Continuous Intra-Annual Changes of Lake Water Level and Water Storage from 2000 to 2018 on the Tibetan Plateau

Abstract

There is a large amount of lakes on the Tibetan Plateau (TP), which are very sensitive to climate change. Understanding the characteristics and driving mechanisms of lake change are crucial for understanding climate change and the effective use of water resources. Previous studies have mainly focused on inter-annual lake variation, but the continuous and long-term intra-annual variation of lakes on the TP remains unclear. To address this gap, we used the global surface water (GSW) dataset and the Shuttle Radar Topography Mission (SRTM) DEM to estimate the water level and storage changes on the TP. The results indicated that the average annual minimum lake water level (LWLmin) and the average annual maximum lake water level (LWLmax) increased by 3.09 ± 0.18 m (0.16 ± 0.01 m/yr) and 3.69 ± 0.12 m (0.19 ± 0.01 m/yr) from 2000 to 2018, respectively, and the largest change of LWLmin and LWLmax occurred in 2002–2003 (0.45 m) and 2001–2002 (0.39 m), respectively. Meanwhile, the annual minimum lake water storage change (LWSCmin) and annual maximum lake water storage change (LWSCmax) were 125.34 ± 6.79 Gt (6.60 ± 0.36 Gt/yr) and 158.07 ± 4.52 Gt (8.32 ± 0.24 Gt/yr) from 2000 to 2018, and the largest changes of LWSCmin and LWSCmax occurred in the periods of 2002–2003 (17.67 Gt) and 2015–2016 (17.51 Gt), respectively. The average intra-year changes of lake water level (LWLCintra-year) and the average intra-year changes of lake water storage (LWSCintra-year) were 0.98 ± 0.23 m and 40.19 ± 10.67 Gt, respectively, and the largest change in both LWLCintra-year (1.44 m) and LWSCintra-year (62.46 Gt) occurred in 2018. The overall trend of lakes on the TP was that of expansion, where the LWLC and LWSC in the central and northern parts of the TP was much faster than that in other regions, while the lakes in the southern part of the TP were shrinking, with decreasing LWLC and LWSC. Increased precipitation was found to be the primary meteorological factor affecting lake expansion, and while increasing glacial meltwater also had an important influence on the LWSC, the variation of evaporation only had a little influence on lake change.

1. Introduction

The Tibetan Plateau (TP) is located in central Asia, with an average elevation more than 4000 m, and is called the third pole in the world along with other surrounding areas [1]. The climate change is obvious on the TP; the increasing rate of temperature was 0.039 °C/yr (p < 0.001) from 1960 to 2014 based on 71 stations [2], which was approximately twice the global average temperature rise rate (0.018 °C/yr) in the same period. The cause of warming was perhaps due to snow albedo feedback [3], and pronounced stratospheric ozone depletion [4]. Warming had affected the change of water resource on the TP, quick melt from glaciers and permafrost, and precipitation on the TP showed an upward trend [5,6]. At the same time, the evaporation in the western part of the TP had been decreasing and the trend of evaporation in the eastern part of the TP significantly increased [7]. Lakes are a sensitive indicator of climate change and an important part of the terrestrial hydrosphere; its space–time changes can well reflect climate and environmental changes [8]. The TP is home to the world’s highest and largest number of highland lakes, known as the Asian Water Tower [9,10]. According to the survey statistics, the TP currently has a total area of more than 4.65 × 10⁵ km² of 1171 lakes with an area more than 1 km² [11]. Lakes are little affected by human activities due to the high altitude and poor accessibility in the TP. Against the background of warming, humid climate and glacier retreat on the TP, the state of its lakes has undergone dramatic changes, and the changes have had an important impact on climate, ecology, and the lives of local residents [12,13,14]. Therefore, the changes of the lakes on the TP is an important index to study the changes of regional climate and ecological environment [15].

Continuous lake level monitoring is of great significance in climate change research. Traditional lake water level monitoring allows for real-time observation through hydrological sites. However, the TP is characterized by a large number of lakes with wide distribution range, high altitude, and remote location, and the actual hydrological monitoring environment is relatively harsh. At present, there are only several hydrological observation stations built on the TP (e.g., Qinghai Lake, YamzhoYumco, and Nam Co), which are very limited compared to the more than 1200 lakes in the region. As such, the measured lake water level data cannot fully reflect the change in lake level throughout the whole of the TP [16]. With the continuous progress of altimetry satellite technology, altimetry satellites have become an important tool for monitoring the changes in LWL and LWS [17,18,19]. For example, Hydroweb, DAHITI, and G-REALM data have been used to estimate the seasonal water volume change (of 1390.91 ± 79.91 km³) in 463 lakes and reservoirs with an area more than 10 km², aiming to advance the existing understanding of the role of the seasonal change of lakes and storage change in regulating global and regional water cycles and the contribution of surface water storage to sea level rise [20]. The Cryosat-2 SARIn mode data have a spatial resolution of 250 m and a 30-day sub-cycle along the orbital direction, and can be used to monitor lake water level changes [21]. CryoSat-2 has been used to monitored three stages of water level changes in more than 200 lakes on the TP from 2010 to 2019, showing that lake levels have risen by an average of 2.19 m in the past ten years, 56% of which occurred in the last three years [22]. The accuracy of ICESat/ICESat-2 data is higher than that of other altimetry satellites and, so, they have been widely used in LWL and LWSC monitoring [23,24,25,26]. ICESat data have been used to monitor the water levels of 111 lakes on the TP, indicating that 84% of lakes and 89% of salt lakes on the TP showed lake level rises, which was consistent with the observed accelerated melting of glaciers [24]. Using ICESat and ICESat-2, the water level and water storage changes of 242 lakes with an area more than 1 km² on the TP have been determined; the rate of change in water level was 0.20 ± 0.04 m/yr, and the total water storage of all lakes on the TP was 11.51 ± 2.26 Gt/yr (Gt = km³ with an assumption that the density of lake water is 1000 kg/m³) in the period 2003–2019 [10,27]. The changes in water level and water storage of 62 lakes on the TP from 2003 to 2018 have been observed using ICESat/ICESat-2, and the 62 lakes examined showed a mean rate of water level change of 0.28 ± 0.03 m/yr from 2003−2018, with 58 lakes increasing (mean rate 0.30 ± 0.03 m/yr) and 4 lakes decreasing (mean rate −0.12 ± 0.06 m/yr), and a total change of lake water storage of ~14 Gt/yr was estimated [28]. Lake water level change is an important index in the study of lake water storage change (LWSC); however, ICESat and Cryosat-2 can only monitor a part of the lakes on the TP [21,28].

