Comparison of Hydrological Patterns between Glacier-Fed and Non-Glacier-Fed Lakes on the Southeastern Tibetan Plateau

: Lakes on the Tibetan Plateau have experienced variations over the last several decades, and the delineation of lake dynamics is favorable for the regional water cycle and can serve as important information for plateau environmental research. This study focused on 57 lakes near the Tanggula Mountains on the southeastern Tibetan Plateau. Yearly inundations of the lakes in 1989–2019 and altimeter data available for 2003–2020 were integrated to illustrate the changing patterns of glacier-fed and non-glacier-fed lakes. These two groups of lakes presented very similar evolution stages. They both increased in 1989–1992, decreased in 1992–1996, increased rapidly in 1998–2005, and had batch-wise ﬂuctuations since 2005, with respective areas of around 5305.28 and 1636.79 km 2 in the last decade. The non-glacier-fed lakes were more sensitive to precipitation variation, and glacier-fed lakes were more sensitive to temperature changes. Based on lakes with obvious changes in water level, the whole water storage variations of the studied lakes were 1.90 Gt/y in 2003–2009, including 1.80 Gt/y for glacier-fed lakes and 0.10 Gt/y for non-glacier-fed lakes. The contribution from glacier melting in 2003–2009 amounted to 16.11% of the whole lake volume increase. In 2010–2020, water mass changes were 0.42 Gt/y for glacier-fed lakes and − 0.14 Gt/y for non-glacier-fed lakes, respectively. The volume increase of glacier-fed lakes in 2010–2020 was mainly due to the expansion of Selin Co. Selin Co experienced a water increase of about 0.46 Gt/y, and the other glacier-fed lakes experienced a decreasing volume of − 0.04 Gt/y. In 2010–2020, 99.43% of the glacier contribution supplied Selin Co.


Introduction
The Tibetan Plateau (TP) is famously named "The third pole" and the "Water tower of Asia" [1]. This plateau covers an area of approximately 2,500,000 km 2 , with an average altitude of more than 4000 m. The TP is the third-largest area of frozen water: around 46,000 glaciers with an area of 100,000 km 2 . The total area of lakes (>1 km 2 ) is greater than 45,000 km 2 [2]. Due to the large glaciated area, the TP is the source of several famous rivers-the Yellow River, Yangtze River, Ganges River, India River, Brahmaputra River, and so on-and thus provides water for the whole of Southeast Asia [3]. Glaciers influence the water resources of lakes on the TP [4]. The glacier area in west China has retreated by 18% since 2000 due to climate warming [5]. The illustration of how the Tibetan lakes have responded to climate change over the past few decades is a key scientific issue.
Due to its high altitude and complex terrain, the TP has a harsh environment, with strong solar radiation, low temperatures, and thin air [3]. Therefore, the intensity of human activity in the TP is low and the dynamics of the lakes are sensitive to climate variability.
As direct measurements on the plateau are limited [6][7][8], many researchers have used remote sensing techniques to account for the hydrologic changes of lakes in this remote region [9,10]. Lakes on the TP have shown large variability, with most lakes undergoing expansion since the late 1990s [11].
In the last decade, the total area of lakes greater than 10 km 2 on the TP increased from 37,111 km 2 to 40,015 km 2 [12], and the two largest lakes, Nam Co and Selin Co on the TP, raised in level by 4.37 m and 2.03 m, respectively [13]. In 2003In -2009, with the aid of precipitation and glacier meltwater, the lake volume in the Inner Basin on the TP increased at a rate of 7.72 Gt/y [14]. ICESat and CryoSat-2 data were used in conjunction to study the water level changes of Nam Co and the results showed that Nam Co's level increased at a rate of 24 cm/y in [2003][2004][2005][2006][2007][2008], then dropped at a rate of −9 cm/y in 2009-2013 [15]. Area-level curves were constructed based on ICESat data and Landsat images for 30 lakes on the TP, and the results indicated that the lake volume increased by 92.43 km 3 from the 1970s to 2011 [16]. Zhang used ICESat data to monitor the water level changes of 62 lakes on the TP and found that the average annual raise was 23 cm/y, and the results illustrated that the lake level rise was consistent with glaciers melting [17]. Sun found that seven out of 10 lakes in Inner Bain presented increasing trends and that ground runoff was the major contributor [3]. Meng established a water balance equation to quantify the contribution of glaciers to lake volume dynamics [18].
The driving forces for expansion were precipitation increases or permafrost/glacier meltwater induced by rising temperatures [11]. Lei et al. studied the lake growth in the interior TP since the 1970s and increasing precipitation since the late 1990s was revealed as the main driving factor [19]. Moreover, Tibetan lake expansions were highly related to permafrost degradation caused by warming [19]. Song investigated the water levels of Tibetan lakes based on multiple satellite altimeters, and attributed the lake evolution to changing precipitation and evaporation [15,20]. The percentage contributions of precipitation to lake volume dynamics are also indicated according to some modeling methods [6,[21][22][23].
Although the reasons were indicated according to the aforementioned research, the proportional contributions from of factors are not clear. It is also a challenge to carry out quantitative estimations due to insufficient field observations and uncertainties about modeling methods [14]. This research used multisource remotely sensed data to illustrate the changing patterns (area, level, and volume) of 57 lakes (greater than 1 km 2 ) located near Tanggula and the Nianqing Tanggula Mountains. The 57 lakes studied were separated into two types according to whether they were dependent on meltwater from surrounding mountain glaciers. Twenty-three lakes in glacier-fed basins were named glacier-fed lakes, while the rest were named non-glacier-fed lakes. Our assumption is that the glacier-fed and non-glacier-fed lakes had similar climatic conditions, with similar influences from climatic factors, so the different evolutionary patterns between glacier-fed and non-glacier-fed lakes can be attributed to the contribution from surrounding mountain glaciers. The delineation of lake dynamics can reflect hydrological responses to climate change [24] and the results are favorable for plateau environmental research [3,25].

