial,and agricultural activities, because high levels may threaten life and property security, while low levels may affect ecological, environmental, and economic sustainability. However, the long-term history and causal mechanisms on these variations are poorly understood, partly due to few and/or short instrumental records. Therefore, reconstructing lake level variations and understanding the corresponding laws and mechanisms are particularly important for understanding both climatic mechanisms and changes in regional water availability.

The methods used to reconstruct past lake level variations can be broadly divided into three categories.The first is to reconstruct lake levels using beach evidence collected from paleoshorelines, terraces,and stratigraphic profiles. By dating beach samples using optically stimulated luminescence, U-Th, and14C, the timing and elevation of historical highstands can be constrained. A number of lake histories have been reconstructed using this approach, including Lake Dali in Inner Mongolia (Goldsmith et al, 2017),Lake Ngangla Ring Tso (Hudson et al, 2015) and Lake Qinghai (Liu et al, 2015) over the Qinghai-Tibetan Plateau, Lake Chenghai in southwestern China (Xu et al, 2020), and the Aral Sea in central Asia (Reinhardt et al, 2008), etc. The advantage of this approach is that the lake level is a first-order record of hydrological change and thus has clear hydroclimate significance.However, this method is limited in applicability to a small number of lakes because of the scarcity of wellpreserved beach evidence and dating material. The second category uses proxy indicators in lake sediment cores, such as sediment grain size, pollen records, total organic carbon (TOC) and carbon/nitrogen ratio (C/N),etc. Examples where these methods were employed include Lake Melincué in the pampas of Argentina(Guerra et al, 2015), Lake Nhauhache in Mozambique(Holmgren et al, 2012), Lake Ahung Co on the Tibetan Plateau (Morrill et al, 2006), Argentina Lake Laguna Potrok Aike in Argentina (Haberzettl et al, 2005) and Lake Titicaca on the Peru-Bolivian border (Rowe et al, 2003), etc. High-resolution and continuous information can be obtained by these methods,but sedimentary proxies are affected by multiple processes, which makes quantification difficult and can result in considerable uncertainties (especially when large lake level drops/rising occurred). The third category involves obtaining historical lake level information from the literature and/or archeological evidence. For example, the inscriptions on the “Taihu Shuizebei”, a stone tablet that describes variations in the level of Lake Taihu, Wujiang, China, recorded four major floods in the early and middle 20th century and 25 large floods between 1194 and 1920 AD (Yu et al, 2013). Three intervals of frequent water level fluctuations in the Dead Sea have been inferred from biblical records, namely 2000 — 1500, 1500 — 1200,and 1000 — 700 BC (Frumkin and Elitzur, 2002).However, this method is also obviously limited due to a lack of historical records.

A large number of lakes exist in southwest China and account for ~1.5% of the total lake area in China(Wang and Dou, 1998; Xu et al, 2011). These lakes are mainly distributed along the fault zone of central and western Yunnan Province, and most are tectonic rift lakes (e.g., Fuxian, Lugu, Yangzong and Chenghai lakes). These lakes are generally characterized by steep basins, deep water, and a relatively small catchment to lake area ratio, resulting in a high sensitivity of the lake level to regional climate/environmental change (Liu, 2015; Xu et al, 2019; Xu et al, 2020). Although several studies have discussed lake level variations in southwestern China, the majority used sedimentary proxies to indicate water depth, such as Yamzhog Yumco in Tibet (Guo et al,2016), and Lake Xingyun (Hillman et al, 2018) and Lake Chenghai (Ye et al, 2018) in Yunnan Province.An integrated approach, especially utilizing beach evidence and historical records, is therefore crucial for reconstructing lake level changes in this area. In this study, we report lake level changes over the last millennium in southwest China by integrating lake level data from previous studies and information inferred from historical records. We then compare lake level changes in southwest China with those over the extended ISM area and discuss the potential forcing mechanisms that drive these changes.

1 Background and methods

Southwest China is located in subtropical Southeast Asia and has distinct wet and dry seasons(Fig. 1a). Precipitation is mainly controlled by the ISM, with more than 80% of the annual precipitation occurring from May to October (Fig. 1b; Wang and Dou, 1998). Modern instrumental records show that precipitation in most areas of Yunnan Province decreased slightly over the past several decades (Fig. 2).Precipitation in most other areas in southwestern China, such as Guizhou, Chongqing and Sichuan provinces, also showed decreasing trends (Compilation Committee of the Southwest China Climate Change Assessment Report, 2013). The lakes in this study are closed or semi-closed lakes and the lake water is mainly supplied by precipitation and related surface runoff, including Lake Yangzong, Lake Chenghai,Lake Fuxian, Lake Xingyun, Lake Qilu, and Lake Yilong (Fig. 1a; Tab. 1). (Fig. 1a; Tab. 1).

