A comparative study of satellite altimetry-based and DEM-based methods for estimating lake water volume changes

Lakes play an important role in the global water cycle and act as key indicators of climate change. China has more than 2,000 lakes (>1 km²), cumulatively covering an area exceeding 80,000 km² (Zhang & Song 2022). These lakes show significant spatial heterogeneity in terms of water volume change, hydrological regimes, and climate interactions. Considering the spatial characteristics of natural environments and resource utilization, the geographic distribution of Chinese lakes has been categorized into five distinct regions: the Tibetan Plateau (TP), Eastern China Plain, Mongolia-Xinjiang, Northeastern China Plain, and Yunnan-Guizhou Lake Region (Wang & Dou 1998). Among these, the TP and Mongolian Regions, which together constitute more than 70% of China, have the largest lakes with individual lakes exceeding 20 km² in area (Liu et al. 2022). In particular, the TP has the largest number of lakes in China with more than 1,000 lakes in areas larger than 1 km² (Zhang 2018). The Xinjiang and the Eastern China Plains occupy 9.1 and 21.73% of the national total lake area, respectively (Shang et al. 2021).

The spatial variability in lake dynamics, increasingly influenced by both climate change and anthropogenic activities, has become more pronounced in recent years (Zhang et al. 2013). In the Mongolia-Xinjiang Lake Region, lakes such as Boston Lake have shrunk due to an increase in irrigation for agriculture (Liu et al. 2006). In the Eastern China Plains, especially in the Yangtze River Basin (such as Tai Lake, Poyang Lake), lakes are subject to significant hydrological alterations driven by monsoonal shifts and human interventions (Guo et al. 2008; Dronova et al. 2011; Ji et al. 2019). However, the TP alpine lakes are generally expanding, a trend attributed to accelerated glacial melt and increased precipitation (Zhang et al. 2011).

The water balance of a lake refers to the net effect of inputs (such as surface runoff inflow and precipitation) against outputs (such as surface runoff outflow and lake evaporation) (Zhang et al. 2013). In remote areas with harsh conditions, these variables are often not directly measurable at hydrological stations. Remote sensing technology provides a solution, enabling the estimation of lake water balance through the analysis of satellite-derived water level (from radar or laser altimeters) combined with surface area data obtained from satellite imagery (Zhang et al. 2013; Duan et al. 2018). The advent of advanced remote sensing techniques has facilitated the development of multiple methods for estimating lake water volumes, typically incorporating data from individual satellites to estimate lake levels and areas, or integrating data from multiple remote sensing sources to improve temporal and spatial coverages (Zhang et al. 2011; Duan & Bastiaanssen 2013; Feng et al. 2022). For example, Zhang et al. (2013) derived lake levels from ICESat-1 altimetry data and lake surface area from Landsat images and further used them to estimate the water balances of the 10 largest lakes in China during 2003–2009. Xu et al. (2022) used multi-mission satellite data to systematically analyze water storage changes in Inner Mongolia's major lakes.

Existing studies usually use empirical formulas to calculate the water volume change in a single area based on lake area and water level (Crétaux et al. 2016; Cai et al. 2020). The method of integrating digital elevation model (DEM) and satellite optical images is also an important technical supplement for estimating water volume change, and this method has acceptable accuracy with an overall average error of 4.98% (Yang et al. 2017). Most existing studies have relied on a single estimation method, with comparative analyses between different methods remaining scarce. Satellite altimetry-based estimation methods have been widely used, particularly for large lakes with comprehensive data sets such as Selin Co and Qinghai Lake. Satellite altimeters are profiling tools rather than imaging devices, and they can only observe lakes along their ground tracks, meaning that many lakes are not or only partially measured. Thus, questions remain regarding the reliability of water volume change estimates for lakes with incomplete data records. The DEM-based method has been well applied in the TP region (Yang et al. 2017), its suitability for lakes in other regions has not been sufficiently investigated. Moreover, comparative analyses of the two different methods are even more scarce.

To address the above-mentioned gaps, this study aimed at conducting a comparative study of satellite altimetry-based and DEM-based methods for estimating lake water volume changes in China. Specifically, we leveraged satellite altimetry-derived lake water levels provided by the Database for Hydrological Time Series of Inland Waters (DAHITI) in combination with Landsat-derived lake areas to estimate the water volume changes for 34 selected lakes from 2005 to 2020. For the DEM-based method, we used the Shuttle Radar Topography Mission (SRTM) DEM and Landsat images to estimate water volume changes in the same lakes during the same period. The objectives of this study are to (1) evaluate the two methods and examine their strengths and limitations for estimating lake water volume changes and (2) assess the regional applicability of the two methods across a large geographical region. This study is expected to help us achieve a better understanding of the performance and applicability of these two methods and potentially to synthesize these two methods for an improved estimate of lake water volume changes.