Study Area
The studied area is bordered to the northeast by the Tanggula Mountains, and to the south by the Nianqing Tanggula range on the southeastern TP, where several large glaciers distribute ( Figure 1). The studied region has a semi-arid climate. The annual temperature is about −5-7 • C. The annual precipitation is 300-600 mm, and more than 80% of the rainfall occurs in May-October. All 57 lakes researched are greater than 1 km 2 and are located in 21 sub-basins, as Figure 1 shows. Four of the 21 basins are glacier-fed type, including 23 glacier-fed lakes. Selin Co and Nam Co, named "Holy lakes," are the two largest lakes on the TP with respective areas of 2447.74 km 2 and 2027.91 km 2 in 2019 [3]. Nam Co was about 139.51 km 2 larger than Selin Co in the 1990s, while Selin Co expanded by about 481.06 km 2 over the last two decades. The expansion of Nam Co was constrained by the surrounding ridges-the Nianqing Tanggula Mountains, where mountain glaciers Remote Sens. 2021, 13, 4024 3 of 15 are widely distributed-and Selin Co became the largest lake on the TP in the 2000s. The third-largest lake studied is Tso Ngön, which is around 286 km 2 . Tso Ngön is adjacent to Selin Co and the two lakes are sometimes interlinked through the northern end of Tso Ngön. The fourth-largest lake studied is Bamu Co at around 232 km 2 , located 30 km north of Nam Co; this lake mainly depends on land surface runoff in the basin. Twenty-eight lakes had their altimeter-derived water level and they occupied around 98.64% of the 57 studied lakes. The spatial distribution of the studied lakes is shown in Figure 1 and the characteristics of the 28 lakes are listed in Table 1. largest lakes on the TP with respective areas of 2447.74 km 2 and 2027.91 km 2 in 2019 [3]. Nam Co was about 139.51 km 2 larger than Selin Co in the 1990s, while Selin Co expanded by about 481.06 km 2 over the last two decades. The expansion of Nam Co was constrained by the surrounding ridges-the Nianqing Tanggula Mountains, where mountain glaciers are widely distributed-and Selin Co became the largest lake on the TP in the 2000s. The third-largest lake studied is Tso Ngön, which is around 286 km 2 . Tso Ngön is adjacent to Selin Co and the two lakes are sometimes interlinked through the northern end of Tso Ngön. The fourth-largest lake studied is Bamu Co at around 232 km 2 , located 30 km north of Nam Co; this lake mainly depends on land surface runoff in the basin. Twenty-eight lakes had their altimeter-derived water level and they occupied around 98.64% of the 57 studied lakes. The spatial distribution of the studied lakes is shown in Figure 1 and the characteristics of the 28 lakes are listed in Table 1. Figure 1. Spatial distribution of the studied lakes, watersheds, and meteorological stations on the southeastern TP. Red points on the lake indicate that CryoSat data are available for the lake and yellow points indicate that ICESat data are available. The Arabic numbers in black circles indicate the glacier-fed basins.