The Shuttle Radar Topography Mission (SRTM) DEM (Digital Elevation Model) provides reliable terrain data for the TP. We can estimate the LWSC, according to the relationship between lake area and lake elevation. Yang et al. (2017) have used the SRTM DEM and Landsat imagery to estimate the LWSC (larger than 50 km²) on the TP, and showed that the LWSC had increased by 102.64 km³ (2.77 km³/yr) from 1976 to 2013, and compared the accuracy with the bathymetric results for five lakes, indicating that the average relative error was only 4.98% [29]. Yao et al. (2018) have suggested that the LWSC increased by 7.34 ± 0.62 Gt/yr during 2002–2015 in the inner TP by using a similar method as Yang et al. (2017), and the accuracy was considered to be approximately 7.2%, by comparing 18 lakes with Hydroweb hypsometry [30]. Qiao et al. (2019a) have also suggested that the average error between this method and ICESat data was approximately 7.2% [31]. Qiao et al. (2019b) have analysed the LWSC of 315 lakes (larger than 10 km²) on the TP, and showed that the LWSC had decreased by 23.69 Gt from 1976 to 1990 and increased by 140.8 Gt from 1990 to 2013 [32]. Zhang et al. (2021) combined with SRTM DEM, have suggested that the LWSC of 1132 lakes (larger than 1 km²) increased by 169.7 ± 15.1 Gt from 1976 to 2019, and by 157 ± 11.6 Gt in the entire inner TP, where the contribution of glacial meltwater to the increase of LWSC was less than 30% [33].

However, most previous studies have focused on the inter-annual LWSC, while there has been little research on the intra-annual and continuous analysis of the LWLC, LWSC, LWLC_intra-year, and LWSC_intra-year on the TP. Therefore, we analysed the changes and causes of the intra-annual LWL and LWSC on the TP from 2000 to 2018. The purpose of this study is to (1) extract the lake area of maximum and minimum water bodies from 2000 to 2018 based on the GSW data set; (2) estimate the LWLC and LWSC by utilizing the SRTM DEM, as well as analysing its change characteristics and spatial differences; and (3) analyse the driving mechanism of lake change on the TP considering the change trends of meteorological factors and glacier distribution.

2. Materials

We used the GSW data set from the European Joint Research Centre (JRC). The construction of the GSW data set involved the use of more than 4 million remote sensing images from Landsat 5, Landsat 7, and Landsat 8 from 1984 to 2018 to extract global surface water bodies in an expert system, which provides flexibility and visual analytics combining human cognitive and perceptual abilities, and the precision of boundary extraction was controlled in a pixel (30 m) [34].

Precipitation data are an important part of the water cycle when analysing terrestrial water storage; therefore, we obtained the Tropical Precipitation Measuring Mission (TRMM) data from the National Aerospace and Space Administration Goddard Space Flight centre (https://trmm.gsfc.nasa.gov/). Yang et al. (2018) used measured precipitation data from the China Meteorological Administration to test the uncertainty of the TRMM, and suggested that the accuracy of this data is high in the Taihang Mountains (R² = 0.97, RMSE = 10 mm, p < 0.01), Hengduan Mountains (R² = 0.98, RMSE = 8 mm, p < 0.01) and Karst area (R² = 0.98, RMSE = 12 mm, p < 0.01) in some typical mountainous regions in China [35], and suggested that the relationship was 0.87 at the annual precipitation and 0.91 at monthly scales by comparing with gauge stations [36].

The National Centre for Environmental Prediction (NCEP) Climate Forecast System (CFS) is a fully coupled model representing the interaction between the Earth’s atmosphere, oceans, land, and sea ice [37]. The RMSE of this dataset was near 1 °C below 200 hPa, and 1.5 °C at 100 hPa in the eastern TP, and 1.8 °C in the western part of TP [38]. We used the meteorological factors of surface temperature from the NCEP data set to describe the annual average temperature change on the TP.

For endorheic lakes, lake surface water evaporation may be the only way for lake water to escape. Therefore, lake evaporation is also an important factor influencing lake water balance. However, as no complete and mature data set for lake evaporation exists at present, we used the actual evaporation data set for the Tibet Plateau, obtained from the National Qinghai-Tibet Plateau Scientific Data Center (http://data.tpdc.ac.cn, accessed on 1 February 2023). This data set contains the actual monthly average surface evaporation of the TP from 2001 to 2018, with a spatial resolution of 0.1 degrees. The data set mainly takes satellite remote sensing data (MODIS) and re-analysis meteorological data (CMFD) as input, and is calculated using the surface energy balance system model (SEBS). In the process of calculating turbulent flux, a sub-grid terrain drag parameterization scheme is introduced, in order to improve the simulation of surface sensible heat flux and latent heat flux. In addition, the evaporation output from the model was verified using the observation data of six turbulent flux stations on the TP, showing high accuracy with small RMSE (9.3–14.5 mm/month). This data set can be used to study the land–atmosphere interaction and water cycle characteristics of the TP [7].

The data set of China’s Second Glacier Inventory was obtained from the National Qinghai-Tibet Plateau Scientific Data Center (http://data.tpdc.ac.cn), comprising vector data. This data set is based on Landsat series remote sensing satellite data, combined with a global digital elevation model to obtain glacier boundaries and glacier attributes within China. The accuracy of this dataset was evaluated with two positioning accuracies, suggesting that this dataset had a high accuracy with ±3.2% [39,40].

