The Dynamic Changes of Lake Issyk-Kul from 1958 to 2020 Based on Multi-Source Satellite Data

Lake Issyk-Kul is the largest alpine lake in arid Central Asia. In recent years, the lake has become a subject of special concern due to the dramatic fluctuations in its water level. In this study, the long-term continuous changes in the water level of Lake Issyk-Kul were derived from hydro-meteorological stations, CryoSat-2, and ICESat-2 satellites. Changes in area were analyzed by the Joint Research Centre (JRC) Global Surface Water (GSW) dataset based on the Google Earth Engine and the variations in water volume were estimated by an empirical formula. The results indicate that the water level of Lake Issyk-Kul fluctuated between 1606.06 m and 1608.32 m during 1958–2020, showing a slight decrease of 0.02 m/year on average. The water level first experienced a significant decreasing trend of 0.05 m/year from 1958 to 1998, and then began to rise rapidly by 0.10 m/year during 1998–2006, followed by a fluctuating decline after 2006. The area of Lake Issyk-Kul exhibited a downward trend before 1998, then a rapid expansion during 1998–2006, and short-term fluctuations in decline thereafter. Meanwhile, changes in water volume of Lake Issyk-Kul followed a similar pattern of variations in water level and area. According to comprehensive analyses, the continuous downward trend of the water level before 1998 was primarily affected by substantial anthropogenic water consumption in the basin. However, since the 21st century, the increases in precipitation and glacier meltwater and the reduced water consumption have collectively facilitated the short-term recovery of Lake Issyk-Kul in water level, area, and water volume.


Introduction
As essential components of global hydrological and biogeochemical water cycles, lakes are extremely susceptible to climate and environmental changes. Endorheic lakes located in arid and semi-arid environments are highly sensitive to climate oscillation [1][2][3], so they serve as crucial indicators of regional climate change and also play indispensable roles in maintaining the regional ecological balance [4][5][6].
The acquisition of long-term gauge observations is restricted by the complex topography, the varying weather conditions, data confidentiality, and data incompleteness [7,8]. Satellite altimetry, an emerging water level determination approach to overcome the limitations of in situ observations, has been widely used by researchers to obtain continuous water levels due to its precise periodic operations and large spatial coverage [9]. Some of the most commonly used altimetric satellites in current research include ERS-1/2, Envisat, GFO, Jason-1/2, ICESat-1/2, CryoSat-2, etc. A number of studies have demonstrated the usefulness of satellite altimetry data for small lakes in ungauged regions and larger lakes such as typical lakes on the Tibetan Plateau [3,[10][11][12][13][14][15][16]. The Geoscience Laser Altimeter The climatic conditions in the Issyk-Kul Lake Basin are mostly continental with a mild and dry climate throughout the year. The average temperature is 19-20 °C in July and 2-3 °C in January [51]. The annual total precipitation varies dramatically over different regions of the basin, increasing from west to east [52]. For example, the annual precipitation ranges 255-258 mm in the central area of the valley and 300-400 mm in the eastern valley [53]. In alpine regions, the annual precipitation increases from 300-400 mm in the east to 800-900 mm in the west [23].

Lake Level Dataset
The lake level time series data (m above mean sea level) of Lake Issyk-Kul were obtained from in situ data, CryoSat-2, and ICESat-2. The main characteristics of the two satellite missions are shown in Table 1 [54][55][56]. CryoSat-2 is one of the missions of the Earth Exploration Program developed by the European Space Agency (ESA). SIRAL is the primary instrument on-board CryoSat-2, which operates in three modes, i.e., low-resolution mode (LRM), synthetic aperture mode (SAR), and SAR interferometric mode (SARIn). Cryosat-2/SIRAL contains four data products. The Level 2 GDR product includes measurement time, latitude, longitude, and height information that can be downloaded from EO-CAT, a Next Generation Earth Observation system (https://eocat.esa.int/sec/#data-services-area, access date: 18 October 2021) [56]. In the latest updated GDR product, the LRM mode data applied 3 waveform retracking algorithms: Refined CFI, UCL, and Refined The climatic conditions in the Issyk-Kul Lake Basin are mostly continental with a mild and dry climate throughout the year. Based on the meteorological observation data, the average temperature is 19-20 • C in July and 2-3 • C in January [51]. The annual total precipitation varies dramatically over different regions of the basin, increasing from west to east [52]. For example, the annual precipitation ranges 255-258 mm in the central area of the valley and 300-400 mm in the eastern valley [53]. In alpine regions, the annual precipitation increases from 300-400 mm in the east to 800-900 mm in the west [23].