Yearly Data for Lake Inundation
The Global Surface Water Dataset (GSWD) was created by the European Commission's Joint Research Center based on more than 30 million scenes of Landsat series images from 1984 to 2019 [26]. Expert systems, visual analysis, and reasoning techniques were integrated to extract water surfaces on a global scale. Detailed information on this program is given in the research of Pekel et al. [26]. Monthly and yearly inundation, occurrence change intensity, seasonality, and transition products are also presented in the dataset. According to an accuracy assessment with the aid of 40,124 test points distributed both temporally (spanning 32 years) and spatially, the omission errors were less than 5% and the commission errors were less than 1% for GSWD [27]. Yearly data mapped the union of seasonality and permanent water distribution, depicting the largest state in a year. Since the yearly inundations of the studied lakes in 1984-1988 were missing, a series of data from 1989-2019 was chosen in this research. The annual products were visually checked based on Google Earth high-resolution images and a series of 30 m interpretation results. If obvious errors existed in the extracted inundation of a particular lake in a particular year, such as a large area of a lake being missed due to cloud coverage, then for that lake, the record for that year was removed. If there was some salt-and-pepper noise in the lake surface extraction, then these errors were manually corrected. To check the accuracy of the GSWD yearly data of the studied lakes, six series of 30 m interpreted results based on Landsat series images acquired in wet seasons in the years 1990, 2000, 2005, 2010, 2013, and 2014 were collected. We compared the areas of the studied lakes between GSWD and the interpretation results. The correlations of the six pairs of data were around 0.99-1. This demonstrated that the GSWD yearly data were able to accurately map the inundation dynamics of the studied lakes. orbit height of 717 km and an inclination of 92 • . The frequency of CryoSat data monitoring is every 369 days. The time interval of adjacent cycles is 30 days, and the space of adjacent ground tracks is around 1.5-10 km. CryoSat-2 operates in three modes: Low-Resolution Mode (LRM), Synthetic Aperture Radar (SAR) mode, and SAR Interferometric (SARIn or SIN) mode. The SARIn mode was employed on the TP [15]. The Ice, Cloud, and Land Elevation Satellite (ICESat) was launched in January 2003 and stopped working in 2009. The ICESat sent laser pulses, which were easily interfered with by clouds. The ICESat revisited the same cycle every 3-4 months, covering the latitude of ±86 • , with a footprint of 70 m in diameter. The resolution in the along-track direction was 175 m. ICESat data were suitable for monitoring small lakes and the estimation accuracy for inland lakes was around 3-10 cm [28,29]. The ICESat product Global Land Surface Altimetry Data (GLA14) were also used in this research. Eight lakes were covered by the ICESat data and 28 lakes were covered by the CryoSat data, as indicated in Figure 1 by the red and yellow points, respectively.

Altimeter Data for Measuring Lake Levels
The following formulas indicate how to derive the surface level from altimeter data. In Formula (1), surface level relative to the referenced ellipsoid is equal to altitude minus range. Altitude is the orbit altitude of the altimeter. Range is the distance between an altimeter and lake surface derived from the time taken for a radio signal to be reflected back [30]. As they are influenced by the solid earth tide and the pole tide, the two items need to be subtracted. In Formula (2), the range needs to have interferences from the environmental and geophysical corrections removed.
Corrected Range = Range + Wet troposphere correction + Dry troposphere correction + Ionosphere correction + Electromagnetic bias (2) The measurements from ICESat and CryoSat-2 data followed the above procedure. In this study, eight lakes, including the two glacier-fed lakes Selin Co and Nam Co, were covered by the ICESat data, as shown in Table 1, and the process of ICESat data was that in Zhang et al. [31]. ICESat were reported to have performed well on detecting Tibetan lake levels, about several to ten centimeters [28,29,32]. The level-2 SIRAL SARIn product of CryoSat-2 data was used in this research, and the processing was similar to that applied by Song et al. [15], who has used this product to detect the level changes of Nam Co and Selin Co, with the level bias of less than 0.1 m [15].
After obtaining altimetry measurements, several procedures need to be performed to refine the results. First, due to the variation in lake inundation, altimeter footprints sometimes fell on the lakeside. Thus, the internal buffered lake boundaries were applied as filters to remove footprints that fell outside the boundaries, avoiding the contamination of land signals. Second, the 3-sigma rule was applied to remove noise from the filtered results, and the mean value of the test results indicated the daily lake surface altitude. Then the daily results were assembled to create a time series of lake levels. Finally, as ICESat data used Topex/Poseidon reference ellipsoid, while CryoSat-2 used the World Geodetic System 84 (WGS84) as a reference [3], we converted the ICESat data into the WGS84 reference to maintain consistency.

Meteorological Data
Daily gauged climate factors from five meteorological stations in the study area from 1966 to 2019 were obtained from the China Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/). To compare with the studied results, daily measured records were accumulated on a monthly and yearly scale. The lake surface evaporation was estimated based on the Penman-Monteith equation, according to the measurements from the nearest gauged station. A detailed introduction of the Penman-Monteith equation is given in the research of Allen et al. [33].

Accuracy Assessment of Lake Level
To evaluate the accuracy of altimeter-derived results, gauged measurements of water level for Selin Co in 2000-2020 were obtained. Figure 2 displays comparisons between altimetry results and in situ records for the lake.
(http://cdc.cma.gov.cn/). To compare with the studied results, daily measured records were accumulated on a monthly and yearly scale. The lake surface evaporation was estimated based on the Penman-Monteith equation, according to the measurements from the nearest gauged station. A detailed introduction of the Penman-Monteith equation is given in the research of Allen et al. [33].