3. Methods

3.1. Lake Area Extraction from the GSW

The GSW data set has a raster data format with a spatial resolution of 30 m, and is divided into two categories: map products and water history. We mainly used the maximum water extent from the map products and the yearly water classification history from water history, and all data were downloaded using JavaScript. The maximum water extent included all water bodies that can be monitored using remote sensing imagery from 1984 to 2018. We extracted lakes with a maximum extent (LME) of more than 10 km² on the TP from 1984 to 2018 using Google Earth Engine. Seasonal water bodies can reflect seasonal changes in water bodies during the year, while changes in permanent water bodies can truly reflect annual changes in water bodies. The details of the processing are given below.

(1) We obtained the LME and yearly water classification history (included permanent water and seasonal water) data from the Google Earth Engine. (2) According to the yearly water classification history data, we combined LME areas of more than 10 km² on the TP to select seasonal and permanent water bodies to calculate the annual maximum areas of lakes (LA_max), and used the permanent water to calculate the annual minimum areas of lakes (LA_min) using the Google Earth Engine. (3) As the GSW data set was obtained from the original images (e.g., Landsat 5, Landsat 7, and Landsat 8), some lake boundaries may have been strip-damaged or subjected to natural factors (e.g., clouds, rivers, glaciers, wetland), resulting in partial water bodies not being fully extracted. Thus, we deleted low-quality data and visually interpreted and manually modified the boundaries of the lakes in conjunction with Google Earth. (4) For missing annual lake water body data, we fit the area using the years near the missing data. (5) All data were defined uniformly using the Albers conical equal area projection coordinate system. The distribution of water bodies in 2000 and 2018 are shown in Figure 1, where Figure 1a (2000) and Figure 1b (2018) show the seasonal water bodies, Figure 1c (2000) and Figure 1d (2018) show the minimum lake extent, and Figure 1e (2000) and Figure 1f (2018) show the maximum lake extent.

3.2. LWL and LWSC Estimation Based on the GSW and SRTM

The SRTM was jointly surveyed by NASA and the National Bureau of Surveying and Mapping (NIMA) of the U.S. Department of Defense on 11 February 2000 [41]. The mission obtained radar image data between 60 degrees north latitude and 56 degrees south latitude, covering more than 80 percent of the Earth’s land surface, and processed it to create a digital elevation model. Here, we used the one arc-second SRTM-1 DEM with a mean error (ME) of 0.11 m and a mean absolute error (MAE) of 3.50 m, which has been previously validated for High Mountain Asia [42]. The holes in this data set have been filled using open-source data such as ASTER GDEM 2, GMTED 2010, and NED. Since 2000, the water level of most lakes in the TP has increased and, so, the SRTM DEM offers the possibility of establishing the relationship between lake area and lake level, thus allowing for estimation of the LWSC [29,32]. We used a 100 m buffer outward of the LME, recorded the corresponding area from the lowest elevation based on the SRTM DEM and, for every 1 m increase in elevation, we summed the recorded area until the recorded area was equal to the 100 m buffer area outward of the LME. At this point, we stopped the area recording and accumulation. Based on the SRTM DEM, we established a relationship between lake area and elevation. The LWSC can be obtained according to changes in lake area and water level [43]. This equation had been used to estimate the LWSC in previous studies [29,31,32,33].

$$\mathsf{\Delta }\mathrm{V}=\frac{1}{3}\times \left({\mathrm{S}}_1+\sqrt{{\mathrm{S}}_1\times {\mathrm{S}}_2}+{\mathrm{S}}_2\right)\times \mathsf{\Delta }\mathrm{h}$$

(1)

△V represents the LWSC during the two periods, S₁, S₂ represent the lake surface areas of the two periods, and △h represents the change of the LWL during the two periods.

At the same time, we calculated the LWLC_intra-year and LWSC_intra-year based on the LA_min and LA_max obtained in each year. ${\mathsf{\Delta }\mathrm{h}}_{\mathrm{j}}$ represents the average LWLC_intra-year of lake j from 2000 to 2018. ${\mathrm{h}}_{\mathrm{jmin}}$ and ${\mathrm{h}}_{\mathrm{jmax}}$ represent the LWL_min and LWL_max of the lake j in each year, respectively. According to the acquired lake area and the change in LWL, the LWSC was then calculated. ${\mathsf{\Delta }\mathrm{V}}_{\mathrm{j}}$ represents the LWSC_intra-year of lake j from 2000 to 2018, while ${\stackrel{-}{\mathrm{A}}}_{\mathrm{jmin}}$ and ${\stackrel{-}{\mathrm{A}}}_{\mathrm{jmax}}$ represent the average LA_min and average LA_max from 2000 to 2018, respectively.

$${\mathsf{\Delta }\mathrm{h}}_{\mathrm{j}}=\frac{1}{\mathrm{n}}{{\displaystyle \sum }}_{\mathrm{i}}^{\mathrm{n}}\left({\mathrm{h}}_{\mathrm{jmax}}-{\mathrm{h}}_{\mathrm{jmin}}\right)$$

(2)

$${\mathsf{\Delta }\mathrm{V}}_{\mathrm{j}}=\frac{{\mathsf{\Delta }\mathrm{h}}_{\mathrm{j}}\times \left( {\stackrel{-}{\mathrm{A}}}_{\mathrm{jmax}}+ {\stackrel{-}{\mathrm{A}}}_{\mathrm{jmin}}+\sqrt{ {\stackrel{-}{\mathrm{A}}}_{\mathrm{jmax}}\times {\stackrel{-}{\mathrm{A}}}_{\mathrm{jmin}}}\right)}{3}$$

(3)

