Dynamic Changes in Lakes within the Selin Co Basin and Potential Drivers in Tibet

Prevailing lake changes on the Qinghai-Tibetan Plateau (QTP) have occurred. Selin Co, a representative saline lake in the central region of the QTP, has experienced signicant expansion, but the main cause for its dramatic expansion is still under debate. Based on Landsat images, meteorological data, and glacier and permafrost data, the dynamic changes in Selin Co and its surrounding small lakes were systematically discussed, and the driving factors behind these changes were further explored. The results suggest that from 1988–2017, the areas of Bange Co and Cuoe Lake showed slow, overall increasing trends at rates of 0.28 km 2 /yr and 0.11 km 2 /yr, respectively, and they exhibited upward trends before 2005 but downward trends afterward. The area of Selin Co substantially increased by 685.8 km 2 with a growth rate of 30.39 km 2 /yr, with a slow increase of 27.11 km 2 during the period from 1988–1997, a rapid increase of 510.53 km 2 from 1997–2005 and an increase of 148.16 km 2 from 2005–2017. Accordingly, the lake level and water volume of Bange Co slightly increased by 1.64 m and 0.088 km 3 , respectively, whereas those of Selin Co signicantly rose by 8.138 m and 17.47 km 3 , respectively. The changes in the areas of Bange Co and Cuoe Lake were mostly related to annual precipitation (AP). Enhanced glacial meltwater owing to rising, with a rapid reduction of 165.1 km 2 (39%) in the glacier area in the basin between 1980 and 2010, predominantly drove the dramatic expansion of Selin Co, followed by accelerated permafrost degradation, with signicant increases in the active layer thickness (ALT) (7.44 cm/yr) and soil temperatures at a 15-m depth (0.0346°C/yr). the area and lake height variations in a single lake (Selin Co), while comparative studies from the perspective of lake bodies lumped together under similar climate conditions are rarely involved. This is insucient to accurately analyze the inuence of local climate conditions on lake change, as well as the inuences of glaciers and permafrost. Therefore, it is necessary to analyze the relationship between the changing trends of lake group areas and inuencing factors. In this paper, based on multi-remote sensing images, meteorological data, glacier data and permafrost monitoring data, the dynamic variations in Selin Co and its surrounding lakes during the period from 1988–2017 were discussed, and the driving factor behind their changes was further explored with a comprehensive consideration of various factors. The main objective of this study is to improve the understanding of the signicant responses of lake changes to climate conditions at the basin scale over the QTP. heterogeneity, a weak positive correlation between lake area increase and permafrost coverage in basins was observed by examining 39 expanding lakes with high permafrost coverage ranging from 65 ~ 99% in an endorheic basin on the QTP (Liu et al. 2020b). The results were mainly attributed to permafrost types, ground ice contents and permafrost thermal conditions. The continuous permafrost coverage in the Selin Co basin was 1.3×10 4 km 2 , accounting for 28.6% of the total catchment area. Despite the low