Accuracy Assessment of Lake Level
To evaluate the accuracy of altimeter-derived results, gauged measurements of water level for Selin Co in 2000-2020 were obtained. Figure 2 displays comparisons between altimetry results and in situ records for the lake. For Selin Co, the ICESat-and CryoSat-derived results were about 3 m higher than the gauged data due to the in situ measurements being based on the local reference, which showed a large bias with WGS reference. The fluctuations in the altimetry measurements agreed well with the in situ records, as shown in Figure 2a. Selin Co had 23 ICESat results and 50 CryoSat records. We selected the altimeter-derived results and in-situ results on the same dates to check the consistency. Correlations between altimeter results and gauged data were high, with R 2 values of 0.99 for ICESat levels and 0.90 for CryoSat results, indicating that the derived altimetry results had high accuracy when capturing lake fluctuations. However, Figure 2b,c suggests that the accuracy of the CryoSat results was lower than that of the ICESat results. In addition, some researchers have used several kinds of altimeter data with overlapping periods to study water level changes. In fact, for different altimeters, system differences exist in terms of orbit height, inclination, and revisit period. To remove the differences and integrate the data, correlations between different kinds of results were constructed during overlapping periods, and then the altimeter results were transformed based on relationship equations to make the level series dense. However, the ICESat data and CryoSat-2 data used in this research covered the time periods 2003-2009 and 2010-2020, respectively. There was no overlap period between them. Therefore, the lake level dynamics of the two time intervals were analyzed separately.

Interannual Lake Areal Variations
The studied lakes include 23 glacier-fed lakes and 34 non-glacier-fed lakes. Both groups showed expansion, and the total area increased from 5825.81 km 2 in 1989 to 6965.74 km 2 in 2019. The non-glacier-fed lakes had an area of around 1313.49-1617.76 km 2 , and the glacier-fed lakes had an area of around 4517.14-5352.88 km 2 during the studied period. The glacier-fed lakes occupied about 76.56% of the total area. Glacier-fed and non-glacierfed lakes presented very similar evolutionary patterns, accompanying several batch-wise increases ( Figure 3). These two groups of lakes presented very similar evolutionary stages. In the 1990s, both glacier-fed and non-glacier-fed lakes endured fluctuations. In the 2000s, For Selin Co, the ICESat-and CryoSat-derived results were about 3 m higher than the gauged data due to the in situ measurements being based on the local reference, which showed a large bias with WGS reference. The fluctuations in the altimetry measurements agreed well with the in situ records, as shown in Figure 2a. Selin Co had 23 ICESat results and 50 CryoSat records. We selected the altimeter-derived results and in-situ results on the same dates to check the consistency. Correlations between altimeter results and gauged data were high, with R 2 values of 0.99 for ICESat levels and 0.90 for CryoSat results, indicating that the derived altimetry results had high accuracy when capturing lake fluctuations. However, Figure 2b,c suggests that the accuracy of the CryoSat results was lower than that of the ICESat results. In addition, some researchers have used several kinds of altimeter data with overlapping periods to study water level changes. In fact, for different altimeters, system differences exist in terms of orbit height, inclination, and revisit period. To remove the differences and integrate the data, correlations between different kinds of results were constructed during overlapping periods, and then the altimeter results were transformed based on relationship equations to make the level series dense. However, the ICESat data and CryoSat-2 data used in this research covered the time periods 2003-2009 and 2010-2020, respectively. There was no overlap period between them. Therefore, the lake level dynamics of the two time intervals were analyzed separately.

