Skip to Main Content
Satellite radar altimetry has effectively been used for monitoring the water level change in recent years. In this study, Qinghai Lake was taken as an example to simulate and forecast water level using the multi-altimeter data from Envisat/RA-2, Cryosat-2/Siral, and Jason-1/Poseidon-2. First, using the robust least square method and system bias correction algorithms, abnormal water levels and the system bias were eliminated, and an accurate lake-level time series was obtained. Then, singular spectrum analysis (SSA) algorithms were used to extract the effective fluctuation signal from the accurate lake-level time series, and the accuracy of the altimetry data was improved. Based on an analysis of SSA algorithms' characteristics, comparison of the SSA-extracted fluctuation signal, and in-situ gauge measurements of Qinghai Lake, the accurate lake-level time series was affected by white noise of zero-mean and 0.5-m variance and colored noise of 0.2202-0.2473-m mean and 0.252-0.2800-m root-mean-square difference. After eliminating the white noise, the accuracy of the altimeter data reached the decimeter level in inland lake monitoring. Next, the SSA-extracted fluctuation signal was decomposed into linear composition, periodic components, and a residual component, and a combined linear-periodic-residual model was established using simple regression, a trigonometric function, and autoregressive-moving-average models. Using the model, the water level change of Qinghai Lake was simulated and forecasted to 2 years, with its accuracy reaching the decimeter level. The experiences of this study can provide an effective reference for the other lakes.