Lake Level Dataset
The lake level time series data (m above mean sea level) of Lake Issyk-Kul were obtained from in situ data, CryoSat-2, and ICESat-2. The main characteristics of the two satellite missions are shown in Table 1 [54][55][56]. CryoSat-2 is one of the missions of the Earth Exploration Program developed by the European Space Agency (ESA). SIRAL is the primary instrument on-board CryoSat-2, which operates in three modes, i.e., low-resolution mode (LRM), synthetic aperture mode (SAR), and SAR interferometric mode (SARIn). Cryosat-2/SIRAL contains four data products. The Level 2 GDR product includes measurement time, latitude, longitude, and height information that can be downloaded from EO-CAT, a Next Generation Earth Observation system (https://eocat.esa.int/sec/#data-services-area, accessed on 18 October 2021) [56]. In the latest updated GDR product, the LRM mode data applied 3 waveform retracking algorithms: Refined CFI, UCL, and Refined OCOG GDR with a footprint diameter of~1.65 km on the water surface and an interval of~300 m [55]. By comparing the number of altitude outliers derived by the three algorithms in Lake Issyk-Kul, the Refined OCOG, which is a tracking method developed by the Mullard Space Science Laboratory that uses an estimate of the pulse width to track the return waveform, was finally used to estimate lake level changes of Lake Issyk-Kul from 2010 to 2020 in this study [57]. ICESat-2, launched in September 2018, continues long-term elevation observations begun by the original ICESat-1 mission with the enhanced Advanced Topographic Laser Altimeter System (ATLAS). The diameter of laser footprints is~17 m and interleaved bỹ 0.7 m along the track [54]. The two-beam instrument setting and smaller footprint help it achieve more accurate measurements. Among the data products of ICESat-2, ATL13 is a dataset specifically used for inland water surface height determination; data from 13 October 2018 to 1 January 2020 are available from the National Snow and Ice Data Center (NSIDC; https://nsidc.org/data/icesat-2/products/, accessed on 18 October 2021) [58]. This study applies height information from ATL13 to obtain the lake level data from 2018 to 2020, which can be used for the accuracy validation of overlapping time periods with data derived from Cryosat-2.
The in situ data used in this study was obtained from HYDROLARE (http://hydrolare. net/data_availability.php, accessed on 18 October 2021), which provides the water level data of Lake Issyk-Kul updated to 2017, with missing data for 1998 and 1999. The composition of water level data is shown in Figure 2.
October 2021) [56]. In the latest updated GDR product, the LRM mode data applied 3 waveform retracking algorithms: Refined CFI, UCL, and Refined OCOG GDR with a footprint diameter of ~1.65 km on the water surface and an interval of ~300 m [55]. By comparing the number of altitude outliers derived by the three algorithms in Lake Issyk-Kul, the Refined OCOG, which is a tracking method developed by the Mullard Space Science Laboratory that uses an estimate of the pulse width to track the return waveform, was finally used to estimate lake level changes of Lake Issyk-Kul from 2010 to 2020 in this study [57]. ICESat-2, launched in September 2018, continues long-term elevation observations begun by the original ICESat-1 mission with the enhanced Advanced Topographic Laser Altimeter System (ATLAS). The diameter of laser footprints is ~17 m and interleaved by ~0.7 m along the track [54]. The two-beam instrument setting and smaller footprint help it achieve more accurate measurements. Among the data products of ICESat-2, ATL13 is a dataset specifically used for inland water surface height determination; data from 13 October 2018 to 1 January 2020 are available from the National Snow and Ice Data Center (NSIDC; https://nsidc.org/data/icesat-2/products/, accessed on 18 October 2021) [58]. This study applies height information from ATL13 to obtain the lake level data from 2018 to 2020, which can be used for the accuracy validation of overlapping time periods with data derived from Cryosat-2.
The in situ data used in this study was obtained from HYDROLARE (http://hydrolare.net/data_availability.php, accessed on 18 October 2021), which provides the water level data of Lake Issyk-Kul updated to 2017, with missing data for 1998 and 1999. The composition of water level data is shown in Figure 2.

