Filtering Radar Interferometry Time Series with Univariate Least Squares Noise Matrix Analysis

Abstract

Human life is always affected by various natural events such as earthquakes, volcanoes, subsidence, etc. One of the suitable tools for investigating and analyzing these hazards is synthetic aperture radar interferometry. This geodetic technique has the capability of resolving the displacement of the Earth's crust and analyzing the deformation through phase differences of radar images. The main advantage of the InSAR is the high temporal and spatial resolution. Analogous to other geodetic methods, the accuracy of the result depends on the modeling of observational disturbances and noises. Despite progress in recent decades, these disorders have received little attention. The case study is northwest of Hawaii Island. In this study, filtering and reducing the turbulence in time series is based on the most appropriate functional model and stochastic model. This process is done using the MLE test. In this study, functional models include trend, cyclic, and offset. Statistical models also include white noise, flicker, and random walk, whose components are identified through univariate least squares noise analysis. Time series are reproduced through the best functional and statistical models. The results indicate that the best model is the linear trend with the presence of cyclic and offset, and white noise for all pixels. By implementing the univariate least squares noise analysis method, the accuracy of the results improved on average by 43%. In addition, applying both high-pass and low-pass filters resulted in an average improvement of 28%.