Introduction
The Qinghai-Tibetan Plateau (QTP) contains 1236 lakes (> 1 km 2 ) with an area of ~ 41,800 km 2 , which accounts for 39.2% and 51.4% of total lake number and area in China (Ma et    found that the contribution of glacier mass changes to lake level rise in the Selin Co basin is only 5.52%, and glacier melting is not the main driving force of the lake level rise. Zhang et al. (2011) demonstrated that both participation runoff increases and evaporation decreases have led to the Selin Co expansion. Lei et al. (2013) suggested that although decreased lake evaporation and glacier mass loss contribute to the growth of the lake to a certain extent, the overall lake growth is mainly owing to the signi cant increase in precipitation. Zhou et al. (2015) found that the main contribution of lake in ows to the water storage change of Selin Co was approximately 49.5%.
As rising air temperatures on the QTP, permafrost has been degraded seriously ( (Song et al. 2020; Ji 2020). Recent studies have even quantitatively indicated that meltwater from permafrost degradation was responsible for 20%~21% of the outlet runoff in the Hulugou River and Shiyang River basins Li et al. 2016) and for 24%±2.4% of the mean runoff in the Gulang River basin (Gui et al. 2019). The melting of ground ice at shallow depths below the permafrost table in an endorheic basin on the QTP was found to account for 21.2% of the increase in lake volume from 2000 to 2016 (Liu et al. 2020b).
For the Selin Co basin, because of the lack of permafrost information, the effect of permafrost thawing on lake change has rarely been considered in previous studies. In reality, continuous permafrost is widely distributed in the Selin Co catchment with a large amount of underground ice content in permafrost regions. It is likely that permafrost degradation plays a signi cant role in in uencing the hydrological regime and water resources in the catchment and eventually results in enhanced lake growth. In addition, most of the above studies focused on the area and lake height variations in a single lake (Selin Co), while comparative studies from the perspective of lake bodies lumped together under similar climate conditions are rarely involved. This is insu cient to accurately analyze the in uence of local climate conditions on lake change, as well as the in uences of glaciers and permafrost. Therefore, it is necessary to analyze the relationship between the changing trends of lake group areas and in uencing factors. In this paper, based on multi-remote sensing images, meteorological data, glacier data and permafrost monitoring data, the dynamic variations in Selin Co and its surrounding lakes during the period from 1988-2017 were discussed, and the driving factor behind their changes was further explored with a comprehensive consideration of various factors. The main objective of this study is to improve the understanding of the signi cant responses of lake changes to climate conditions at the basin scale over the QTP.

Study area
Bange Co, Cuoe Lake and Selin Co are enclosed salt lakes in the central QTP; these lakes are located in Nagqu Prefecture, Tibet (Fig. 1). The study area covering Selin Co-Bange Co has a characteristic semiarid climate, with a mean annual air temperature (MAAT) of 0.78°C, annual precipitation (AP) of 315 mm, annual evaporation (AE) of 2,080 mm, wind speed of 3.9 m/s and relative humidity of 42%, as well as an average annual sunshine duration of 2,950 hours (Zhang et al. 2011). Precipitation mainly occurs from May to September. The Selin Co lake (31°32.7′N ~ 32°7.8′N, 88°31.7′E ~ 89°21.7′E) is distributed at the junction of Shenzha, Bango and Nima counties. It had an average surface elevation of 4,545.7 m a.s.l. during the period from 2000-2015 and an area of 2,059.5 km 2 during the period from 1988-2017 (from this study), with a maximum length of 80 km and width of 66 km (Zhang et al. 2013). The Selin Co catchment has an area of 45,530 km 2 with the Tanglha and Nyamqentanglha Mountains to the south. Several major rivers ow into Selin Co from different locations, including the Zajia Zangbo River from the north shore of the lake, Boqu Zangbo River from the east shore of the lake and Zagen Zangbo River from the west shore of the lake; among these rivers, Zajia Zangbo River is the longest river on the QTP with a length of 409 km (Bian et al. 2010;Du et al. 2014). The lake is mainly supplied by Geladandong Glacier, which is located in the upstream region of the basin and serves as a major water source for the lake (Zhang et al. 2011). There are many small lakes over the Selin Co basin, among which Cuoe Lake is nearest to Selin Co and has an area of 271.85 km 2 . Its main supply originates from the Daergawa Zangbo River and Xiagang Zangbo River, which ow from the southeast into the lake. Bange Co (31°47.21′N, 89°30.34′E) is located in Bango County. The average surface area was 126.37 km 2 during the period from 1988-2017, and the surface elevation was 4,521.22 m during the period from 1959-2003 (from this study). As shown in Fig. 1, it consists of Bange I, II and III. Bange I is a seasonal lake that is lled only by water in the rainy season and covers an area of 5.4 km 2 . Since 1983, due to the rising water level, Bange II and Bange III have merged into one larger lake (named Bange Co) with an area of 130 km 2 and depth of approximately 1.1 m (Zhao et al. 2006). The lake is supplied by snow meltwater via the Qiaga Zangbo River. A small wetland with an area of 35.89 km 2 is distributed in the catchment, covering an area of 2,479.77 km 2 .