Lake Area Dataset
The lake area time-series data of Lake Issyk-Kul were derived from JRC Monthly Table 2. The JRC dataset shows the temporal and spatial changes of surface water over the years; every pixel is discriminated as water or non-water one by one through the expert system, and the final results are recorded as monthly historical data for the entire time period between 16 March 1984 and 31 December 2020. The Landsat datasets have been atmospherically corrected using the Land Surface Reflectance Code (LaSRC) and the Landsat ecosystem disturbance adaptive processing system (LEDAPS). Moreover, these data were also produced using the C Function of Mask (CFMASK) to improve the data quality [59].

Meteorological Data
The meteorological stations in the Issyk-Kul basin were established at different times and are located at different elevations and latitudes, with some stations changing location several times [24]. Each meteorological station provides observations for different periods of time, which makes it difficult to obtain comprehensive long-term continuous meteorological observations for the entire basin. Therefore, the annual temperature, annual total precipitation, and annual total evaporation during 1958-2020 were derived from ERA5-Land monthly averaged data (https://www.ecmwf.int/en/forecasts/datasets/reanalysisdatasets/era5, accessed on 18 October 2021). ERA5-Land is a reanalysis dataset providing a consistent view of the evolution of land variables over several decades at an enhanced

Glacier Data
The glacier data of the Issyk-Kul Lake Basin in 2002 were obtained from the Randolph Glacier Inventory (RGI) 6.0 published by Global Land Ice Measurements from Space (GLIMS, http://www.glims.org/RGI/, accessed on 18 October 2021) [63]. The glacier data in 2015 were derived in this study. There are numerous debris-covered glaciers distributed in the Issyk-Kul Lake Basin, which are difficult to identify through automatic classification. Therefore, the glaciers in 2015 were manually delineated based on the Landsat images, and the unrecognizable parts were identified by higher-resolution Google Earth images. Afterward, the glacier area was calculated on the WGS84 spheroid in an Albers equalarea map projection. The glacier volume was estimated by the volume-area (V-A) scaling method based on empirical parameters proposed by Radić and Hock (2010) [64].

Other Relevant Data
The extent of the water body used in this study was obtained from Global Lakes and Wetlands Database (GLWD). The cropland data of 2020 were from the Globalland30

Methods
The methods of both lake level and area extraction and the volume estimation in this study are shown in Figure 3.
equal-area map projection. The glacier volume was estimated by the volume-area (V-A) scaling method based on empirical parameters proposed by Radić and Hock (2010) [64].

Other Relevant Data
The extent of the water body used in this study was obtained from Global Lakes and Wetlands Database (GLWD). The cropland data of 2020 were from the Globalland30 dataset published by the Ministry of Natural Resources of the People's Republic of China (http://www.globallandcover.com/, accessed on 18 October 2021).