Historical literature has clear time control and therefore can provide important historical lake level information (O’Hara, 1993; Frumkin and Elitzur,2002). We compiled historical drought and flood disaster records around six lakes in southwest China(Yangzong, Chenghai, Xingyun, Fuxian, Yilong and Qilu; Tab. 1 and Tab. 2) from literature such asYunnan provincial chorography(Compilation Committee of Yunnan Provincial Chorography, 1998b),Yunnan chorography(Liu, 1991),Wanli Yunnan general chorography(Li, 2011) andYiliang County chorography(Compilation Committee of Yiliang County Chorography, 1998), etc. We use red and blue symbols to indicate relatively high and low lake level status, respectively (Fig. 3). The relative lake level variations were then compared with lake levels reconstructed from beach evidence and/or sediment core proxy records (Fig. 3; Tab. 2). As the historical literature records do not contain absolute elevation information, high and low lake levels inferred from the records of each lake can only indicate a relative amplitude. However, as beach evidence-based lake level reconstructions are very limited and literatureinferred lake level information is available over the entire study area, the combination of these two approaches is crucial to constrain both the timing and amplitude of historical change in regional water availability.

2 Results

2.1 Southwest China lake level variations in historical records

The literature records of potential lake level changes in southwestern China are shown in Tab. 2.Historical drought and flood disaster records mainly fall within the Ming and Qing dynasties (see the time span of each dynasty in Fig. 3), suggesting that water availability was generally high in this area from 1400 to 1800 AD (see details in Tab. 2 and Fig. 3).

Tab. 1 Overview of the lakes investigated in this study

Tab. 2 Lake level fluctuations in southwestern China inferred from historical records

Fig. 1 Locations of the studied lakes in southwest China and other lakes in the Indian summer monsoon region(a),Average seasonal precipitation and temperature in Yunnan Province(data from the China Meteorological Data Service Center; http://data.cma.cn/en) (b)

Lake Yangzong:Lake Yangzong is a freshwater lake with a modern basin area of ~192 km2and a lake area of ~31 km2(Wang and Dou, 1998). Recent lake levels have been decreasing due to an imbalance between water supply, evaporation and water usage(Wang and Dou, 1998). It was named “Big Lake”before the Ming Dynasty (1368 — 1644 AD; Zheng,2006), possibly suggesting it used to be a larger lake. The lake was once closed and there used to be a small outlet less than one chi (1 chi = 1/3 m) in the northeast (Compilation Committee of Yiliang County Chorography, 1998). In 1396, an official named Muchun led the building of the Tangchi Channel(length of 18 km, width of ~4 km) for irrigation.However, when official Wen Heng inspected the channel in 1562, parts of it were no longer useable and had to be rebuilt, indicating that the water level of Lake Yangzong decreased during this period (Compilation Committee of Yiliang County Chorography, 1998). In 1687, floods in Yiliang County damaged the channel bank, suggesting that the lake level in Lake Yangzong rose at this time (Compilation Committee of Yiliang County Chorography, 1998).

Lake Chenghai:According to theNew Yunnan provincial chorography, fishermen accidentally found a stone tablet submerged in Lake Chenghai, called the“Chenghai Ancient Tablet”. Based on the inscriptions on the stone tablet, it was assumed that the stone tablet was constructed during the Dali Kingdom (937 —1254 AD) before the Yuan Dynasty (1271 — 1368 AD;Niu, 2007), which implies that Lake Chenghai was low before the Yuan Dynasty, and Lake Chenghai “became as a sea” during the Yuan Dynasty (Niu, 2007).Combined with the records in theYongsheng County chorographythat translates to “before the middle of the Ming dynasty, water flowed from Lake Chenghai into the Jinsha River”, it can be inferred that before the middle of the Ming dynasty (see time span in Fig. 3),the water level of Lake Chenghai was relatively high and outflowed (Compilation Committee of Yongsheng County Chorography, 1989; Wang and Dou, 1998). The Haizha dam was built at the south spillway during the Wanli Empire (1573 — 1620),although its exact construction dates are not known.The dam was opened in spring and closed in summer to maintain a water supply for irrigation, which suggests that the lake level could have decreased during this interval. However, it was still a semiclosed lake, so the water levels were likely relatively high at that time (Ji, 2014; Niu, 2007). However,statements in theYunnan general chorography, such as “leading the public to dredge the southern estuary of Lake Chenghai” and “later filled up, the dam was also long ruined” (1662 AD), imply that lake level declines could have occurred during the interval from the Wanli Empire to the early Kangxi Empire (~ 1573 to ~ 1662), although the exact timing and magnitude cannot be constrained. Our previous lake level curve,which was based on beach evidence, did not capture the details in short-term water level variations during this period because the beach evidence from several recent centuries was seriously damaged and/or masked(Xu et al, 2020). Using historical literature records,we have modified our previous curve and recorded a decrease during this interval (see dotted line in Fig. 3).