Remote sensing
Landsat data (http://glovis.usgs.gov) provided more available information for delineating the lakes. Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager (OLI) images free of or with small fractions of cloud coverage (< 10%) were chosen to extract the lake surface areas and create a continuous time series of areal changes between 1988 and 2017. All of the images were acquired in October because lake areas in this month are relatively stable (Zhang et al. 2017c). Thirty scenes were used to delineate lakes between 1988 and 2017, and 36 scenes (one scene for each month) were acquired to determine the seasonal scale of the lake areas during the period from 2015-2017. The lake boundaries were extracted from the false color compositions (bands 5, 4, and 3 for Landsat TM/ETM + and bands 7, 6 and 5 for Landsat OLI as red, green, and blue, respectively) of raw Landsat images for each lake in ENVI 5.3. Then, visual examinations and manual editing of lake boundaries were conducted to delineate lakes in ARCGIS 10.2. All the map and image data were projected into the UTM coordinate system Zone 45 using the WGS-84 geodetic datum. The accuracy of manual digitization was controlled within one pixel.
Lake level changes over the QTP are poorly understood due to remoteness, high altitude, thin atmosphere, and harsh weather conditions in the region.
Satellite altimetry can serve as a powerful and complementary tool for hydrologic monitoring and studies. The water level changes of Selin Co were studied using the Globe Lake Level Change Dataset (http://www.geodoi.ac.cn/webcn/doi.aspx?id=990), which was produced by Liao et al. (2018). This dataset was compiled through lake boundary delineations, water level calculations, outlier removal, Gaussian ltering, and elevation system conversions based on multialtimeter data (Cryosat-2, Jason-2 and ENVISAT/RA-2). It contains lake level during 2002-2017 for 87 lakes in High Mountain Asia, and the error between the water levels of lakes from this dataset and from the Hydroweb water level products was less than 1 m.

Meteorological data
The air temperature, precipitation and evaporation data from the Bange (90.