Methods
The methods of both lake level and area extraction and the volume estimation in this study are shown in Figure 3. The scattering or refraction of the CryoSat-2 satellite radar pulse signal in the propagation process will affect its propagation speed and delay the round-trip time of the observation signal. Moreover, affected by natural factors, the distance estimated may be biased, therefore, various deviations need to be corrected. The calculation formula of the lake level is as follows: where is the orthometric height based on the EGM96 geoid, is the height from the satellite's centroid to the WGS-84 reference ellipsoid, is the distance from the satellite to the lake surface, is the correction, and is the difference between the local geoid and the EGM96 geoid, which can be calculated by the geoid height function in MATLAB.
Compared with ICESat-1, the reference ellipsoid of ICESat-2 has changed from T/P to WGS84, so its calculation formula is basically the same as CryoSat-2, and the elevation measurement results can be directly compared with CryoSat-2. Satellite altimetry data pre-processing The scattering or refraction of the CryoSat-2 satellite radar pulse signal in the propagation process will affect its propagation speed and delay the round-trip time of the observation signal. Moreover, affected by natural factors, the distance estimated may be biased, therefore, various deviations need to be corrected. The calculation formula of the lake level is as follows: where H is the orthometric height based on the EGM96 geoid, H alt is the height from the satellite's centroid to the WGS-84 reference ellipsoid, R is the distance from the satellite to the lake surface, c is the correction, and N is the difference between the local geoid and the EGM96 geoid, which can be calculated by the geoid height function in MATLAB. Compared with ICESat-1, the reference ellipsoid of ICESat-2 has changed from T/P to WGS84, so its calculation formula is basically the same as CryoSat-2, and the elevation measurement results can be directly compared with CryoSat-2.

2.
Outlier removal To ensure the validity of the data, it is necessary to filter out all the outliers of the water level to extract the time series. The specific removal steps are as follows: First, a 200 m buffer zone from the lake boundary to the center of the lake was established, and satellite footprint points were filtered based on the buffer zone to make sure the points completely fall into the lake, thereby avoiding interference from elevation points that may touch the lake shore.
Second, all points were manually filtered to eliminate outliers that are tens or even hundreds of meters away from most water level values ( Figure 4).
First, a 200 m buffer zone from the lake boundary to the center of the lake was established, and satellite footprint points were filtered based on the buffer zone to make sure the points completely fall into the lake, thereby avoiding interference from elevation points that may touch the lake shore.
Second, all points were manually filtered to eliminate outliers that are tens or even hundreds of meters away from most water level values (Figure 4).  Third, the 3σ criterion was adopted to eliminate the outliers in the data within a day, and then the remaining effective values were averaged as the daily lake level. The results are shown in Figure 5   Third, the 3σ criterion was adopted to eliminate the outliers in the data within a day, and then the remaining effective values were averaged as the daily lake level. The results are shown in Figure 5, taking Cysosat-2 data on 13 August 2017 ( Figure 5b) and ICESat-2 data on 2 March 2019 ( Figure 5c) as examples. The usability of the 3σ criterion has been proven in the application of large lakes with enough footprints [65].

Modification Notes
The following is a description of the modified part: (1) In the second paragraph of section 2.1, "The average temperature is 19-20 °C in July and 2-3 °C in January" need to be replaced with "Based on the meteorological observation data, the average temperature is 19-20 °C in July and 2-3 °C in January".
(2) The Figure 5 need to be replaced with the following figure. (3) The Figure 7 need to be replaced with the following figure. Finally, the overall screening of all daily water levels was operated by the 3σ criterion to further obtain the monthly average water level and the annual average water level.

3.
Multi-source water level data merging As shown in Figure 3, the time series of water levels are synthesized by in situ data, CryoSat-2 products, and ICESat-2 missions. However, as the satellite altimetry data and the in situ hydrological data are referenced to WGS-84 and Baltic systems, respectively, the conversion of the elevation system must be completed first. According to the average difference between the water level obtained by CryoSat-2 and the in situ observations during the overlapping time periods, the altimetric data were converted into the same reference system as in situ data, thereby obtaining the lake level of Lake Issyk-Kul from 1958 to 2020.

Lake Area Extraction
Considering that the water level of Lake Issyk-Kul was at a relatively stable high value in August during the year, we selected remote sensing images of August from the JRC datasets based on the GEE to analyze long-term area variation patterns of Lake Issyk-Kul in the past three decades. For the years when the area in August cannot be obtained due to poor image quality, the images of the adjacent months were utilized as substitutes. The lake area extracted automatically is affected by some redundant elements such as rivers and beaches, so the lake boundaries derived from remote sensing images were further manually modified to improve accuracy.