Interannual Lake Areal Variations
The studied lakes include 23 glacier-fed lakes and 34 non-glacier-fed lakes. Both groups showed expansion, and the total area increased from 5825.81 km 2 in 1989 to 6965.74 km 2 in 2019. The non-glacier-fed lakes had an area of around 1313.49-1617.76 km 2 , and the glacier-fed lakes had an area of around 4517.14-5352.88 km 2 during the studied period. The glacier-fed lakes occupied about 76.56% of the total area. Glacier-fed and nonglacier-fed lakes presented very similar evolutionary patterns, accompanying several batchwise increases (Figure 3). These two groups of lakes presented very similar evolutionary stages. In the 1990s, both glacier-fed and non-glacier-fed lakes endured fluctuations. In the 2000s, the two groups showed rapid growth of 5262.70 and 1654.13 km 2 , respectively. In the 2010s, glacier-fed and non-glacier-fed lakes experienced small fluctuations, with respective areas of around 5305.28 and 1636.79 km 2 over the last decade. This finding coincided with the results of Zhang et al. [14], who studied the lake volume changes of the whole inner TP and found that the lakes experienced a slight decrease before 1995, followed by a rapid increase in the 2000s, and then a recent deceleration in the 2010s.
the two groups showed rapid growth of 5262.70 and 1654.13 km 2 , respectively. In the 2010s, glacier-fed and non-glacier-fed lakes experienced small fluctuations, with respective areas of around 5305.28 and 1636.79 km 2 over the last decade. This finding coincided with the results of Zhang et al. [14], who studied the lake volume changes of the whole inner TP and found that the lakes experienced a slight decrease before 1995, followed by a rapid increase in the 2000s, and then a recent deceleration in the 2010s. We divided the study period into six intervals and the change patterns of each interval for the two groups are shown in Table 2 Figure 4 shows three lakes with obvious expansions during the studied period. Nam Co and Selin Co are two representative glacier-fed lakes, and Zige Tangco is a non-glacierfed lake. Nam Co showed overall expansion over the past three decades. It experienced a fluctuation before 1996. Nam Co increased from 1960.45 km 2 in 1989 to 1967.36 km 2 in 1991, and then dropped to a minimum of 1955.33 km 2 in 1996. The lake rapidly increased at a rate of 8.16 km 2 /y in 1996-2005 and presented stability (around 2027.56 km 2 ) in the following years. Figure 4a shows expansion the most in the west and east due to the restriction of range in the southeast. Selin Co also underwent a fluctuation before 1996 with a mean value of 1813.37 km 2 . Since 1996, it experienced a fast expansion at a rate of 57.14 km 2 /y in 1996-2005, and then the increase rate dropped to 16.64 km 2 /y in 2005-2014. The annual area increased from 1803.59 km 2 in 1996 to 2411.30 km 2 in 2014. Then the lake was in a comparatively stable state in the following years. From the two curves in Figure 4a and b, we found that, before 2001, Selin Co was smaller than Nam Co, while after 2001, We divided the study period into six intervals and the change patterns of each interval for the two groups are shown in Table 2 Figure 4 shows three lakes with obvious expansions during the studied period. Nam Co and Selin Co are two representative glacier-fed lakes, and Zige Tangco is a non-glacierfed lake. Nam Co showed overall expansion over the past three decades. It experienced a fluctuation before 1996. Nam Co increased from 1960.45 km 2 in 1989 to 1967.36 km 2 in 1991, and then dropped to a minimum of 1955.33 km 2 in 1996. The lake rapidly increased at a rate of 8.16 km 2 /y in 1996-2005 and presented stability (around 2027.56 km 2 ) in the following years. Figure 4a shows expansion the most in the west and east due to the restriction of range in the southeast. Selin Co also underwent a fluctuation before 1996 with a mean value of 1813.37 km 2 . Since 1996, it experienced a fast expansion at a rate of 57.14 km 2 /y in 1996-2005, and then the increase rate dropped to 16.64 km 2 /y in 2005-2014. The annual area increased from 1803.59 km 2 in 1996 to 2411.30 km 2 in 2014. Then the lake was in a comparatively stable state in the following years. From the two curves in Figure 4a and b, we found that, before 2001, Selin Co was smaller than Nam Co, while after 2001, Selin Co presented continuous expansion and became the largest lake on the TP. The expansion of Selin Co mainly occurred in the north and southeast due to the flat and open terrain, as Figure 4b shows. Zige Tangco increased by 6.12 km 2 during the 1990s, with a rate of 1.22 km 2 /y, and then expanded at a rapid rate of 2. Selin Co presented continuous expansion and became the largest lake on the TP. The expansion of Selin Co mainly occurred in the north and southeast due to the flat and open terrain, as Figure 4b shows. Zige Tangco increased by 6.12 km 2 during the 1990s, with a rate of 1.22 km 2 /y, and then expanded at a rapid rate of 2.18 km 2 /y in 2001-2010. The expansion of Zige Tangco mainly occurred in the north part ( Figure 4c) and in the last decade, it maintained an area of 242.94 km 2 .

Fluctuations of the Lake Level
Twenty-eight lakes had altimeter-derived water level results and eight lakes had both ICESat and CryoSat-2 altimetry-derived measurements in 2003-2020. Those eight lakes accounted for about 70% of the total lake area, including three glacier-fed lakes, Selin Co, Nam Co, and Tso Ngön. The number of available ICESat-derived records for the eight lakes ranged from 2 to 23, spanning 1-7 years. Selin Co had the most ICESat results, with 23  If the p-value of a level curve was less than 0.05, then the lake was considered to have experienced clear variations, while, if the opposite was true, it was concluded that there was no obvious change in the lake. As Figure 5 shows, the changing patterns of lake level coincided with that of inundated area. Two glacier-fed lakes, Nam Co and Selin Co, and two non-glacier-fed lakes, Xuguo Co and Zige Tangco, presented obvious increasing trends of around 0.22-0.63 m/y during their own available periods of ICESat data, as Figure 5 shows. Because CryoSat-2 had dense ground paths, more results were available for the eight lakes. The number of CryoSat-derived results was 6-50, spanning the last decade. CryoSat results showed that three lakes Namka Co, Dongqia Co, and Selin Co had