Most of the 27 lakes on the TP are located in the endorheic region and are less affected by human activity; however, they are extremely sensitive to climatic and cryospheric changes. The TP is a regional Earth system showing very strong interactions among its lithosphere, hydrosphere, cryosphere, biosphere, atmosphere, and anthrosphere (Yang et al. 2011; Yao et al. 2015). The inland lakes in the arid region of Xinjiang are an important link in the water resource cycle and a comprehensive reflection of the effect of climate change and human activities on water resources (Wu et al. 2012). Lake volumes in the Eastern China Plain Lake Region show significant inter- and intra-annual variations caused by precipitation fluctuations during the summer monsoon. This phenomenon is more significant in the mid-lower reaches of the Yangtze River basin (Chen et al. 2011; Zheng et al. 2016).

The SRTM DEM is a near-global topographic database generated from satellite images (Jakob & Van 2001) at a resolution of 30 m (Rabus et al. 2003). The SRTM provides a global high-quality DEM which covers the Earth between latitudes 60°N and 56°S and is acquired with the same sensor in a single mission and produced with single-technique synthetic aperture radar (SAR) interferometry (Jakob & Van 2001). A great advantage is the homogeneous quality of the DEM (Rabus et al. 2003). Yang et al. (2017) assessed the accuracy of lake water volume change results calculated using the SRTM DEM and found the accuracy of this method acceptable. Therefore, these two methods in this study were not necessary to verify the accuracy.

First, the lake area A_SRTM was extracted from the SRTM DEM data and the corresponding lake surface area A_i (i = 1, 2, 3 …) at every 1 m increment in elevation. Therefore, A_i (i = 1, 2, 3 …) at an elevation of i m above A_SRTM was calculated. Second, using Equation (3) to calculate V_i which expresses the volume between the lake surface A_SRTM and A_i – the volume change when the lake surface rises from A_SRTM to A_i – we obtained a series of (V_i, A_i) data pairs and established a regression equation (Equation (4)) between them – V_i is estimated as a function of A_i – to calculate the increase in lake volume relative to the lake surface area A_SRTM when the lake reaches a certain surface area.

For example, at the time t_a, the lake surface area increases to A_a. When A_a is substituted into Equation (4), the volume change of the lake relative to A_SRTM can be obtained. Assuming that the slope below A_SRTM is similar to that above A_SRTM within a certain range, the volume change when the lake shrinks from A_SRTM to A_b at moment t_b can be estimated using Equation (4).

Last, by inserting and (obtained from Landsat images) into Equation (5), the water volume change of Aqqikkol Lake from 2015 to 2020 was estimated. Then we repeated the same procedures mentioned earlier to estimate the water volume changes of all lakes with this DEM-based method.

We analyzed the lake water volume changes by using SRTM DEM, which has more complete estimation results relative to DAHITI missing the early water level information seriously (Table 2). The results show that the overall water volume of lakes in the study region is on the rise. From 2005 to 2020, except for the reduction of the water volume of three lakes such as La'nga Co, the total water volume of the remaining lakes has increased to varying degrees. In the three monitoring periods, the fastest increase period is from 2015 to 2020, with a total increase of 26.07 km³ in all lakes. The slowest increase period is from 2005 to 2010 with some lakes showing a significant decrease, and the total water volume of all lakes only increased by 11.47 km³. Therefore, the water volume growth rate of China's lakes in the past 15 years shows an accelerating trend.

Water volume changes derived from the SRTM DEM showed considerably spatial heterogeneity with 24 lakes in the TP Lake Region showing a marked increasing water volume trend. Among these, Selin Co and Qinghai Lake, with vast water storage capacities, showed the largest increase (both above 10 km³) with average annual changes of 0.70 and 0.95 km³/a, respectively. Lakes with lower water storage capacities, such as Dorge Co and Mugchu Tso, increased more slowly, with an average annual change of 0.003 km³/a. In contrast, the Bairad Co, La' nga Co, and Gyaring Co Lakes exhibited a certain degree of shrinkage. The water volume in the Xinjiang Lake Region showed a slowly increasing trend. The annual increase in water volume in Ulunger Lake was 0.09 km³/a, whilst the water volume changes in the East China Plain Lake Region showed fluctuating tendencies. The water volume change in Hulun Lake increased by more than 1 km³.

The satellite altimetry-based method with the DAHITI data integrates multi-source satellite data, and the monitoring period for each satellite is different. This resulted in partially missing water level data for some lakes which affected the lake water volume change estimation for some lakes in the TP Lake Region. Based on the data of the ‘water balance’ column in Tables 1 and 2 (the total water balance over the past 15 years), there is a marked difference in the total water balance of 21 of the lakes. These lakes limited water level data from DAHITI for the past 15 years, making it impossible to estimate the results of water volume changes during this period. Therefore, missing water level data is an extremely unfavorable factor for exploring the temporal variation characteristics of lake volumes using satellite altimetry data.