Accuracy Verification
The assessment of water level data contains absolute verification and relative verification. The absolute verification refers to verifying CryoSat-2 data with in situ data, while relative verification means evaluating CryoSat-2 data through higher-precision ICESat-2 data. Before the accuracy evaluation, the CryoSat-2 data and ICESat-2 data have been converted by the method introduced in Section 2. Moreover, we used a modified normalized difference water index (mNDWI) and a maximum between-cluster variance (OTSU) image segmentation algorithm to extract the lake area from the Landsat datasets based on the Google Earth Engine. The results were used as validation data to evaluate the accuracy of the area extracted by the JRC datasets. The mNDWI has been widely used to discriminate water bodies from other land surfaces [66]. It can be expressed as: where ρ MIR and ρ Green represent the middle-infrared and green bands, respectively.

Estimation of Water Volume Change
Changes in water volume are determined by both lake area and water level. In this study, we used an estimation equation to calculate the variations in water volume [67][68][69]. The estimation equation is as follows: where ∆V represents the water volume change, ∆H represents the lake level change during the two periods, and S 1 and S 2 represent the lake area for two periods.

Variations in the Water Level of Lake Issyk-Kul
Based on the in situ observations, the CyroSat-2 dataset, and the ICESat-2 dataset, the temporal variations in the water level of Lake Issyk-Kul over the period of 1958 to 2020 are shown in Figure 6. The Mann-Kendall trend test was used to determine whether a time series has a monotonic upward or downward trend with a p-value lower than 0.01 (<0.01) being statistically significant. The water level of Lake Issyk-Kul dropped from 1608.08 to 1606. 55

Variations in the Water Level of Lake Issyk-Kul
Based on the in situ observations, the CyroSat-2 dataset, and the ICESat-2 dat temporal variations in the water level of Lake Issyk-Kul over the period of 1958 are shown in Figure 6. The Mann-Kendall trend test was used to determine w time series has a monotonic upward or downward trend with a p-value lower t (<0.01) being statistically significant. The water level of Lake Issyk-Kul dropp 1608.08 to 1606. 55   The intra-annual variations in the water level of Lake Issyk-Kul indicated r regular seasonal variabilities (Figure 6b,c). The water level was relatively stable f uary to March, and then gradually increased, reaching the maximum of 1607.09 m The intra-annual variations in the water level of Lake Issyk-Kul indicated relatively regular seasonal variabilities (Figure 6b,c). The water level was relatively stable from January to March, and then gradually increased, reaching the maximum of 1607.09 m in August. Subsequently, the water level exhibited a downward trend until the monthly average value dropped to 1606.93 m in November. In general, the rising trend during the year mainly occurred from May to August, and the lake level fluctuated more drastically in the warm season (April to October) than in the cold season (November to March of the following year).

Variations in Area of Lake Issyk-Kul
According to the relationship between the SRTM DEM and the in situ water level, it can be inferred that the lake area decreased from 6242.85 km 2 to 6217.62 km 2 during 1958-1990. Owning to the poor quality of satellite images in the 1980s and the inability to fully cover the study area, only part of the effective area data after 1990 can be obtained through Landsat datasets, as shown in Figure 7. From 1990 to 2020, the lake area declined slightly from 6217.62 km 2 to 6214.86 km 2 at a rate of −0.09 km 2 /year. It stands out that the smallest lake area appeared in 1998 (6204.71 km 2 ), becoming a turning point in the trend of lake area fluctuation, which was consistent with the findings of previous studies [31]. Between 1990 and 1998, the area of Lake Issyk-Kul reduced dramatically by 1.61 km 2 /year (p < 0.01) for a total of 12.91 km 2 , which means the lake shrank by about 0.21% during this period. From 1998 to 2006, the lake area expanded from 6204.71 km 2 to 6219.51 km 2 at a rate of 1.84 km 2 /year (p < 0.01). After 2006, the lake area entered a period of shrinkage at an average rate of 0.33 km 2 /year, accompanied by several short-term fluctuations.