Fluctuations of the Lake Level
Twenty-eight lakes had altimeter-derived water level results and eight lakes had both ICESat and CryoSat-2 altimetry-derived measurements in 2003-2020. Those eight lakes accounted for about 70% of the total lake area, including three glacier-fed lakes, Selin Co, Nam Co, and Tso Ngön. The number of available ICESat-derived records for the eight lakes ranged from 2 to 23, spanning 1-7 years. Selin Co had the most ICESat results, with 23  If the p-value of a level curve was less than 0.05, then the lake was considered to have experienced clear variations, while, if the opposite was true, it was concluded that there was no obvious change in the lake. As Figure 5 shows, the changing patterns of lake level coincided with that of inundated area. Two glacier-fed lakes, Nam Co and Selin Co, and two non-glacier-fed lakes, Xuguo Co and Zige Tangco, presented obvious increasing trends of around 0.22-0.63 m/y during their own available periods of ICESat data, as Figure 5 shows. Because CryoSat-2 had dense ground paths, more results were available for the eight lakes. The number of CryoSat-derived results was 6-50, spanning the last decade. CryoSat results showed that three lakes Namka Co, Dongqia Co, and Selin Co had obvious change trends, including three lakes that dropped at a rate of −0.18 m/y to −0.14 m/y, while Selin Co rose at a rate of 0. obvious change trends, including three lakes that dropped at a rate of -0.18 m/y to -0.14 m/y, while Selin Co rose at a rate of 0.19 m/y in 2010-2020. The other 20 lakes, including six glacier-fed lakes, only had CryoSat-2 results for the last decade. Seven of the 20 lakes presented decreasing trends with rates of −0.24 to −0.11 m/y, and the other 13 lakes showed no obvious changes. The water level fluctuations of Puga Co and Ringco Ogma with the largest decrease rates agreed well with the areal variations, as Figure 6 shows.  Table 3. Based on the lakes that had obvious level changes, we calculated the average level variation rates for glacier-fed and non-glacier-fed lakes during two separate periods, 2003-2009 and 2010-2020. We compared the difference between glacier-fed and non-glacier-fed lakes by area-weighted averaging of the change rates to calculate the average growth rate, which was the growth rate of water level in per square kilometer. In respect of average rate, glacier-fed lakes showed a greater rise (0.43 m/y) than non-glacier-fed lakes (0.36 m/y) in 2003-2009, while non-glacier-fed lakes retreated more (−0.17 m/y) than glacier-fed lakes (0.15 m/y) in 2010-2020. The glacier-fed lakes had a growth state overall during the last decade, mainly due to the constant expansion of Selin Co, and the other glacier-fed lakes showed a declining rate (−0.15 m/y) in 2010-2020. The lakes involved in the two separate study periods are shown in Table 3. Table 3. Comparison between glacier-fed and non-glacier-fed lakes in terms of water level change rate, area, and lake volume changes in 2003-2020.

Altimetry
ICESat ( Table 3. Based on the lakes that had obvious level changes, we calculated the average level variation rates for glacier-fed and non-glacier-fed lakes during two separate periods, 2003-2009 and 2010-2020. We compared the difference between glacier-fed and non-glacier-fed lakes by area-weighted averaging of the change rates to calculate the average growth rate, which was the growth rate of water level in per square kilometer. In respect of average rate, glacier-fed lakes showed a greater rise (0.43 m/y) than non-glacier-fed lakes (0.36 m/y) in 2003-2009, while non-glacier-fed lakes retreated more (−0.17 m/y) than glacier-fed lakes (0.15 m/y) in 2010-2020. The glacier-fed lakes had a growth state overall during the last decade, mainly due to the constant expansion of Selin Co, and the other glacier-fed lakes showed a declining rate (−0.15 m/y) in 2010-2020. The lakes involved in the two separate study periods are shown in Table 3. Table 3. Comparison between glacier-fed and non-glacier-fed lakes in terms of water level change rate, area, and lake volume changes in 2003-2020.

Altimetry
ICESat ( 1 The average growth rate means the growth rate of water level in per square kilometer. 2 The whole area is the sum of average annual area for glacier-fed or non-glacier-fed lakes during the two separate periods. 3 Changed volume = Average growth rate × Whole area × Water density (10 3 kg/m 3 ).

Driving Forces
The water balance of the studied lakes in the enclosed basins was mainly influenced by several input and output factors. The input variables were precipitation, ground runoff, and glacier/snow/permafrost meltwater. The ground runoff is the accumulated water, determined by the precipitation and evaporation over the land of the basin. The output variables include lake surface evaporation, water exchange between adjacent lakes, some groundwater outflow and infiltration. Of these driving variables, precipitation, evaporation, and temperature were the dominant factors. We analyzed the relationship between climate factors and lake area changes during the studied period to illustrate the influences of climate on lake dynamics. Daily gauged climate variables from 1966-2020 were