Fig. 2 Observed lake level changes at Lake Chenghai (Ji, 2014), Lake Fuxian, and Lake Yangzong (He et al, 2019); temperature,wind speed, and precipitation at Kunming and Lijiang stations (data from the China Meteorological Data Service Center,http://data.cma.cn/en) and simulated evaporation at Lijiang (Xu et al, 2020)

In 1735 AD, official Qiaosun Jiang donated their salary to dredge the river and rebuild the dam (Niu,2007). Although the river required dredging and dam maintenance, water was still frequently able to flow out of Lake Chenghai. In 1762, the level of Lake Chenghai declined. The water level of Lake Chenghai experienced a sharp drop by more than 30 m in a short interval beginning at ~1779 AD, and the water could not outflow to the Jinsha River since then (Xu et al,2016). At this point, Lake Chenghai switched from an open to a closed lake (Xu et al, 2016). The Yongsheng County Water Affairs Bureau conducted textual research on the Haizha dam and showed that the lake cannot outflow when the water level is lower than the dam elevation (~1540 m) (Ji, 2014).

Xingyun and Fuxian lakes:Lake Xingyun and Lake Fuxian were once connected and formed a single lake. As inferred fromYunnan chorographyandYunnan general chorography(first published during the Ming dynasty), Lake Xingyun and Lake Fuxian could have been frequently connected during 1573 — 1627 AD. Today, as a result of continuous water level decline, the two lakes are separated (Liu,1991; Li, 2011). During the Qing Dynasty, there were records of flood disasters in both lakes, such as “the farmlands of three counties were flooded”,“the river bank was destroyed, the river level rose,and the surrounding farmland was flooded” for Lake Fuxian (Fang, 2001; Niu, 2007) and “dredging the outlets of Lake Xingyun and Lake Fuxian again, the floods of farmland and villages around the lake have been alleviated” for Lake Xingyun (Compilation Committee of Yiliang County Chorography, 1998).These historical records suggest that both lakes experienced wetter conditions from 1723 — 1780 AD,causing severe economic loss to local people.

Lake Yilong:Lake Yilong had multiple highlevel overflow records during the Ming and Qing dynasties (see time span in Fig. 3), such as “deep rain in Shiping County, lake water rose, and farmlands were flooded” in 1573 AD, “Lake Yilong and Lake Chirui overflowed” in 1879, and “Heavy rains for several months, floods outbreaking, Yilong and Chirui lakes were flooded and crops were inundated” in 1879(Liu, 1991; Compilation Committee of Yiliang County Chorography, 1998).

Lake Qilu:According to theYunnan provincial chorography, Lake Qilu had a wider lake area(corresponding to a higher lake level) during the Yuan Dynasty (Fig. 3), with a natural sinkhole located on the eastern coast. After ~ 1284 AD, the sinkhole was excavated to construct an outlet, and the surrounding shallow lake area was used as farmland. This suggests that the lake level was possibly low before the Yuan Dynasty, and people could have used the lake periphery as farmland. In 1284, the lake level began to rise and people needed to construct an outlet to artificially keep the lake low. In 1382, the lake dropped to~1801 m. In the later part of the Ming Dynasty (~ 1580 —1644 AD), the lake level dropped again (~1797.5 m,similar to the modern lake level). At the end of the Qing Dynasty (1893 — 1894), the lake level had risen by ~3 m, resulting in floods that inundated farmhouses and caused human fatalities (Compilation Committee of Yiliang County Chorography, 1998). However, as human activity largely influences the lake levels, the historical Lake Qilu water level information is less convincing in comparison with other lakes.

