Monitoring and attribution analysis of water resources change in Tibetan Plateau from 1991 to 2021 based on remote sensing

Abstract

The climate and environment in the TP (Tibetan Plateau) have been undergoing significant change under the general context of global climate change. Its lake system is important for the supply and storage of fresh water for billions of people. Although, remotely sensed big data and cloud computing can facilitate enhanced monitoring of water resources changes at large scale, their application to the macro-monitoring of water resources dynamics on the TP has been scarcely employed. Previous work on the dynamics of lakes on TP focused on lake monitoring on multi-year or epoch-based scales and therefore lack sufficient temporal information. Small lakes (< 10 km2 was relatively rare but important in order to gain more knowledge of the smaller lakes, which are more sensitive to the climate changes. Accordingly, this PhD study focus on monitoring the interannual variation of all lakes (>0.01 km2) on the TP for 1991–2021 and quantify the mechanisms behind the drivers.
The main findings are as follows:
(1) This study provides a continuous temporal profile of lake dynamics through the generation of high-accuracy (overall accuracy of ~ 97.7%) lake maps for all lakes larger than 0.01 km2. For the lake area on the TP, we found that both lake area and number generally showed increasing trends during 1991-2021 (638 km2/y and 587.5 lakes /y, respectively), and identified a hotspot with the most rapid increase in 2000 (5884.3 km2 and 3713 lakes, respectively) across the TP. The area and quantity of lakes on the TP experienced an increase with fluctuations, especially the lake numbers. The area and number changes of large (>10 km2) and small lakes (0.01 – 10 km2) were mostly consistent from 1991 to 2021. Large lakes’ area changed more drastically and dominated area changes of all lakes with area larger than 0.01 km2 on the TP. The small lakes’ number changed more obviously and dominated the number dynamics of lakes on the TP. Also, the changes of the number of large lakes will have normally one year lag compared to the changes of small lakes’ number. A stable period from 2002 to 2012 was found in number and area for both large and small lakes. Two great rises happened in 2000 and 2017 in number and area for both large and small lakes. Among 12 basins on the TP, the Inner basin not only had the most lakes and largest lake areas on the TP on average, but also had the largest lake area and number increase rates. Yangtze basin had the second most lakes on the TP and the second largest lake number increase rate. Yellow basin had the second largest lake area on the TP. Specifically, Qinghai Lake accounted for the most lake areas in Yellow basin, its area changes also dominated the area dynamics of Yellow basin. Qaidam basin had the second highest lake area increase rate. The basins that have more lakes averagely contributed more to the overall increase in lake numbers on the TP and the basins that have larger lake areas contributed more to the overall increase in lake areas.
(2) From 1991 to 2021, four key drivers of Tibetan Plateau (TP) lake dynamics— Water Availability (WA), snowmelt, runoff, and skin temperature—exhibited distinct spatiotemporal variations. WA showed no significant long-term trend with prominent short-term fluctuations; snowmelt declined slightly but significantly (slope = -0.38 mm/y, p < 0.05); runoff lacked an obvious overall trend but had large interannual variations; skin temperature rose significantly (0.04 K/y, R² = 0.32, p = 0.00), with stronger warming in the southeastern TP. Notably, the dominant drivers of lake dynamics across the TP’s 12 basins displayed strong spatial heterogeneity. Temperature was the most widely influential driver, governing lake changes in 7 basins (Ganges, Qaidam, Mekong, Tarim, Yangtze, Yellow, Brahmaputra). WA dominated 2 basins (Amu Darya, Inner basin subset), runoff led changes in 2 basins (Hexi for lake number, Indus for lake area), and snowmelt was the key driver in 1 basin (Salween), where it interacted with temperature. This spatial differentiation was closely linked to each basin’s climatic background, topographic conditions, and hydrological processes. Basins with strong spatiotemporal heterogeneity (e.g., Inner basin) showed weak correlations between annual average driver data and lake changes. In contrast, drivers with drastic changes and small seasonal differences (e.g., Tarim basin’s temperature) formed significant correlations with lake variations. Additionally, single drivers often exerted significant impacts only during specific periods of the 31-year study. Generally, smaller spatial or temporal scales yielded higher accuracy in driver analysis.
(3) Case studies
Several case studies were conducted to deepen the understanding of lake dynamics and driver impacts on the TP. Focusing on three representative lakes (Dogai Coring, Selin Co, and Qinghai Lake) with distinct basin characteristics, this study analyzed their long-term changes and responses to four key drivers (Water Availability, snowmelt, runoff, temperature) using the GEE platform and multiple linear regression. Temperature emerged as the significant driver for Selin Co’s area changes (adjusted R² = 0.162, p = 0.014), while Dogai Coring and Qinghai Lake showed no statistically significant single-driver contributions, reflecting the complexity of local hydrological processes. Seasonal detection in 2000 revealed differential driver effects: Dogai Coring and Selin Co exhibited notable area increases linked to concentrated seasonal water supply, whereas Qinghai Lake’s slight shrinkage correlated with its lower dependence on seasonal precipitation input compared to the other two lakes. These case studies confirm that lake responses to drivers are scale-dependent, with small lakes and glacial fed lakes showing greater sensitivity to seasonal driver fluctuations.

(4) An advancing method was conducted to detect all lakes larger than 0.01 km2 on the whole TP accurately and generate a continuous change process of lakes on the TP. This method used all the available Landsat (Land Satellite) imagery, improved water body mapping algorithm in conjunction with Google Earth Engine cloud computing platform. Therefore, in this study, we took advantage of the ability to map all lakes (including small lakes) accurately and continuously on the TP to map the lake annual changes to a higher accuracy and explore the relationship between lakes changes and climate dynamics even further and deeper.

Date of Award	15 Apr 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Ping Fu (Supervisor), Stephen Grebby (Supervisor) & Jinwei Dong (Supervisor)