Glacier and permafrost data
The First and Second Glacier Inventory of China (http://westdc.westgis.ac.cn) and the latest glacier dataset from Ye et al. (2017) were used to extract glacier coverage in the basin. Permafrost monitoring data were provided by the Cryosphere Research Station on the Qinghai-Tibet Plateau, Chinese Academy of Sciences. Soil temperatures at the active layer observation site (QT04) were monitored using 105T thermistor sensors with an accuracy of 0.1°C. Soil temperatures at the borehole site (QTB15) were monitored using thermistor sensors, which were made at the Chinese State Key Laboratory of Frozen Soil Engineering at Lanzhou and exhibited excellent sensitivity (± 0.01°C) in laboratory tests. These thermistor sensors were deployed at different intervals at different depths ( Table 1). The instruments were attached to a CR1000 data logger (Campbell Scienti c). Soil temperatures for the active layer were collected once every 30 min, and soil temperatures for the boreholes were automatically recorded 12 times per day at 2-hour intervals. The ALT was determined by measuring the maximum depth of the 0°C isotherm, as observed from the soil temperature pro les. Soil temperatures at a 15-m depth recorded from the borehole sites were used to describe permafrost temperatures in this study, because the zero-amplitude depths ranged from 10 to 15 m on the QTP. Permafrost temperatures at a 15-m depth were generally within − 2°C from the freezing point except in a few mountainous areas, which avoids effect of seasonal temperature variation (Wu et al. 2010). The ground ice content in the basin was extracted from the distribution of the ground ice content over the QTP, which was calculated by the following formula: , where Gi is the ground ice content (kg), is the bulk density of the soil (kg·m − 3 ), is the gravimetric water content of the soil (%), Z is the permafrost thickness (m) and S is the permafrost area (m 2 ). Note: MAGT is the mean annual ground temperature; DZAA is the depth of the zero annual amplitude of the ground temperature of permafrost. Lake changes during the period also occurred simultaneously in sizes. The shapes of the three lakes changed accordingly, especially dramatic changes in Selin Co (Fig. 3). The shape of Bange Co changed slightly; these changes occurred in the shallow areas east of Bange II. Cuoe Lake expanded southward continuously. There was an opening in the west that was closed in some years so that a large rock island in the west of the lake was inundated. Mainly, a signi cant change in the shape of Selin Co was observed; the lake expanded rapidly northward, southward and southeast-southward, and marked expansion occurred along the lake shoreline in the southeast direction. The maximum expansion distance was approximately 146.7 m in the northwest direction. Based on lake area and lake surface elevation data, the lake volume was calculated as follows: , where is the lake volume change from area (A1) and lake surface elevation (H1) to area (A2) and lake surface elevation (H2) (   The increasing trends in AP and MAAT exhibited three stages. The AP sharply decreased from 1988-1995, gradually increased from 1995 to 2005 and then slowly decreased (Fig. 7a). The MAAT gradually fell from 1988 to 1997, rapidly rose from 1997 to 2007 and then slowly rose (Fig. 7b). A signi cant decreasing trend in AE was observed from 1988 to 2017, which also displayed three stages (Fig. 7c). AE gradually declined from 1988 to 1997 but increased from 1997-2007; these variations were attributed to the decreasing and increasing MAAT, respectively. After 2007, AE sharply declined, and the decrease in wind speed may have been responsible for these changes. The area changes of Bange Co and Cuoe Lake corresponded well to the variations in the AP. The lake areas decreased slightly from 1988-1995, which agreed with the decrease in the AP. The lake areas continuously increased from 1995 to 2005 with increasing AP. After 2005, the lake areas decreased with increasing AP. This result suggested that the changes in the areas of Bange Co and Cuoe Lake were closely associated with the AP, and precipitation was the most critical in uencing factor. However, the area change of Selin Co coincided with the increasing MAAT. The area of Selin Co slowly increased from 1988 to 1997 with a slight increase in the MAAT, quickly increased from 1997 to 2005 due to a rapid increase in the MAAT and slowly increased from 2005 to 2017 with a slight increase in the MAAT. This indicates that the rapid expansion of Selin Co had a close relationship with the continuous increase in the MAAT, as the MAAT has important impacts on glacier ablation processes and permafrost thawing. The area changes of the three lakes were inconsistently correlated with AE; the areas decreased slightly from 1988-1995 as AE declined, continuously increased from 1995 to 2005 with increasing AE and then decreased in spite of the declining AE.