Fig. 3 Lake level variations in southwest China

Modern lake level decline:Drought and flood records in theYunnan chorographyspanning the past 100 a indicate an apparent drought in 1960 — 1990,during which the level of most lakes in Yunnan Province dropped. For example, in 1960, Lake Yangzong dropped by more than 1 m relative to previous years due to a decrease in precipitation. In 1963, the level in Lake Yangzong declined, exposing the water pumps along the lake. In the same year, Lake Qilu dropped by more than 1 m. In 1975, most of the small ponds in the Yuxi area dried up and the level of Lake Qilu dropped. In 1978, almost all of the lakes in the province declined. Among those, the Nanpan River system was perhaps the most striking, because the Yiliang County pumping station was unable to operate normally due to the decline in level of Lake Yangzong.From 1982 to 1983, the whole year was dry and the rivers stopped flowing; as a result, the levels of most lakes in Yunnan Province decreased (Compilation Committee of Yiliang County Chorography, 1998).In 1991, the level in Lake Chenghai was 1499 m,which was lower than the level in 1961 (1504 m),demonstrating a decrease of 5 m in 30 a (Ji, 2014).All of these results suggest that local precipitation is an important influence on annual/seasonal lake level variations in southwestern China. The lake level variations in the study area are similar in trend as revealed from modern observations (Fig. 2), which should be most likely ascribed to both natural forcing(like rising temperature and declining precipitation,etc.) and human activities (see below).

2.2 Comprehensive lake level variations in southwest China

By comparing lake levels reconstructed from beach evidence, such as at Lake Chenghai (Xu et al,2019; Xu et al, 2020), and lake level fluctuations inferred from historical literature records, we compiled an integrated water level profile for lakes in southwestern China (Fig. 3). The lake levels in southwestern China over the past ~1000 a can be roughly divided into four stages (Fig. 3), including two periods of high levels (before ~ 900 AD and 1350 — 1800) and two intervals of declining levels(900 — 1350 and 1800 — present).

StageⅠ(before ~ 900 AD):The beach evidence from Lake Chenghai indicates that lake level was ~40 m higher than modern lake levels (Fig. 3). Sedimentary proxies in Lake Lugu, Lake Erhai and Lake Chenghai also indicated broadly contemporary high lake level conditions in southwest China (Zeng et al, 2012; Xu et al, 2015a; Xu et al, 2016).

StageⅡ(900 — 1350, MWP):Levels in Lake Chenghai were low on average. Although some abrupt and short rises (e.g., ~ 1155 — 1185) occurred, the overall lake level was low during this stage (Xu et al,2019).

StageⅢ(1350 — 1800, LIA):Literature records in theYunnan chorographyandYiliang County chorographysuggest that the level of Lake Yangzong was high around 1400 and low in ~ 1562. TheYunnan chorographyrecorded that the water level of Lake Qilu decreased in the late Ming Dynasty (1580 — 1644).Historical records of other lakes indicated that most lakes in southwestern China were high from 1400 to 1800. Lake Chenghai was also at high in this stage.

StageⅣ(1800 — present):According to the historical records of drought and flood in southwestern China and instrumental records of the past 50 a, we conclude that the levels of most lakes declined during this period, which is also consistent with declining monsoon precipitation inferred from other proxies (Xu et al, 2012; Xu et al, 2016).

3 Discussion

3.1 Hydroclimate in southwest China

The hydroclimate patterns in southwestern China over the past millennium were “cold/wet-warm/dry”as revealed by numerous previous studies (Xu et al,2019; Xu et al, 2020). Specifically, the MWP was warm and dry, and the LIA was cold and humid (Zeng et al, 2012; Xu et al, 2015a; Xu et al, 2016). Both lake level information and historical records are consistent with these patterns.

StageⅠ(before ~ AD 900):The lake levels in southwest China were generally high, corresponding to humid conditions. High lake levels are supported by proxy indices in lake sediment cores. For example,C / N ratios and median grain size in Lake Lugu were high and coarse, respectively (Sheng et al, 2015),suggesting higher rainfall, stronger surface runoff, and higher terrestrial organic matter loads. Conifer pollen concentrations in a sediment core from Lake Erhai were relatively high, also indicating a relatively humid environment (Sheng et al, 2015).