4. Results

4.1. Accuracy Verification of Lake Water Level Changes

To verify the feasibility of retrieving the LWL and LWSC between the lake areas obtained from GSW and the SRTM DEM, we compared the LA_min and LA_max of 410 lakes in 2000 with the area of lakes obtained from the SRTM DEM in 2000, and found that they were highly correlated (Figure 2). Comparing the LA_min (Figure 2a) and LA_max (Figure 2b) with the SRTM DEM lake area in 2000, the correlation coefficient (R²) was greater than 0.99 and the standard error (RMSE) was 15.277 km² and 17.931 km², respectively. Qinghai Lake has continuous lake water level observation data from 1976 to 2018. Zhang et al. (2019) have compared the ICESat/ICESat-2 altimetry satellite observation data from 2000 to 2018 with the measured water level of Qinghai Lake, where R² = 0.95 and RMSE = 0.1 m, and the research results demonstrated that altimetry satellite data could be well-applied to lake water level monitoring. We utilized the water levels of 61 lakes from Zhang et al. (2019), which could be monitored on the TP observed by ICESat/ICESat-2 from 2003 to 2018, for comparison with the water levels of the 61 lakes monitored in our study, and found that the overall correlation between the two was relatively high. Figure 2 shows that the change of LWL_min (Figure 2c) and LWL_max (Figure 2d), compared with the ICESat/ICESat-2 satellite altimetry water level, where the R² values were both greater than 0.75, and the average error was 0.77 m.

4.2. The Change of LWL and LWSC on the TP

The lakes were found to be generally expanding from 2000 to 2018 on the TP (Figure 3 and Figure 4). The LA_min increased by 5161.32 ± 340.81 km² (271.65 ± 17.94 km²/yr; Figure 3a), with 58 lakes shrinking and 352 lakes expanding from 2000 to 2018 (Figure 4a). The LA_min increased the most in 2003 (818.52 km²) and decreased the most in 2013 (27.24 km²). The LA_max increased by 5808.81 ± 255.88 km² (305.73 ± 13.47 km²/yr), with 19 lakes shrinking and 391 lakes expanding from 2000 to 2018. The LA_max increased the most in 2012 (896.18 km²) and decreased the most in 2014 (58.51 km²; Figure 3a and Figure 4b).

We integrated the LA and the LWLC to calculate the average LWLC in multiple lakes and found that the LWL and LWSC showed an overall increased trend from 2000 to 2018 (Figure 3b,c). Compared with 2000, the average LWL_min increased by 3.09 ± 0.18 m (0.16 ± 0.01 m/yr) in 2018, increased the most in 2003 (0.45 m), and changed the least in 2013 (0.02m) (Figure 3b and Figure 4d). The average LWL_max increased by 3.69 ± 0.12 m (0.19 ± 0.01 m/yr) from 2000 to 2018, increased the most in 2001 (0.39 m), and changed the least in 2014 (0.05 m) (Figure 3b and Figure 4e).

The total LWSC_min increased by 125.34 ± 6.79 Gt (6.60 ± 0.36 Gt/yr), where the LWSC_min decreased by 3.06 ± 0.35 Gt (0.16 ± 0.02 Gt/yr) in shrinking lakes, and increased by 128.40 ± 6.77 Gt (6.76 ± 0.36 Gt/yr) in expanding lakes from 2000 to 2018. The LWSC_min changed the most in 2003 (17.67 Gt) and the least in 2013 (0.97 Gt; Figure 3c and Figure 4g). The total LWSC_max increased by 158.07 ± 4.52 Gt (8.32 ± 0.24 Gt/yr), where the LWSC_max decreased by 6.42 ± 0.49 Gt (0.34 ± 0.03 Gt/yr) in shrinking lakes and increased by 164.50 ± 4.30 Gt (8.66 ± 0.23 Gt/yr) in expanding lakes from 2000 to 2018. The LWSC_max changed the most in 2016 (17.51 Gt) and the least in 2014 (2.08 Gt).

Next, we analysed the intra-annual changes in lakes. The average annual change in lake area was 1614.53 ± 289.87 km² from 2000 to 2018 on the TP, with the least change in 2003 (1119.83 km²) and the most in 2016 (2304.61 km²; Figure 3a and Figure 4c). The average annual change in LWL was 0.98 ± 0.23 m, with the least change in 2003 (0.61 m) and the most in 2018 (1.44 m; Figure 3b and Figure 4f). The average LWSC_intra-year was 40.19 ± 10.67 Gt from 2000 to 2018, which changed the most in 2018 (62.46 Gt) and the least in 2003 (24.05 Gt).

4.3. The LWLC of Typical Lakes on the TP

As shown in Figure 5, we selected the continuous LWLC of six typical lakes (four expanding and two shrinking lakes) in the study area for further analysis. The LWL of Qinghai Lake increased significantly faster. Over the past 19 years, the LWL_min in Qinghai Lake rose by 1.17 ± 0.29 m, the LWL_max rose by 3.87 ± 0.60 m, and the LWLC_intra-year was 0.67 ± 0.90 m during this period (Figure 5a). Affected by the overflow of Zhuonai Lake in the upper reaches of Kusai Lake and Salt Lake, the water level of Kusai Lake rose rapidly in 2011. Since 2011, the LWL of Kusai Lake and Salt Lake has risen sharply, and then stabilized. The LWL_min of Kusai Lake increased by 10.79 ± 0.67 m, the LWL_max of Kusai Lake increased by 11.48 ± 1.24 m, and the LWLC_intra-year was 2.60 ± 1.34 m (Figure 5b). Meanwhile, the LWL_min of Salt Lake rose by 19.85 ± 2.44 m and the LWL_max of Salt Lake increased by 23.66 ± 3.45 m. The LWLC_intra-year was 1.12 ± 0.29 m during this period, which changed the most in 2011, with a value of 6.9 m (Figure 5c). The LWL of Selin Co has continued to rise, with the LWLC_intra-year showing an upward trend. The LWL_min of Selin Co increased by 7.77 ± 0.85 m, its LWL_max increased by 8.64 ± 0.81 m, and the LWLC_intra-year was 1.12 ± 0.69 m from 2000 to 2018 (Figure 5d). Zhuonai Lake broke out in 2011 and, consequently, the water level began to drop. The LWL_min of Zhuonai Lake decreased by 0.49 ± 0.03 m, its LWL_max decreased by 5.74 ± 1.40 m, and the LWLC_intra-year was 2.86 ± 2.19 m during this period (Figure 5e). The LWL of the Yamdrok Lake presented an overall downward trend, its LWL_min decreased by 3.21 ± 0.51 m, its LWL_max decreased by 5.79 ± 0.65 m, and the LWLC_intra-year was 0.92 ± 0.88 m during this period (Figure 5f).

