Periodic displacement accurate extraction of reservoir active slopes through InSAR observation and independent component analysis-based wavelet transform

Abstract

The stability of reservoir slopes is often greatly influenced by seasonal rainfall and periodic water level fluctuation. To reveal the spatiotemporal characteristics of displacement and the mechanism of the active slopes, it is of great significance to identify active slopes on reservoir banks and extract periodic displacements. The wavelet transform method based on least squares decomposition has been used to extract periodic displacements of reservoir slopes, which only divides displacements into two components, neglecting errors such as random terms. This paper proposes a method that combines Independent Component Analysis (ICA) and wavelet transform to investigate the temporal relationship between periodic displacements and water level fluctuations. Taking the Maoergai Hydropower Station in Heishui County as example, based on ascending and descending SAR images acquired from 2018 to 2020, a total of 21 active slopes were detected. The time series InSAR results were decomposed by ICA. Through separate analysis and validation on typical slopes, it was demonstrated that the obtained periodic displacements are highly consistent with the water level fluctuations, and displacement changes lag behind water level fluctuations. Cross-validation was performed and proved the stability and reliability of the time lag (about 80–88 days derived from ascending and descending observation) results in this paper. This study improves the accuracy and stability of the periodic displacement extraction and provides technical support for understanding the relationship between the water level fluctuations and the slope displacements.

Keywords

• Maoergai Reservoir
• Active slopes
• Periodic displacements
• Independent Component Analysis (ICA)
• Wavelet transform