Glacier contribution
Bange Co and Cuoe Lake are nonglacier-fed lakes, and their water budgets mainly depend on precipitation and evaporation. As mentioned above, precipitation was the main driver behind their observed changes. Selin Co is a glacier-fed lake with high glacier coverage in its basin. There were 297 glaciers with a total area of 423.09 km 2 in 1980, accounting for approximately 1% of the basin area. Lakes supplied by many glaciers within their catchments may be more affected by increasing glacial meltwater than by precipitation (Zhang et al. 2011). As the MAAT recorded at Tanggula meteorological station increased at a rate of 0.03°C/yr between 2005 and 2015 (Fig. 8a), glaciers in the Selin Co catchment considerably retreated. The statistical results showed that from 1980 to 2010, the total glacier area decreased by 165.1 km 2 (39%); during this period, it decreased by 25.9 km 2 (6%) from 1980 to 2001 and by 139.1 km 2 (35%) from 2001 to 2010 (Fig. 8b). For Selin Co, a highly glacierized lake in a catchment on the QTP, such considerable glacier ablation generated a large amount of surface runoff from glacier meltwater, which contributed to the extension of Selin Co. Strong positive relationships between lake area change and glacier area change were observed for most large lakes on the QTP ( As permafrost degradation occurred, a large amount of ground ice content in the catchment melted; this melt was not only more likely to provide more water resources but also increased aquifer activation to enhance hydrological processes in the basin and further supply rivers and lakes, given the rise to a marked expansion of Selin Co. Meltwater directly in uenced the groundwater recharge and water levels of Selin Co or increased the amount of groundwater discharge as surface drainage. Some of the meltwater even directly drained to become surface runoff and supplied Selin Co. Moreover, the presence of icerich permafrost in the Selin Co basin served as a barrier layer due to its low hydraulic conductivity and permeability (Yang et al. 2003;Woo et al. 2008;Wang et al. 2009). Ice-rich permafrost impeded liquid water in ltration and the interaction between surface water and groundwater, which nally resulted in a large amount of direct surface runoff due to both rain and snow-glacier melting due to the lack of a water storage buffer effect.
As shown in Table 3, precipitation from May-September mainly accounted for 93% of the annual precipitation in the study site. Both precipitation and the air temperature increased from April, reached maximum values in July, and decreased after July. Evaporation showed an obvious increase from January to April and then decreased, but it continued to increase after September and appeared to peak in October. Large areas of Bange Co and Cuoe Lake observed from May-August were mainly attributed to the most signi cant increases in precipitation. For Selin Co, glacier ablation runoff mainly occurred from July to September due to rising temperatures (Zhang et al. 2009). According to thawing and freezing processes of the active layer of permafrost near the Tanggula region, the active layer begins to thaw downwards from the ground surface at the end of April, and the thawing process reaches its maximum depth in late autumn Hu et al. 2014). Glacial meltwater and ground ice meltwater signi cantly participate in mountainous discharge in glacial regions, and this discharge is greater than precipitation in this basin (Zhang et al. 1997). Therefore, glacier melting and permafrost thawing with high air temperatures during the period from June-August can account for the large area of Selin Co observed during the period from August-November well, with a corresponding response lag. At the seasonal time scale, the relationship between lake group areas and in uencing factors further con rmed that changes in the areas of Bange Co and Cuoe Lake were mainly related to increasing precipitation, and glacial and permafrost meltwater were signi cant factors in uencing Selin Co growth.

Hydraulic connection
Due to the strong summer monsoon-driven climate with temperatures 2 to 4°C higher and precipitation 40% to > 100% higher than the current values (Shi et al.1999;Shi et al. 2001), lakes on the QTP appeared to be in the "pan-lake stage" in the period of 40 ~ 25 ka B.P. During this time, the QTP was covered by large interconnected pan-lake systems with a total area of ~ 360,000 km 2 and a total volume of lake water > 530 million km 3  After the greatest lake period, because of the reduced lake level arising from the rapid uplift of the QTP and the cold climate, the great lake contracted and fell apart, so Bange Co, Cuoe Lake, Wuru Co and Yagedong Co were separated from Selin Co as independent lakes during the late Late Pleistocene, when the QTP gradually evolved into its present appearance (Lv et al. 2003;Meng et al. 2012b;Zhao et al. 2018). Therefore, a signi cant viewpoint of the hydraulic connection between Selin Co and its surrounding lakes has been illustrated by many studies but is under debate.
Bange Co, an isolated lake from the eastern part of Selin Co (Fig. 10), was connected with Selin Co by a sandy spit that consisted of sandy sediments with high permeability. Some authors have stated that a strong hydraulic connection despite a straight-line distance of approximately 8 km was the main cause of the dynamic change in Bange Co as the Selin Co lake level increased (Zhao et al. 2006  permafrost conditions strongly affect the discharge regime for regions with high permafrost (> 60%). Although the general applicability of the above statistical relationships exists at larger spatial scales in cross-regional permafrost basins with heterogeneity, a weak positive correlation between lake area increase and permafrost coverage in basins was observed by examining 39 expanding lakes with high permafrost coverage ranging from 65 ~ 99% in an endorheic basin on the QTP (Liu et al. 2020b). The results were mainly attributed to permafrost types, ground ice contents and permafrost thermal conditions. The continuous permafrost coverage in the Selin Co basin was 1.3×10 4 km 2 , accounting for 28.6% of the total catchment area. Despite the low coverage percentage, permafrost in the basin was located on the southern boundary of the continuous permafrost on the QTP, which exhibited more rapid permafrost warming and degradation than that on the central QTP. Additionally, as a consequence of accelerated permafrost degradation, a large amount of ground ice has melted signi cantly, resulting in profound impacts on lake change.
Similar phenomena also occurred in thermokarst lakes along the Qinghai-Tibet Highway. A eld investigation of unmanned aerial vehicles (UAVs) in August 2018, con rmed the existence of abundant thermokarst lakes close to the Qinghai-Tibet railway and highway (Fig. 11). Thermokarst lakes are a typical manifestation of permafrost degradation and develop as a result of the thawing of ice-rich permafrost or the melting of massive ground ice. Thermokarst lakes were spread between the Kunlun Mountain pass and the Fenghuo Mountain pass along the Qinghai-Tibet railway, where ice-rich and warm permafrost exists. The distribution of thermokarst lakes is closely related to the ice content and the permafrost temperature; 83.8% of thermokarst lakes are located in rich-ice permafrost regions, and 54.9% are located in high-temperature permafrost regions ). Permafrost degradation with melting ground ice initially promotes thermokarst, but subsequent lake drainage can further accelerate permafrost degradation (Smith et al. 2005).
The rapid expansion of Selin Co was signi cantly attributed to the increase in glacier meltwater, followed by that of permafrost meltwater. However, glacier melting and permafrost thawing into lakes should not increase the overall mass of the lakes and may decrease the mass of lakes because a portion of the melted water is lost through evaporation or is discharged to rivers that leave the TP. Additionally, the effect of permafrost degradation on the hydrological regime is complex and involves the potential interactions among climate change, permafrost degradation and groundwater ow; thus, it is di cult to access groundwater discharge to lake recharge. Since it remains challenging to evaluate glacier mass balances and permafrost meltwater, we could not quantify the associated contribution to lake growth.