4.4. Spatial Difference Analysis of Lake Changes on the TP

To analyse the spatial changes of the lakes, the TP has previously been divided into multiple sub-regions, according to the size of lakes, trend of lake changes, and basin boundaries [32]. We referred to this watershed division method and divided the study area into six parts to analyse the spatial distribution of lakes on the TP. We divided those large lakes with large changes into sub-area A, including Nam Co., Selin Co., Zharinanmu Co., and Tangra Yumco, and the area in the northern part of the TP (where the LWSC was slightly smaller than that of area A) was divided into sub-area B. The lake area having LWSC with the smallest increase was designated as area C, located in the middle of the study area. The part where glaciers are widely distributed, the average annual temperature is low, the annual precipitation is less, and the lake area and water storage changes are greater than sub-area C was divided into sub-region D. The southern part of the study area, where lake changes showed a downward trend, was divided into area E. Finally, the eastern part of the TP was divided into area F. The six parts of the study area comprised four sub-regions (A, B, C, and D) in the inner TP, and two further sub-regions (E in the north-eastern part of the TP and F in the southern part of the TP); see Figure 6.

As shown in

Table 1. Changes in LA, LWLC, and LWSC in different Areas of the TP from 2000 to 2018.

	Area A	Area B	Area C	Area D	Area E	Area F	Total
Number of lakes	116	75	93	33	30	33	380
LACmin (km2/yr)	88.12 ± 9.84	105.33 ± 5.74	30.08 ± 3.08	25.78 ± 0.96	−7.59 ± 1.05	20.28 ± 1.18	262.01 ± 17.42
LACmax (km2/yr)	92.29 ± 9.27	121.24 ± 4.90	35.48 ± 2.84	27.71 ± 0.63	−5.80 ± 1.15	23.23 ± 1.41	294.15 ± 13.29
Average of LACinter-year (km2)	410.08 ± 54.43	624.02 ± 131.54	149.91 ± 67.60	56.16 ± 20.47	104.15 ± 22.97	147.52 ± 35.29	1491.83 ± 272.24
LWLCmin (m/yr)	0.18 ± 0.02	0.28 ± 0.01	0.15 ± 0.01	0.24 ± 0.01	−0.04 ± 0.01	0.08 ± 0.01	0.17 ± 0.01
LWLCmax (m/yr)	0.20 ± 0.02	0.31 ± 0.01	0.17 ± 0.01	0.27 ± 0.01	−0.06 ± 0.01	0.18 ± 0.02	0.20 ± 0.01
Average of LWLCinter-year (m)	0.88 ± 0.14	1.68 ± 0.35	0.60 ± 0.27	0.59 ± 0.19	0.89 ± 0.21	0.77 ± 0.70	0.97 ± 0.23
LWSCmin (Gt/yr)	2.95 ± 0.31	1.95 ± 0.08	0.43 ± 0.03	0.67 ± 0.02	−0.11 ± 0.02	0.44 ± 0.06	6.32 ± 0.34
LWSCmax (Gt/yr)	3.34 ± 0.29	2.40 ± 0.08	0.53 ± 0.03	0.75 ± 0.02	−0.19 ± 0.03	1.11 ± 0.14	7.94 ± 0.23
Average of LWSCinter-year (Gt)	14.70 ± 2.63	12.14 ± 3.20	1.74 ± 0.88	1.64 ± 0.59	2.74 ± 0.67	4.52 ± 4.16	37.48 ± 12.11

, there were 116 lakes in area A, 75 lakes in area B, 93 lakes in area C, 33 lakes in area D, 30 lakes in area E, and 33 lakes in area F. We found that the expansion rates of LA_min and LA_max were 262.01 ± 17.42 km²/yr and 294.15 ± 13.29 km²/yr from 2000 to 2018, with 380 lakes, and the average LAC_inter-year was 1491.83 ± 272.24 km². The area of lakes in the north-eastern area B and the south-eastern area A of the inner TP presented a faster rate of change; the LA_min and LA_max in area B increased by 105.33 ± 5.74 km²/yr and 121.24 ± 4.90 km²/yr, respectively, and the LA_min and LA_max in area A increased by 88.12 ± 9.84 km² and 92.29 ± 9.27 km²/yr, respectively. The lakes in area E (in the southern TP) shrank, and the LA_min and LA_max in area E decreased by 7.59 ± 1.05 km²/yr and 5.80 ± 1.15 km²/yr, respectively. The LAC_inter-year in area B (624.02 ± 131.04 km²) changed drastically, where the LAC_inter-year in area A was 410.08 ± 54.43 km² and the LAC_inter-year in area D changed slightly (i.e., 56.15 ± 20.47 km²) from 2000 to 2018.

The LWL_min and LWL_max increased by 0.17 ± 0.01 m/yr and 0.20 ± 0.01 m/yr, respectively, and the LWLC_inter-year increased by 0.97 ± 0.23 m from 2000 to 2018. The LWL rose faster in area B and area D, where the LWL_min and LWL_max increased by 0.28 ± 0.01 m/yr and 0.31 ± 0.01 m/yr in area B, respectively, and the LWL_min and LWL_max increased by 0.24 ± 0.01 m/yr and 0.27 ± 0.01 m/yr in area D, respectively. The LWL in area E showed a downward trend, and the decrease rates of LWL_min and LWL_max in area E were 0.04 ± 0.01 m/yr and 0.06 ± 0.01 m/yr, respectively. The LWLC_inter-year in area A (0.88 ± 0.14 m), area E (0.89 ± 0.21 m), and area F (0.77 ± 0.70 m) did not differ much, and the LWLC_inter-year in area C (0.60 ± 0.27 m) and area D (0.59 ± 0.19 m) of the TP presented relatively small changes.