Variations in the Water Volume of Lake Issyk-Kul
Based on the past area changes and the Equation (3), it was estimated that the water volume of Lake Issyk-Kul dropped by about 9.57 km 3 during 1958-1990, and the specific variations after 1990 were pre-sented in Figure 8.

Accuracy Verification
In this section, we first evaluate the accuracy of the lake level data derived from Cry-oSat-2, and then the accuracy of the water area data extracted from the JRC dataset is explored systematically.

Accuracy Verification of Water Level
As shown in Figure 9a, the CryoSat-2 data are significantly correlated with in situ data. The Pearson correlation coefficient (CC) is 0.919 (Figure 9b), and the mean absolute error (MAE) and root-mean-square error (RMSE) are 0.060 m and 0.075 m, respectively. There is also a good match between CryoSat-2 data and ICESat-2 data (Figure 9c), with a CC of 0.960 (Figure 9d), an MAE of 0.018 m, and an RMSE of 0.020 m. In short, the combination of in situ observations, CryoSat-2 datasets, and ICESat-2 datasets can provide comprehensive and long-term information to monitor the water-level variations of Lake Issyk-Kul. The specific evaluation metrics are listed in Table 3. Several researchers have used ICESat-2 or CryoSat-2 to monitor water levels in various types of lakes, and the results showed that both satellites have great potential to accurately determine lake level changes [3,58,[70][71][72]. Our study further demonstrates that combining the two satellites can provide a more accurate exploration of lake levels with higher consistency with in situ data.

Accuracy Verification
In this section, we first evaluate the accuracy of the lake level data derived from CryoSat-2, and then the accuracy of the water area data extracted from the JRC dataset is explored systematically.

Accuracy Verification of Water Level
As shown in Figure 9a, the CryoSat-2 data are significantly correlated with in situ data. The Pearson correlation coefficient (CC) is 0.919 (Figure 9b), and the mean absolute error (MAE) and root-mean-square error (RMSE) are 0.060 m and 0.075 m, respectively. There is also a good match between CryoSat-2 data and ICESat-2 data (Figure 9c), with a CC of 0.960 (Figure 9d), an MAE of 0.018 m, and an RMSE of 0.020 m. In short, the combination of in situ observations, CryoSat-2 datasets, and ICESat-2 datasets can provide comprehensive and long-term information to monitor the water-level variations of Lake Issyk-Kul. The specific evaluation metrics are listed in Table 3. Several researchers have used ICESat-2 or CryoSat-2 to monitor water levels in various types of lakes, and the results showed that both satellites have great potential to accurately determine lake level changes [3,58,[70][71][72].
Our study further demonstrates that combining the two satellites can provide a more accurate exploration of lake levels with higher consistency with in situ data.

Accuracy Verification of Water Area
The comparison between the extraction results of the JRC dataset and the Landsat dataset using the mNDWI and OTSU algorithm is shown in Figure 10. The RMSE of the lake area derived from the JRC dataset was 0.518 km 2 with a CC of 0.990 and an MAE of 0.470 km 2 , further demonstrating the feasibility of obtaining long-term area data of Lake Issyk-Kul from the JRC dataset.

. Accuracy Verification of Water Area
The comparison between the extraction results of the JRC dataset and the Landsat dataset using the mNDWI and OTSU algorithm is shown in Figure 10. The RMSE of the lake area derived from the JRC dataset was 0.518 km 2 with a CC of 0.990 and an MAE of 0.470 km 2 , further demonstrating the feasibility of obtaining long-term area data of Lake Issyk-Kul from the JRC dataset.

The Driving Force of the Dynamic Changes of Lake Issyk-Kul
In order to explore the potential driving factors of the long-term dynamic changes of Lake Issyk-Kul, we further analyzed the lake's response to climate change, glacier meltwater, and human activities in the basin.