In addition, for water volume estimation in a single lake, the limited DAHITI water level data may also affect the value of the estimation results. Table 3 shows the deviation between the estimated results for Junshan and Dulishi Lakes. The deviation for Dulishi Lake is 0.01 km³, while that of Junshan Lake is 0.12 km³. Junshan Lake is located in the Eastern China Plain Lake Region, and its daily water level changes significantly owing to seasonal precipitation changes and human activities. However, the time difference between the water level and area is approximately two months, leading to a notable deviation in the estimation results. In contrast, the water level time series for Dulishi Lake is relatively complete, and the estimated results had a small deviation.

The DEM-based method has the advantage for estimating lake water volume changes without temporal constraints and can be applied to inland lakes worldwide because of the coverage of DEM data. However, the DEM-based method is highly dependent on lakeshore terrain. For some lakes with complete water level time series, such as the Hulun and Poyang Lakes, the satellite altimetry-based method is more suitable for monitoring the water volume change. However, the limited satellite altimetry-derived water levels from the DAHITI data constrain the applicability of this method.

As summarized in Table 4, the two methods demonstrate different regional applicability. The DEM-based method has a good applicability rate of 100% in the TP Lake Region which has a greater lakeshore slope, while in the Xinjiang and Eastern China Plain Lake Region with flatter lakeshore slopes, the water volume changes of some lakes cannot be estimated as there are applicability rates of only 50 and 40%, respectively. In contrast, the lack of long-term water level data across a period of 10 years for 10 lakes, such as Gongzhu Lake, greatly restricts the use of DAHITI data for estimating volumetric changes in these lakes.

In summary, these two methods, along with their limitations and advantages, are technical complements. Therefore, when assessing lake water volume changes over a large region and over a long time series, more suitable methods can be selected based on the different lake region types. For lakes in regions that lack water level data for a certain period, the DEM-based method can be prioritized to estimate water volume changes during this period. For lakes with a relatively flat lakeshore terrain in the region and the availability of altimeter observations, satellite altimetry-based methods can be used to estimate water volume changes.

This study used and compared two different methods, the satellite altimetry-based method with the DAHITI data and the DEM-based method with the SRTM DEM data, to estimate the water volume change results of 34 lakes in China. This study discussed the limitations and regional applicability of the two methods. The estimation results obtained using the two methods were highly consistent, with an R² of 0.93. However, both methods have advantages and limitations. The satellite altimetry-based method with the DAHITI data can only obtain lake water volume changes during the time period having altimetry data. In addition, for the eastern lakes, the huge daily change in lake water levels requires that the water level and lake range monitoring time points, i.e., the corresponding date, must be completely consistent. The DEM-based method with the SRTM DEM data can be used to estimate the water volume change of any closed lake in any region since 2000; but its accuracy is dependent on the lake shore topography. In our studied lakes in China, extracting accurate lake shorelines is challenging for lakes with slopes less than 1.5°. Because of the accuracy of lake shore extraction, the DEM-based method with the SRTM DEM is more suitable for estimating lake water volumes in the TP Lake Region (with an applicability rate of 100%); however, its applicability is lower in Xinjiang and Eastern China Plain Lake Region (with an applicability rate of 50 and 40%, respectively). Owing to the limited water level data for some lakes, the satellite altimetry-based method with the DAHITI data is more suitable for lakes with more complete water level data. A strategic selection of methods is recommended for analyzing lake water volume changes over extensive regions and timeframes. The DEM-based method is recommended for lakes with steeper shorelines and missing water level data. The satellite altimetry-based method with the DAHITI data is better suited for lakes with flatter shore terrain and comprehensive water level data. By combining these two methods, more complete and accurate estimates of lake water volume changes can be achieved.

This study provides valuable insights for researchers to effectively integrate these methods and obtain more complete and accurate estimates of lake water volume changes, especially when estimating changes over large spatial regions and long time series to understand the underlying patterns. Future researchers can further enhance the effectiveness of these methods by incorporating additional data. The new SWOT mission (launched at the end of 2022) promises to significantly advance lake monitoring capabilities. It is interesting and relevant to evaluate the actual performance of SWOT data for estimating lake water volume changes.

This work is supported by National Natural Science Foundation of China (41901129) and the University Natural Sciences Research Project of Anhui Educational committee (KJ2020JD06). Zheng Duan acknowledges the support from the Joint China–Sweden Mobility Grant funded by NSFC and STINT (CH2019-8250).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.