The increases in LWSC_min and LWSC_max were 6.32 ± 0.34 Gt/yr and 7.94 ± 0.23 Gt/yr, respectively, and the LWSC_inter-year was 37.48 ± 12.11 Gt from 2000 to 2018. The increases in LWSC in area A and area B were faster from 2000 to 2018. The LWSC_min and LWSC_max in area A were 2.95 ± 031 Gt/yr and 3.34 ± 0.29 Gt/yr, respectively, while those in area B were 1.95 ± 0.08 Gt/yr and 2.40 ± 0.08 Gt/yr, respectively. Although the area of lakes in area B increased the fastest, compared with area A, the LWSC in area A increased at a faster rate than that in area B. The LWSC in area E decreased, where the LWSC_min and LWSC_max in area E were 0.11 ± 0.02 Gt/yr and 0.19 ± 0.03 Gt/yr, respectively. The LWSC_inter-year in the middle of the TP, the southwestern area A of the inner TP, and the north-eastern area B of the inner TP presented relatively large changes. The LWSC_inter-year was high in area A (14.70 ± 2.63 Gt) and area B (12.14 ± 3.20 Gt), while that in area D (1.64 ± 0.59 Gt) was the smallest.

4.5. Climate Change on the TP

We analysed the trends of the main meteorological factors for the TP, including the annual average evaporation, annual average surface temperature, and the annual average precipitation of the TP. The average annual evaporation of the TP decreased from 2001 (525.22 mm) to 2018 (511.05 mm), the overall average annual evaporation in the study area decreased by 17.62 ± 9.40 mm (0.52 ± 0.98 mm/yr), and the annual average evaporation was the highest in 2005 (561.22 mm) and the lowest in 2014 (490.40 mm; Figure 7a). The annual evaporation increased in area D and area F, with increase rates of annual evaporation of 1.70 ± 2.10 mm/yr and 3.10 ± 0.96 mm/yr, respectively. The annual evaporation in area A, area B, area C, and area E decreased continuously, with change rates of annual evaporation of −6.93 ± 3.03 mm/yr, −5.81 ± 3.02 mm/yr, −1.86 ± 3.53 mm/yr, and −2.06 ± 1.18 mm/yr, respectively (Figure 8a). From 2000 to 2018, the annual average surface temperature of the TP showed an overall upward trend, with an overall increase of 0.23 ± 0.57 °C (0.01 ± 0.03 °C/yr) in the study area. During 2000–2006 and 2009–2016, the average surface temperature of the TP increased continuously, and from 2006 to 2019, the average annual surface temperature of the TP changed dramatically. The average annual surface temperature of the TP was the highest in 2006 (−1.69 °C) and the lowest in 2009 (−4.13 °C; Figure 7b). From 2000 to 2018, the average annual surface temperature of the TP mostly showed an upward trend, and the temperature change trend for each sub-region was similar. The average surface temperature of area D rose at the fastest rate (0.05 ± 0.05 °C/yr). The annual average surface temperature of area F rose slowly (with 0.01 ± 0.03 °C/yr), while the average annual surface temperature increase rates of area C, area A, area B, and area E were 0.03 ± 0.06 °C/yr, 0.02 ± 0.06 °C/yr, 0.02 ± 0.04 °C/yr, and 0.01 ± 0.02 °C/yr, respectively (Figure 8b).

From 2000 to 2018, the average annual precipitation of the TP showed an upward trend. The average annual precipitation increased from 2000 (153.70 mm) to 2018 (167.44 mm) and, overall, the average annual precipitation increased by 10.32 ± 6.19 mm (0.54 ± 0.33 mm/yr). The average annual precipitation was the lowest in 2016 (132.50 mm) and the largest in 2018 (167.44 mm) on the TP (Figure 7c). There are obvious spatial differences in the annual precipitation change of the TP from 2000 to 2018. From 2000 to 2018, the annual average precipitation of area A, area C, area D, and area F showed an overall growth trend, with the increase rates of precipitation of 1.57 ± 0.72 mm/yr, 1.23 ± 0.35 mm/yr, 0.62 ± 0.33 mm/yr, and 1.25 ± 0.48 mm/yr, respectively. The annual average precipitation of region B and region E showed a downward trend, with annual precipitation decrease rates of 0.09 ± 0.37 mm/yr and 0.70 ± 0.76 mm/yr, respectively.

5. Discussion

5.1. Effects of Meteorological Factors on LWLC

From 2000 to 2018, most of the lakes on the TP experienced an expanding state, with increasing seasonal variation. The annual precipitation and average annual surface temperature increased, while the average annual evaporation decreased from 2001 to 2018. From 2000 to 2018, the average LWL_min increased by 0.16 ± 0.01 m/yr, the average LWL_max increased by 0.19 ± 0.01 m/yr, and the evaporation of the TP decreased by 0.52 ± 0.98 mm/yr. The contribution of decreased evaporation to the increase in LWL_min was 0.33%, and the contribution to the increase in LWL_max was 0.27%; therefore, it only had a small influence on the expansion of lakes. Previous studies had suggested that increasing precipitation was the primary cause of lake expansion on the TP, followed by increasing glacial meltwater and permafrost meltwater, and the variation of evaporation had little influence [29,30,31,32,33].

