[Insert Figure 1 here]

2.2 Study data

2.2.1 LEGOS Hydroweb and DAHITI altimetry data archives
Observation data provided by altimetry satellites has been gradually increased since the early 1990s, including TOPEX/POSEIDON(T/P), ERS-1, ERS-2, Jason-1/-2/-3, Envisat and Cryosat-2(Table 1). The combination of different altimetry satellites can increase the spatial coverage of altimetry measurements and extend the temporal span to nearly thirty years. At present, the long-term lake level changes reconstructed by many institutions is based on the altimetry satellite data of different service years. As one data source of our analyzed water levels, we obtained lake levels from the Laboratoire d Etudes en Géophysique et Océanographie Spatiales (LEGOS) Hydroweb altimetry data archive and the Hydrological Time Series of Inland Waters (DAHITI) altimetry data archive. The LEGOS Hydroweb altimetry data archive was derived from merged Topex/Poseidon, Jason, Envisat and Sentinel data (Crétaux et al. 2009; Crétaux et al. 2011b). All background rectification have been used to produce this data archive. Since different altimetry satellites are used for the same lake, the water level calculation is mainly divided into three steps. First, each satellite data was processed separately to obtain the altimetry measurement data of different satellites, and then Topex/Poseidon data was used as a reference standard to minimized potential radar instrument deviations and geoid differences between different satellites. Finally, lake levels from the different satellites were merged. The DAHITI data archive was launched by the Deutsches Geodätisches Forschungsinstitut (DGFI, now DGFI-TUM) in 2013. The processing procedure of DAHITI database in detail can be referred to in Schwatke et al. (Schwatke et al. 2015) which was based on an extended outlier detection and a Kalman filtering of multi-source merged altimetry measurements. In this paper, because the algorithms adopted by the two data archives were different, we only used one of them for a lake. When a lake was covered by the two data archives at the same time, Hydroweb data archive was selected as the priority.