Driving Forces
The water balance of the studied lakes in the enclosed basins was mainly influenced by several input and output factors. The input variables were precipitation, ground runoff, and glacier/snow/permafrost meltwater. The ground runoff is the accumulated water, determined by the precipitation and evaporation over the land of the basin. The output variables include lake surface evaporation, water exchange between adjacent lakes, some groundwater outflow and infiltration. Of these driving variables, precipitation, evaporation, and temperature were the dominant factors. We analyzed the relationship between climate factors and lake area changes during the studied period to illustrate the influences of climate on lake dynamics. Daily gauged climate variables from 1966-2020 were obtained and the data were processed to be in the annual frequency to match the areal series. Figure 7 presents the areal changes of glacier-fed and non-glacier-fed lakes against yearly precipitation, temperature and lake surface evaporation based on gauged measurements from the five nearby meteorological stations.
The annual precipitation at the five stations presented no obvious change trends in 1966-2020 and some wet and dry years appeared alternately, as Figure 7a shows. The red line in Figure 7a  The statistics on lake areas indicate that precipitation was an important driving factor, and non-glacier-fed lakes had more dependence on this climate element.
The annual lake surface evaporation fluctuated and did not present any obvious change trend (Figure 7b), ranging from 91.49 to 131.52 mm. The mean value after 2000 was a little lower than before 2000. The lowest value was 91.50 mm in 1997, when the mean temperature was very low. There was no close correlation between lake surface evaporation and batch-wise patterns of lake area.
The temperature steadily increased during the study period (Figure 7c). The mean value of the five annual temperature records in the study area showed an increase since 1966. It rose from -1.09 • C in 1966 to 1.80 • C in 2020, at a rate of 0.05 • C/y. The annual temperature had the lowest value of −1.96 • C in 1997, when both glacier-fed and nonglacier-fed lakes had their smallest area. In terms of the expansion intensity, non-glacier-fed lakes were not as sensitive to temperature increases as glacier-fed lakes. With an increase in temperature, melting and retreating glaciers became an important water supply to glacierfed basins [14], while the lake area did not maintain constant expansion and grew slowly after 2005. This phenomenon may be related to decreasing precipitation in 2005-2015.
Remote Sens. 2021, 13, 4024 11 of 15 obtained and the data were processed to be in the annual frequency to match the areal series. Figure 7 presents the areal changes of glacier-fed and non-glacier-fed lakes against yearly precipitation, temperature and lake surface evaporation based on gauged measurements from the five nearby meteorological stations. The annual precipitation at the five stations presented no obvious change trends in 1966-2020 and some wet and dry years appeared alternately, as Figure 7a shows. The red line in Figure 7a is the mean value of the five stations and indicates that the average precipitation in 1974 (563.49 mm), 1980 (626.67 mm), and 2008 (627.72 mm) was higher than that of the surrounding years, while the precipitation in 1972 (339.30 mm), 1994 (346.03 mm), and 2015 (281.34 mm) was lower than that of the surrounding years. Annual gauged rainfall from the meteorological station near Nam Co was around 372.58-794.81 mm, as the cyan line shows in Figure 7a-about 178.89 mm/y higher than the measurements near Selin Co, indicated by the orange line in Figure 7a. The annual precipitation and lake area had different fluctuations in 1989-1994 and the relationship between these two factors was not clear. In 1994-2003, the rainfall dynamically increased by 19.17 mm/y, and the glacier-fed and non-glacier-fed lakes constantly expanded at high rates of 55.80 km 2 /y and 23.28 km 2 /y, respectively. The precipitation decreased in 2008-2015 and then got better. The yearly non-glacier-fed lake area was comparatively stable in 2005-2010 and then had a slight decrease after 2011, while glacier-fed lakes showed an obvious expansion trend in 2005-2013, at a rate of 18.32 km 2 /y, indicating that the glacier-fed lakes were not very sensitive to the precipitation as non-glacier-fed lakes. Non-glacier-fed lakes showed a higher On the whole, lake area change patterns are a complicated effect of temperature, lake surface evaporation, and precipitation. The area of non-glacier-fed lakes is more sensitive to precipitation and glacier-fed lakes are more sensitive to temperature, which regulates meltwater from glaciers or snow.