StageⅡ(900 — 1350, the MWP):In this stage, lake levels in southwest China were generally low, reflecting “warm/dry” climate conditions. Lake Chenghai terrace/shoreline records (Xu et al, 2019;Xu et al, 2020) are consistent with lake level trends indicated by the average sediment grain size (Ye et al, 2018), and both suggest low levels. C/N ratios and sedimentary grain size in Lake Lugu were small(even though due to the uncertainties in chronology,the sedimentary C/N ratio and median particle size of Lake Lugu are slightly leading), respectively,suggesting weakened surface runoff under drier conditions (Sheng et al, 2015). Conifer pollen contents in Lake Erhai sediment were low, indicating a decrease in biomass in the catchment area under dry conditions (Xu et al, 2015b). It is worth noting that Lake Chenghai is a closed lake, so coarser grain sizes indicate lower lake levels and an arid environment, whereas Lake Lugu is an open lake and the sedimentary grain size reflects changes in surface runoff intensity.

StageⅢ(1350 — 1800, the LIA):Lake levels of most lakes in the southwest China rose and/or were kept high (even though there may be some small abrupt drops during this period), indicating generally humid hydroclimatic conditions. The C/N ratios and median grain size in Lake Lugu sediment were generally high and coarse, respectively, suggesting enhanced surface runoff.The conifer pollen contents in the Lake Erhai sediment core were high, indicating humid climate conditions (Sheng et al, 2015; Xu et al, 2015b), which generally coincided with changes in the level of Lake Chenghai (Fig. 3).The average grain size of Lake Chenghai sediment was fine in this stage, reflecting a deep-water environment and humid conditions, consistent with the overall “cold/wet” LIA (Ye et al, 2018).

StageⅣ(1800 — present, the CWP):Overall,the lake levels in southwest China decreased in this stage. The median grain size and C/N ratios of the Lake Lugu sediment were low (Sheng et al, 2015).The conifer pollen contents in the Lake Erhai sediment decreased significantly (Xu et al, 2015b), and the average grain size of sediment in Lake Chenghai became coarser (Ye et al, 2018). All of these results reflect dry conditions.

3.2 Hydroclimate of the Indian summer monsoon region over the past millennium

Our comparison shows that the “cold/wet-warm/dry” hydroclimate pattern existed in southwest China and the surrounding area, all of which was influenced by ISM (Fig. 4). For example, Sun et al (2012) showed that the sedimentary grain size in Cattle Pond on Dongdao Island (see location in Fig. 1a) was significantly finer during the MWP (1000 — 1400) than the subsequent LIA (1400 — 1850). Chu et al (2002) used proxies,such as TOC, C/N ratios, and biogenic silica content in sediments of Lake Huguangyan (Fig. 1a) in south China to propose that the climate was warm and dry during the MWP (880 — 1260) and cold and wet during the LIA (1320 — 1700). Warrier et al (2014)used magnetic susceptibility, sand content, and carbonate content to describe lake level variations of Lake Thimmannanayakanakere (Fig. 1a) in south India over the past 3700 a; they showed that the climate during the MWP (850 — 1200) was relatively dry, while during the LIA (1350 — 1750) it was wet. Using sediment stratigraphy, diatom fossils,and mosquito assemblages, Verschuren et al (2000)showed that lake levels in Lake Naivasha (Kenya;Fig. 1a) were lower during the MWP (1000 — 1270)and higher during the LIA (1270 — 1850), although there were three drought intervals during the LIA(Fig. 4c). Halfman et al (1994) showed that Lake Turkana in East Africa (Fig. 1a) was low during the MWP and high during the LIA (regardless of potential age uncertainties) using stacked records of sedimentary carbonate content (Fig. 4d). Alin and Cohen (2003) reconstructed lake level variations during the late Holocene in Lake Tanganyika (East Africa; Fig. 1a) based on sedimentary ostracod assemblages, and showed that lake levels were low during the MWP (1050 — 1250), high during the first part of the LIA (1250 — 1500), and low during the second half of the LIA (Fig. 4e). Mills et al (2014)generated a lake level record spanning ~ 1000 a based on the combined sedimentary diatom records of Lake Nyamogusingiri and Lake Kyasanduka (Fig. 1a) in Uganda, and showed that the MWP (1000 — 1200)was dry and the LIA was generally wet (Fig. 4f and Fig. 4g). Using sedimentary diatom records, Stager et al(2005) reconstructed lake level variations in Lake Victoria (East Africa; Fig. 1a). They showed that the lowest lake levels occurred during the MWP and the CWP, whereas during most of the LIA, the lake levels were abnormally high (Fig. 4h). In summary, changes in water levels of a large number of lakes in the ISM region over the past millennium had broadly similar trends, with low levels during the MWP corresponding to “warm/dry” conditions, and generally high levels during the LIA corresponding to “cold/wet”conditions. This is in contrast to the hydroclimate patterns in areas under the influence of the East Asian Summer Monsoon (EASM). During the LIA, colder tropical Pacific Ocean condition led to southward movement of the Intertropical Convergence Zone(ITCZ) and intensified Walker circulation, resulting in a weaker EASM and stronger ISM (Xu et al, 2015a;Xu et al, 2016).