We analysed the relationship between the changes in lakes and meteorological factors in different regions, and found that the annual precipitation change rate in area A was the fastest (1.57 ± 0.72 mm). At the same time, the average annual average surface temperature increased by 0.03 ± 0.06 °C/yr. The increasing rate of LWL_min and LWL_max was 0.18 ± 0.02 m/yr and 0.20 ± 0.02 m/yr, respectively, while the evaporation decreased by 6.92 ± 3.03 mm/yr. From the above, it can be calculated that the contribution of the decreased evaporation in area A to the increase in LWL_min was 3.84%, and 3.46% to LWL_max.

The lakes in area B expanded the fastest among those areas, but the precipitation and annual average surface temperature decreased, and the annual evaporation also decreased with decreasing rate of 5.81 ± 3.02 mm/yr. The LWL_min and LWL_max in area B increased by 0.28 ± 0.01 m/yr and 0.31 ± 0.01 m/yr, respectively. The contribution of decreased evaporation to the increase in LWL_min was 2.1%, and its contribution to the increase in LWL_max was 1.9%. Therefore, the primary cause of lake expansion was perhaps the high precipitation, which could offset the loss of evaporation.

The lake expansion rate in area C was slower, the precipitation increased quickly, and the average surface temperature and evaporation decreased. The increasing rate of LWL_min and LWL_max was 0.15 ± 0.01 m/yr and 0.17 ± 0.01 m/yr, respectively, while the evaporation decreased by 1.86 ± 3.53 mm/yr. The contribution of decreased evaporation to the increase in LWL_min was 1.2%, and 1.1% to LWL_max. Therefore, the lake expansion could perhaps be mainly attributed to increasing precipitation.

The number of lakes in area D was small, and the average annual precipitation, average annual surface temperature, and annual evaporation were low. The lakes were also in an expanding state, as a whole. The precipitation and average surface temperature increased, but the evaporation also increased, which would hamper the lake expansion.

The annual precipitation, average annual surface temperature, and annual evaporation in the southern region of the TP (area E) were relatively high, and the lake changes showed an overall shrinking trend. The surface temperature increased, while the evaporation decreased. However, area E was the region with the largest decrease in precipitation, out of all of the sub-regions. The average annual decrease in evaporation was 2.06 ± 1.18 mm/yr, and the decreasing rate of LWL_min and LWL_max was 0.04 ± 0.01 m/yr and 0.06 ± 0.01 mm/yr, respectively, indicating that the decreased evaporation slowed down the shrinkage of the lakes to a certain extent. As such, it can be concluded that the shrinkage of lakes was perhaps mainly caused by the decreased precipitation in area E.

Most of the lakes in area F presented an expanding trend, and the precipitation and evaporation increased, while the average annual temperature decreased. The increasing rate of evaporation was 3.10 ± 0.96 mm/yr, and the increasing rate of LWL_min and LWL_max was 0.08 ± 0.01 m and 0.18 ± 0.02 m, respectively, indicating that the increased evaporation slowed down the expansion of lakes, to a certain extent. It can be concluded that precipitation was the main influencing factor leading to lake expansion in area F.

5.2. Effects of Glacieal Water Supply on LWSC

According to the distribution of glaciers, we analysed the difference between glacier-fed and non-glacier-fed lakes’ LWSC in the inner TP. The watershed of the inner TP was divided into glacier-fed and non-glacier-fed lakes areas based on the results of Qiao et al. (2019) [30]. There were 123 glacier-fed lakes and 81 non-glacier-fed lakes in the inner TP. The distribution of different types of lakes is shown in Figure 9.

As shown in Figure 10, glacier-fed lakes had a greater increase on the LWSC than that of non-glacier-fed lakes. The LWSC_min of glacier-fed lakes increased by 75.94 ± 3.02 Gt from 2000 to 2018, while the LWSC_min of non-glacier-fed lakes only increased by 12.84 ± 0.86 Gt; furthermore, the LWSC_max of glacier-fed lakes increased by 85.89 ± 2.77 Gt, while the LWSC_max of non-glacier-fed lakes only increased by 14.25 ± 0.86 Gt. The increasing LWSC of glacier-fed lakes was approximately six times that of non-glacier-fed lakes.

As shown in

Table 2. The LWSC of glacier-fed lakes and non-glacier-fed lakes in the inner TP from 2000 to 2018.

No.	Type	Basin Area (km²)	Lake Number	Glacier Area (km²)	The LWSCmin (Gt)	The LWSCmax (Gt)
1	Non-glacier-fed	12,839.7	9	/	1.02 ± 0.08	1.30 ± 0.18
2	Non-glacier-fed	10569.7	13	/	0.79 ± 0.09	0.99 ± 0.10
3	Non-glacier-fed	21,964.7	15	/	1.51 ± 0.12	1.72 ± 0.15
4	Non-glacier-fed	8046.4	7	/	0.63 ± 0.03	0.69 ± 0.03
5	Non-glacier-fed	17,773.8	15	/	3.16 ± 0.20	3.39 ± 0.28
6	Non-glacier-fed	23,980.3	16	/	4.76 ± 0.51	5.28 ± 0.57
7	Non-glacier-fed	15,837.8	9	/	0.97 ± 0.05	0.92 ± 0.05
8	Glacier-fed	32,329.2	14	330.7	10.22 ± 0.55	10.58 ± 0.57
9	Glacier-fed	43,197.1	19	366.2	3.90 ± 0.19	4.94 ± 0.33
10	Glacier-fed	23,898.3	8	379.7	7.32 ± 0.36	8.01 ± 0.30
11	Glacier-fed	13,515	7	336.9	5.39 ± 0.23	7.12 ± 0.39
12	Glacier-fed	29,057	5	169	16.85 ± 0.87	18.92 ± 0.96
13	Glacier-fed	4482.6	4	22.6	1.26 ± 0.35	1.32 ± 0.14
14	Glacier-fed	10,730.5	1	180	4.51 ± 0.52	5.58 ± 0.57
15	Glacier-fed	52,534.5	28	2682.4	11.81 ± 0.46	12.46 ± 0.45
16	Glacier-fed	54,781.9	17	259.3	11.62 ± 0.74	12.86 ± 0.76
17	Glacier-fed	56,152.5	17	639.2	3.06 ± 0.36	4.04 ± 0.50