Conclusion
Based on Landsat TM images, ETM + images, OLI images, meteorological data, glacier data and permafrost monitoring data, the dynamic changes in Selin Co and its surrounding small lakes were discussed, and the driving forces behind their changes were further explored. The results suggested that, from 2000 to 2017, the areas of Bange Co and Cuoe Lake showed slow overall increasing trends at rates of 0.28 km 2 /yr and 0.11 km 2 /yr, respectively, exhibiting The changes in the areas of Bange Co and Cuoe Lake were consistent with the precipitation trend, and precipitation was the main in uencing factor. The dramatic expansion of Selin Co was mainly related to glacial meltwater, followed by accelerated permafrost.

Con icts of Interest
The authors declare no con ict of interest.

Figure 1
A schematic representation of the study area. The digital elevation model was obtained from ASTER GDEM version 2.0 (http://www.gscloud.cn); the drainage basins were extracted from the HydroSHEDS dataset (http://hydrosheds.cr.usgs.gov); the glacier data shown were obtained from the First Glacier Inventory of China (images during 1950-1980); and the lake areas shown were from 2017. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Annual changes in the areas of Bange Co, Cuoe Lake and Selin Co from 1988-2017.

Figure 3
Changes in the shapes of Bange Co, Cuoe Lake and Selin Co. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.   The annual precipitation and mean annual air temperature at the Shenzha and Bange meteorological stations from 1988 to 2017.

Figure 7
The annual anomaly changes in the AP, MAAT and AE at the Bange meteorological station from 1988 to 2017.  (a) The distribution of ground ice content over the Selin Co and Bange Co catchments; (b) the annual changes in the ALT at the QT04 observation site and in the soil temperature at a 15-m permafrost depth recorded at the QTB15 observation site. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 10
The linkage between Selin Co and the surrounding lakes. The gray-blue area shows the paleo-Selin Co extension delineated by the altitudes of the highstand shorelines; the blue area indicates the current levels of the lakes around Selin Co (modi ed from Meng et al. 2012b). Note: The designations employed and