Climate Change
Temperature, precipitation, and evaporation as specific climatic characterizations are critical factors affecting the variations of the lake [24]. We used a sequential Mann-Kendall (M-K) mutation test that estimates progressive (UF) and backward (UB) series to detect abrupt variation years in the study series [73]. From 1958 to 2020, the mean temperature in the Issyk-Kul Lake Basin rose at a rate of 0.13 • C/10 year and the average annual air temperature was −0.72 • C (Figure 11a). The results of the M-K test (Figure 11b) showed that the upward trend of temperature became more evident since 2005, with a confidence level of more than 95%. Higher temperatures lead to an increase in glacial ablation, which may contribute to an increase in lake recharge and a rise in the lake level. Remote Sens. 2022, 14, x FOR PEER REVIEW 13 of 19 Figure 10. The comparison between the extraction results of the JRC dataset and the Landsat dataset using the mNDWI and OTSU algorithm.

The Driving Force of the Dynamic Changes of Lake Issyk-Kul
In order to explore the potential driving factors of the long-term dynamic changes of Lake Issyk-Kul, we further analyzed the lake's response to climate change, glacier meltwater, and human activities in the basin.

Climate Change
Temperature, precipitation, and evaporation as specific climatic characterizations are critical factors affecting the variations of the lake [24]. We used a sequential Mann-Kendall (M-K) mutation test that estimates progressive (UF) and backward (UB) series to detect abrupt variation years in the study series [73]. From 1958 to 2020, the mean temperature in the Issyk-Kul Lake Basin rose at a rate of 0.01 °C/year and the average annual air temperature was −0.72 °C (Figure 11a). The results of the M-K test (Figure 11b) showed that the upward trend of temperature became more evident since 2005, with a confidence level of more than 95%. Higher temperatures lead to an increase in glacial ablation, which may contribute to an increase in lake recharge and a rise in the lake level. The precipitation fluctuated considerably during the study period (Figure 11c), and the M-K test showed multiple intersections of the UF and UB curves, none of which exceed the 0.05 significance level threshold (Figure 11d), which means that the overall upward trend in precipitation from 1958 to 2020 was not significant. The evaporation over the basin indicated an upward trend from 1958 to 2020, and the significant increase in evaporation occurred after 2013 (Figure 11e,f).
Since the water level of Lake Issyk-Kul increased significantly after 1998, the variations in precipitation and evaporation during different time periods, with 1998 as the turning point, were analyzed. From 1958 to 1998, the precipitation and evaporation exhibited slight increases and decreases, respectively, which should have contributed to the rise of water level. However, the water level dropped significantly during this period, indicating that climatic factors were not the major factors driving the dynamics of the lake before 1998. From 1998 to 2020, both the precipitation and evaporation presented increasing trends, and the lake level, area, and water volume have gradually recovered since 1998. The Pearson correlation coefficient value showed that the water volume was highly correlated with precipitation during this period (CC = 0.515, p < 0.01). These facts signified that the precipitation contributed more to the dynamics of Lake Issyk-Kul than evaporation in the early 21st century.

Glacier Meltwater
Glacier meltwater is the predominant source of replenishment for rivers in the arid regions of Central Asia, which can indirectly affect the water level, area, and volume of lakes [74]. Based on RGI 6.0 and the glacier data derived in this study, we found that the glacier area in the Issyk-Kul Lake Basin reduced from 520.09 km 2 to 458.04 km 2 at an average rate of 4.77 km 2 /year with a loss of around 62.05 km 2 , which means the glacier area shrank by about 11.93% during 2002-2015. The glacier volume decreased by approximately 2.63 km 3 (11.63%) over the period of 2002-2015, equivalent to the loss of glacier mass of 2.23 ± 0.16 km 3 w.e. (water equivalent), by assuming a volume to mass conversion factor of 850 ± 60 kg/m 3 . The water volume of Lake Issyk-Kul increased by about 0.64 km 3 over the same period, which was lower than the loss of glacier mass. This fact suggests that the loss of glacier mass in the Issyk-Kul basin may contribute to the increase in the water volume of the lake during 2002-2015. Therefore, glacier meltwater is a critical factor influencing the variations of Lake Issyk-Kul, especially during 2002-2015. The precipitation fluctuated considerably during the study period (Figure 11c), and the M-K test showed multiple intersections of the UF and UB curves, none of which exceed the 0.05 significance level threshold (Figure 11d), which means that the overall upward trend in precipitation from 1958 to 2020 was not significant. The evaporation over the basin indicated an upward trend from 1958 to 2020, and the significant increase in evaporation occurred after 2013 (Figure 11e,f).
Since the water level of Lake Issyk-Kul increased significantly after 1998, the variations in precipitation and evaporation during different time periods, with 1998 as the turning point, were analyzed. From 1958 to 1998, the precipitation and evaporation exhibited slight increases and decreases, respectively, which should have contributed to the rise of water level. However, the water level dropped significantly during this period, indicating that climatic factors were not the major factors driving the dynamics of the lake before 1998. From 1998 to 2020, both the precipitation and evaporation presented increasing trends, and the lake level, area, and water volume have gradually recovered since 1998. The Pearson correlation coefficient value showed that the water volume was highly correlated with precipitation during this period (CC = 0.515, p < 0.01). These facts signified