, the LWSC in the No. 12 watershed presented the largest change among those watersheds from 2000 to 2018, the increase of LWSC_min and LWSC_max were 16.85 ± 0.87 Gt and 18.92 ± 0.96 Gt, respectively. There are 28 lakes in the No. 15 basin, and the glacier area in this basin was the largest, with a glacier area of 2682.4 km²; as such, the LWSC_min and LWSC_max had increased by 11.81 ± 0.46 Gt and 12.46 ± 0.45 Gt from 2000 to 2018, respectively. In the No. 14 watershed, the LWSC_min and LWSC_max were 4.51 ± 0.52 Gt and 5.58 ± 0.57 Gt, respectively. However, the LWSC was small overall for non-glacier-fed lakes. Among them, in the No. 2 watershed with 13 lakes, the LWSC_min and LWSC_max were 0.79 ± 0.09 Gt and 0.99 ± 0.10 Gt from 2000 to 2018, respectively. The LWSC_min and LWSC_max were 0.97 ± 0.05 Gt and 0.92 ± 0.05 Gt, respectively, in the No. 7 watershed from 2000 to 2018. There were 16 lakes in No. 6 watershed, the largest watershed area, which had the largest change in LWSC of the areas, where the LWSC_min and LWSC_max were 4.76 ± 0.5 Gt and 5.28 ± 0.57 Gt, respectively. These results suggested that glacier-fed lakes had a much larger increase than that of non-glacier-fed lakes.

As shown in Figure 11, we calculated the LWSC per unit area for each basin from 2000 to 2018. The results showed that the LWSC per unit area of glacier-fed lakes was greater (~four times) than that of non-glacier-fed lakes. As shown in Figure 11, for glacier-fed lakes, the LWSC per unit area in basins No. 12, No. 11, and No. 14 was relatively large. The LWSC_min and LWSC_max per unit area for the No. 12 watershed were 34.26 ± 1.74 mm/yr and 30.52 ± 1.58 mm/yr, respectively. For the No. 11 watershed, the LWSC_min and LWSC_max per unit area were 27.74 ± 1.51 mm/yr and 21.00 ± 0.91 mm/yr, respectively. The LWSC_min and LWSC_max per unit area in the No. 14 watershed were 27.39 ± 2.78 mm/yr and 22.13 ± 2.54 mm/yr, respectively. However, for non-glacier-fed lakes, the LWSC per unit area in basins No. 3 and No. 7 were small, with the LWSC_min and LWSC_max per unit area in basin No. 3 being 4.12 ± 0.35 mm/yr and 3.63 ± 0.29 mm/yr, respectively, while the LWSC_min and LWSC_max per unit area in the No. 7 watershed were 3.04 ± 0.17 mm/yr and 3.23 ± 0.16 mm/yr, respectively. These results indicated that the LWSC with glacier supply was greater than that of non-glacier-fed areas.

Even though the increasing LWSC of glacier-fed lakes was approximately six times that of non-glacier-fed lakes, and the LWSC per unit area of glacier-fed lakes was approximately four times that of non-glacier-fed lakes, the glacier-fed lake basins had much larger lake sizes and many more lakes than those of non-glacier-fed lake basins, which perhaps were much more sensitive to climate change than non-glacier-fed lakes. Chen et al. (2022) suggested that glacial meltwater contributed 9% ± 1% to lake expansion of TP’s endorheic basin during 1995–2020 by quantifying the glacier mass balance [44], and Zhang et al. (2017) suggested that glacial meltwater contributed 13% to lake expansion of TP’s endorheic basin during 2003–2009 [45]. Generally, the average precipitation of those glacier-fed lake basins was much higher than that of non-glacier-fed lake basins [31]. Therefore, we also think that increasing precipitation was considered as the primary cause of lake expansion, and glacial supply was also an important factor affecting LWSC, and the variation of evaporation only had a little influence on lake expansion.

6. Conclusions

Lakes have an important impact on climate change and water cycle in the Tibetan Plateau region. Previous studies have focused on the inter-annual change of lake water level and water storage, but there is a lack of analysis regarding the intra-annual change of these factors over a long-term scale and considering a large number of lakes. Therefore, in this research, we analysed the temporal and spatial variation of 410 lakes (larger than 10 km²), regarding their intra-annual change, on the TP from 2000 to 2018. This further allowed us to analyse the driving mechanism of lake changes on the TP.

Most of the lakes were distributed in the central and northern regions of the TP, and we observed an overall expansion trend from 2000 to 2018. Notably, the average LWL_min increased by 3.09 ± 0.18 m (0.16 ± 0.01 m/yr), the average of LWL_max increased by 3.69 ± 0.12 m (0.19 ± 0.01 m/yr), and the average annual change in LWL was 0.98 ± 0.23 m from 2000 to 2018. The total LWSC_min increased by 125.34 ± 6.79 Gt (6.60 ± 0.36 Gt/yr), the total LWSC_max increased by 158.07 ± 4.52 Gt (8.32 ± 0.24 Gt/yr), and the average of LWSC_intra-year was 40.19 ± 10.67 Gt. The lakes expended the fastest in the southern and northern parts of the TP, while lake water storage in the central part of the TP increased a little. The lakes in the southern part of the TP showed a shrinking trend, with decreasing LWSC.

We analysed the changes in precipitation, temperature, and evaporation for each sub-region of the TP, and we found that the increases in precipitation and surface temperature, along with the decrease in evaporation, have a positive impact on the expansion of lakes on the TP, where precipitation perhaps was the dominant factor affecting lake expansion. We also analysed the LWSC of glacier-fed and non-glacier-fed lakes in the inner TP and found that glacier-fed lakes present more significant changes in LWSC than non-glacier-fed lakes, indicating that glacier meltwater had an important impact on lake change. The variation of evaporation only had a little influence on lake change.