Fig. 4 Hydroclimatic comparison of lake level variations (relative (%) or absolute (m)) in southwest China with those in theextended Indian summer monsoon region over the past millennium

3.3 Possible driving mechanism of lake level variations in southwest China

Previous studies have suggested that MWP and LIA hydroclimate patterns in the ISM region are related to changes in sea surface temperatures(SSTs) in the tropical Pacific (Xu et al, 2019; Xu et al, 2020). Higher SSTs in the eastern tropical Pacific correspond to decreases in ISM intensity,and vice versa (Kumar et al, 1999). One possible explanation is that when SSTs rise in the eastern tropical Pacific, corresponding to an El Niño state,the zonal SST gradient in the tropical Pacific may decrease, leading to a weakening of Walker circulation and eastward shift of its central position.Then, the southern Asian monsoon trough moves eastward, resulting in a decrease in ISM intensity and regional precipitation (Tokinaga et al, 2012).However, when SSTs in the eastern tropical Pacific decrease, corresponding to a La Niña state, the east — west thermal gradient in the tropical Pacific increases and Walker circulation strengthens,which may lead to increased precipitation in most ISM regions (Xu et al, 2016). For example, based on sand contents in a sediment core collected at El Junco (Conroy et al, 2008), the MWP was similar to an El Niño-like state during which the tropical/subtropical ISM region was dry, whereas the LIA period corresponded to the La Niña-like state and the corresponding climate was humid.

Recently, Tan et al (2019) proposed a potential pattern of north — south atmospheric circulation based on theδ18O record from stalagmites in Klang Cave in south Thailand and Liang Luar Cave in east Indonesia(Fig. 5c and Fig. 5e). The atmospheric circulation pattern postulated from theδ18O stalagmite record for the past 2000 a corresponded well with lake level variations in southwest China (Fig. 5). For example,in the MWP and CWP, stalagmiteδ18O differences were large, corresponding to lower lake levels in southwestern China and eastern Africa (Fig. 4 and Fig. 5). This further indicates that changes in tropical ocean SSTs cause large-scale atmospheric circulation changes in the ISM region, which in turn affects its hydroclimate conditions.

Fig. 5 Comparison of lake level variations in southwestern China and their potential forcing

3.4 Potential anthropogenic influence on lake level changes over the last millennium

In addition to natural climate forcing effects on lake level changes, water level variations in southwest China may also have been affected by intense anthropogenic activities during the last millennium.Southwest China is densely populated, with people clustered in the peripheral areas around lakes. In general, precipitation in the study area is relatively high, so the use of riverine water and surface runoff in wetter periods should have been relatively small,and lake level variations should have been more dominated by natural forcing. However, in drier periods, people divert rivers and surface runoff for industrial, agricultural usage, and daily life, so rainfall received by the lake would have been greatly reduced.The imbalance in water supply and related decline in lake levels should therefore be more striking in drier intervals, because annual evaporation is much greater than the annual precipitation over the study area(Fig. 2). In the future, as “warm/dry” conditions prevail over the ISM area, lakes in our study area may have a risk of continuous declining during future warm and dry intervals.

4 Conclusions

Knowledge of historical lake level variations and their forcing mechanisms is important for social and economic development, ecological/environmental protection, drought and flood prediction and/or adaptation, and understanding of regional/global climate forcing. In a comprehensive comparison of lake shoreline/terrace records, historical literature records, and other proxy-based reconstructions in southwest China, we found that lake level variations over the past millennium can be roughly divided into four stages: (1) high lake levels before~ 900 AD; (2) average low lake levels during the MWP (~ 900 — 1350 AD); (3) high lake levels during the LIA (~ 1350 — 1800 AD); (4) declining lake levels in the CWP. The results showed that lake level variations and hydroclimate conditions in southwest China are consistent with lake level and hydroclimate patterns over the extended ISM area.We contend that hydroclimate change in southwest China may be related to SST changes in the tropical Pacific Ocean and large-scale atmospheric circulation movement. Lake levels over southwest China and the extended ISM area may risk continuous decline under a warming climate in the intensely human-impacted Anthropocene.