Lake Volume Changes and Glacier Contributions
According to Table 3, two glacier-fed and two non-glacier-fed lakes rose obviously in 2003-2009, two glacier-fed and eight non-glacier-fed lakes dropped, and Selin Co showed constant growth in 2010-2020. We determined the lake water budget changes in the two separate study periods according to the average annual area and average growth rate in Table 3. In 2003-2009, the average area of glacier-fed and non-glacier-fed lakes was 4158.80 km 2 and 264.14 km 2 , respectively, and the average growth rates for these two kinds of lakes were 0.43 m/y and 0.36 m/y, respectively. If the water density was assumed to be 10 3 kg/m 3 , then the whole water storage changes were derived according to Formula (3).
Changed volume = Average growth rate × Whole area × Water density (3) where Changed volume means the average changed water storage in one year, Average growth rate is the rate of level changes in per square kilometers, and the Whole area is the sum of average annual area for glacier-fed or non-glacier-fed lakes. The volume changes of these two groups were 1.90 Gt/y, including 1.80 Gt/y for glacier-fed types and 0.10 Gt/y for non-glacier-fed lakes. In 2010-2019, glacier-fed and non-glacier-fed groups had respective average areas of 2675.80 km 2 and 816.71 km 2 . As the level change rates for the two groups were 0.16 m/y and −0.17 m/y, respectively, we obtained water mass changes of 0.42 Gt/y for glacier-fed lakes and −0.14 Gt/y for non-glacier-fed lakes, as Table 3 shows. In addition, in 2010-2020, Selin Co maintained constant expansion at a rate of 0.19 m/y, and the other glacier-fed lakes experienced a retreat at a rate of −0.15 m/y. Therefore, Selin Co experienced a water increase of about 0.46 Gt/y, while the other glacier-fed lakes retreated at a rate of −0.04 Gt/y. The glacier-fed and non-glacier-fed lakes had the same geographic environments and so experienced similar climatic influences. If we assume that the climate has comparable effects on the two groups of lakes, and causes similar infiltration and permafrost contribution, then the difference in the change rates between glacier-fed and non-glacier-fed lakes must be caused by the supply from glacier meltwater. Therefore, the glacier-fed lakes had the same change rates as non-glacier-fed lakes due to climate: 0.36 m/y in 2003-2009 and −0.17 m/y in 2010-2020 as Table 3 shows. It can be seen that, for glacier-fed lakes, the average growth rate of 0.07 m/y (0.43 m/y-0.36 m/y) for glacier-fed lakes in 2003-2009 was attributed to glacier meltwater, which provided about 0.29 Gt/y water supply in 2003-2009, and this occupied 16.11% of the whole lake storage change (1.80 Gt/y). Zhang et al. (2017) has studied the complex pattern of lake volume changes on the TP and found that glacier mass loss contributed 13% water supply for lake volume increase [14]. This quantitative estimation was comparable with that in this research. In 2010-2020, the growth rate of 0.33 m/y (0.16 m/y-0.17 m/y) for glacier-fed lakes was due to glacier contribution, which supplied water of 0.87 Gt/y and 99.43% of it supplied Selin Co.

Conclusions
Yearly lake inundation data from 1966-2019 were used in this research to depict the area changes of 23 glacier-fed and 34 non-glacier-fed lakes. We found that, since 1989, these two kinds of lakes endured expansions, with variations, and increased greatly in 1998-2005. Glacier-fed lakes have kept growing since 2005, while non-glacier-fed lakes experienced a decrease in 2010-2017 but rebounded later on. The area in the 2010s was about 980.60 km 2 higher than in the 1990s. Selin Co maintained a constant expansion and has been the largest lake on the TP since 2001. Two kinds of altimetry data, ICESat (2003-2009) and CryoSat (2010-2020), were combined to delineate the level changes of 28 lakes. ICESat-derived level records had higher accuracy than CryoSat results when evaluated by in situ measurements. Four lakes presented obvious changes, with an increase rate of 0.22-0.63 m/y. Ten lakes showed an obvious retreat, while Selin Co maintained clear growth in 2010-2020.
Due to the remoteness of the TP, the evolutionary patterns of lakes are highly influenced by climate change. The lake area variations reflected the complex effect of climate factors. The qualitative analyses between lake area and precipitation, and lake surface evaporation and temperature were carried out to depict the correlations. We found that non-glacier-fed lakes were more influenced by precipitation and glacier-fed lakes were more influenced by the rising temperature. Therefore, glacier meltwater was an important water supply to glacier-fed lakes. Based on the lakes showing obvious level changes in 2003-2020, several conclusions can be made. In 2003-2009, the lake volume increase of four studied lakes with obvious level changes was 1.90 Gt/y and glacier-fed lakes accounted for 94.74% (1.80 Gt/y). In 2010-2020, the glacier-fed lakes experienced a volume increase of 0.42 Gt/y, while non-glacier-fed lakes retreated by −0.14 Gt/y. As the glacier-fed and non-glacier-fed lakes were in the same geographic environment, we assumed that there would be comparable effects of climate exerted on the two groups of lakes and calculated the contributions from glaciers. The water supply from glacier melting accounted for 16.11% of the lake volume increase in 2003-2009, and 99.43% of that supplied Selin Co in 2010-2020.
There were some internal differences between ICESat-and CryoSat-derived results due to the different altimeters. As there was no overlap period between the two datasets, we have illustrated the lake hydrologic dynamics of the two time ranges separately. Due to the limited availability of altimeter data, only 28 lakes had level records, so we just used lakes with obvious level changes to analyze the lake volume changes and contrasted the changing patterns between glacier-fed and non-glacier-fed lakes. Though this conclusion cannot reflect the whole state of the study area, it may serve as an important reference to study plateau environments.
This research explored the evolutionary difference between glacier-fed and non-glacierfed lakes, and some key findings were obtained. The results can reflect the effect of climate change and glacier retreat on the TP, and supply scientific data for the land surface processes modeling of the TP. The procedure and research program employed in this research can be expanded in other enclosed basins, and, in future research, the quantitative contributions of several driving forces, such as ground runoff, precipitation, and lake surface evaporation, can be derived based on the water balance equation and lake volume changes.