Most people around the world have heard of Tibet, the land of high mountains, permanent snow, and the Buddhism. Tibet has a unique topography and geography. In addition to high mountain ranges and deep river canyons, vast prairies and grasslands stretch across the plateau, and four of the world’s most important rivers start their course to the seas from the region.
Although the Qinghai-Tibet Plateau is far away from us, it is closely related to global climate changes. Lake storage change serves as a unique indicator of the change on the Tibetan Plateau (TP). There are about 1200 lakes with an area larger than 1 km2 in the region. However, comprehensive lake storage data, especially for lakes smaller than 10 km2, are still lacking in the region.
Due to the harsh environment and few in situ observation, satellite remote sensing has become an indispensable tool for studying lake dynamics in the region. The Landsat satellite has data archive of more than 40 years, which provides the opportunity to develop a long-term data of lake water volume change to study the climate change in the region. Nevertheless, existing studies are limited to a few large lakes or a short time span of less than 15 years. They have neither made full use of available Landsat data, nor have they covered more than 75% of the lakes in the region.
In this research, using the Google Earth Engine (GEE) geospatial analysis platform, we analyzed 30 years (1989 – 2019) of Landsat imagery to obtain annual lake area time series for 976 lakes with a maximum area larger than 1 km2 on the EBTP. We further estimated annual volume change for the lakes based on the relationship between lake area and surface elevation using digital terrain model data. This study provides a lake water volume dataset which covers so far, the largest number of lakes and the longest time span on the EBTP.
In this research, relative lake volume (RLV) is calculated in two steps: derive annual lake area from Landsat imagery and calculate lake water volume change based on lake area. The first step includes four sub-steps: lake identification, lake analysis extent and seed determination, water classification and segmentation, and annual lake area calculation. And the second step includes deriving lake area and elevation relationship, estimating lake surface elevation, and calculating lake water volume change.
Among these steps, the most important contributions of our work are that our methods can recognize the same lake even if the lake became several separate waterbodies in some of the years. In addition, we used several commonly used methods (linear regression, second order polynomial regression, or MCI methods) to model the relationship between lake surface area and elevation. We found that none of them is always the best, and we chose the method based on the characteristics of each lake.
Comparison with three major existing data products in the region indicates that our dataset is reliable and might be more accurate. To the best of our knowledge, the dataset in this study provides the longest and most comprehensive lake water volume change data in the region, especially for small lakes (1-10 km2). The dataset is valuable in studying the impacts of climate change and water balance in the region. The workflow used in this study can be further improved to process individual Landsat image (instead of annual composite image) and create a lake water volume dataset with a higher temporal resolution in the future.