Human Activities
In the 1950s, as the population in the lake basin grew, consequently, a large amount of wasteland was reclaimed. The irrigated area expanded rapidly from around 500 km 2 in 1930 to 1540 km 2 in 1984, which led to the increase in water demand. During 1935-1971, the average river water usage was about 9.4 × 10 8 m 3 , which then increased to around 13.7 × 10 8 m 3 in the period of 1972-1984 [75]. Thus, it was the uptake of water for human activities such as irrigation, industry, and urban domestic life that seemed to be the major factor contributing to the dramatic decline in water levels during this period. Recent lake water consumers are shown in Table 4 [26]. Obviously, over the past few decades, agriculture has always been the highest water consumer compared to manufacturing and household water consumption. The water usage dedicated to human activities after 2000 had been reduced drastically compared with the 20th century, which is consistent with the water level and water area recovery period since 1998. The reduction in water consumption may be due to the decrease in water users in the Issyk-Kul Lake Basin after the collapse of the Soviet Union in 1989-1992, with some residents being deployed to different locations, which meant that they used other water sources [26,76]. Accordingly, human activities are the crucial factors responsible for the long-term dynamic changes of Lake Issyk-Kul. Table 4. The water consumption for agriculture, manufacturing, and household uses in the Issyk-Kul Lake Basin from 1980 to 2014 [26].

Period
Agriculture ( In short, the continuous downward trend of the water level before 1998 is related to the excessive water consumption for human activities in the entire basin. Since the 21st century, the growing precipitation, increased glacier meltwater induced by significantly increasing temperature, as well as reduced water consumption collectively contributed to the restoration of the water level, area, and water volume of Lake Issyk-Kul.

Conclusions
The combination of multisource data proves to be a feasible way to detect variations in Lake Issyk-Kul. According to the monitoring results of in situ data, CryoSat-2, and ICESat-2, the water level of Lake Issyk-Kul declined rapidly from 1958 to 1998, with an average rate of 0.05 m/year. Since 1998, the water level climbed generally, accompanied by several fluctuations. The water level presented relatively regular seasonal changes, which is manifested in how the multi-year average water level reached the highest in spring and the lowest in autumn. The increase mainly occurred from May to August and the fluctuation in the warm season was more severe than that in the cold season. During the period of 1990-2020, the area of Lake Issyk-Kul shrank by around 2.76 km 2 at a rate of 0.09 km 2 /year. Meanwhile, the water volume of Lake Issyk-Kul showed a decreasing trend before 1998, followed by a significant increase from 1998-2006, and then a small fluctuating decrease.
Over the years, the large amount of water withdrawal from lake tributaries for human activities in the basin has been cited as being responsible for the dynamic changes of Lake Issyk-Kul before 1998. However, the recovery of the water level, area, and water volume since 1998 was affected by increased precipitation, glacier meltwater resulting from rising temperature, and decreased water consumption in the early 21st century. Owing to the complexity and uncertainty of environmental changes, the quantitative discussion of the impact of various factors on lake dynamic changes requires further research.