Interferometric Synthetic Aperture Radar (InSAR) is used to explore the ground truth by exploiting the phase difference between two synthetic aperture radar images. The phase of an InSAR image pair (called an interferogram) is used to reconstruct a digital terrain model of the imaged area. Typically, the interferogram is noisy due to different effects (thermal, temporal, etc.). In this paper, the interferogram phase noise is characterized by an additive noise model and multiplicative noise model. These models can be employed in a total variation approach to find the optimal approximation of the uncorrupted phase information. The split Bregman iteration approach was used to implement the algorithm. The experimental results show that the proposed approach can robustly and efficiently remove the noise in a numerically stable way. The signal-noise-ratio was used to evaluate the performance of this noise depression approach. From the results, it was found that the additive noise model is more effective at explaining the noise behaviors occurring in an interferogram.

This paper presents an improved PS-InSAR algorithm and preliminary test results on subsidence determination in central Taiwan. PS-InSAR with ALOS PALSAR images is used to evaluate the land surface subsidence in central Taiwan mostly covered by vegetation and mountains. The PS-InSAR results are analyzed and then evaluated by comparing with precise leveling data. Both results determined by PS-InSAR and precise leveling show similar subsidence pattern. The statistic figures derived from the absolute differences in vertical displacement velocity vectors on 200 PSs and 200 benchmarks show that the absolute differences have the mean 1.4 /year, the minimum 0.0 /year, the maximum 4.5 /year and the root mean square difference (RMSD) 1.4 /year. The conclusion infers that monitoring large-scale subsidence in central Taiwan by means of PS-InSAR is feasible and applicable, and it has a good application potential for a national-level regular subsidence monitoring project particularly with advanced radar data.

Synthetic Aperture Radar (SAR) is capable of day/night and all-weather data acquisition, moreover, detection of centimeter-level deformation is achievable through differential interferometric synthetic aperture radar (D-InSAR) technique. However, due to the atmosphere delay, the surface deformation extracted through D-InSAR processing is incorrect. In order to improve the accuracy, the atmospheric correction is required. This paper focused on the reduction of errors caused by dry delay. A total of seven ALOS PALSAR images acquired from 2007 to 2011 was processed with 2-pass D-InSAR technique to extract surface displacement. Once this was achieved, dry delay effect was computed based on the data collected from local weather stations and then applied to remove the atmospheric dry delay error. It was shown that the accuracy of 36% to 59% of the check points in the six pairs was improved after dry delay correction. The characteristics of the dry delay effect were also examined. It was suggested that dry delay correction is essential for achieving accurate D-InSAR surface displacement results.

The Puli Basin, located between the Western Foothills and the Hsüehshan Range is the largest basin in the active orogeny in central Taiwan. South to the Puli Basin, a series of small basins spread along the NNE direction. Various tectonic models have been proposed in previous studies to interpret the formation of the Puli Basin. In this study, we aim to measure the deformation and to understand the tectonics of Puli Basin by the method of satellite remote sensing and field observation. We applied the Persistent Scatterer Interferometric SAR technique to monitor the surface deformation. Our result shows a relative lower LOS velocity in Puli Basin, which implies the change of depth of the subsurface basal detachment. A steep ramp along the detachment is proposed for the grown structure underneath the Puli Basin in this study. In order to interpret the mechanism of surface deformation in central Taiwan, we implement several forward models based on previous researches and our interferometric results and then compare our observations with accurate GPS and precise leveling data. These results can be compared with our field observations and with the recent seismicity data (20130327 and 20130602) to examine the activity and geometry of faults of this area, where a conjugate fault system below the basins was suggested. Due to the lack of evidence, the relationship between the conjugate system and the basin subsidence mechanism is hard to determine but may be the key to understand the tectonic activity in central Taiwan.

Yunlin area is suffering severe subsidence caused by groundwater withdrawal and the high compressibility of sediment. The conventional geodetic methods including GPS, leveling and multi-layer compaction monitoring well have been used to monitor land subsidence. Those sensors can only deliver point-wise displacements and the monitoring capacity of displacement is limited by point density. In this study, we applied temporarily coherent point SAR interferometry （TCPInSAR） to monitor land deformation rate over Yunlin area. The proposed method can estimate deformation parameters without the effect of unresolved phase ambiguity resolution. The study utilizes 15 ALOS PALSAR acquisitions from March 2007 to March 2011 to derive land deformation. The TCP density over Yunlin is 196 pixels/km2, compared to 0.22 point/km2 of the leveling benchmarks. TCPInSAR yields vertical displacements matching the leveling result to 0.6 cm/year （RMSE）. This research suggests that TCPInSAR with ALOS PALSAR data can effectively monitor land subsidence in a large area like Yunlin area.

The technique of the Synthetic Aperture Radar (SAR) is developed in recent years. Without the limit of the weather and sunshine, SAR is different from the optical technique. The extension of the improved SAR technique, such as Interferometry SAR (InSAR) and Persistent Scatterer InSAR (PSInSAR), are widely used to build the Digital Terrain Model (DTM) and detect the deformation rate of the surface. The basic theory of an InSAR technique is based on time difference in arriving at two different receivers from the signal which was just sent out by the transmitter of a satellite. The changes in height can be converted to phase differences. However, if the observed time is extended, the dependence and precision will decrease because of a lowering coherence of two images. This problem may be solved by using the PSInSAR technique. The method is used to find points with stable signatures. According to the characteristics of these points, the result usually has high precision, which is mentioned in many research papers. If this method is applied to SAR images of Taiwan areas, it is not so satisfactory because the surface deformation is too erratic. So, our research aims to improve this method to make it appropriate for the country. To do this, additional information is needed. We should find other points called the distributed scatterer. Distributed scatterers are selected by the center pixel of a region which has the same signal reflection. They have excellent dependence on account of statistical calculations. As distributed scatterers are included, the improved effect is shown in our research.

Slope land area covers more than 70% of Taiwan. Landslide hazards happen frequently owing to collision of Philippine sea plate and Eurasian plate in terms of earthquakes and typhoons formed by tropical cyclone. Heavy rainfall induced landslide investigation requires huge amount of budget. Landslide investigation usually adopts field investigation, identification from optical satellite images and aero photos. Even optical image based landslide identification has very high accuracy, but the investigation of landslide before total failure lacks of systematic methodology. Thus a potential landslide investigation is proposed in this research. Potential landslide area is defined with differential interferometry radar imagery, which topographic effect has been removed from digital elevation model. The potential landslide map is defined with fringe and compared with landslide area defined from optical satellite imagery. The study is trying to provide evidence that potential landslide map defined by differential interferometry has earlier finding of landslides. Study area is selected as May River watershed. May River watershed locates between Renai and Puli township. Landslides and debris flow happen frequently in this watershed. Despite of direct hazard induced by landslide, secondary hazard attacked with debris flow located in channel is another important issue. The result of this study shows that potential landslide investigation with differential interferometry has high reliability and can be used as other potential landslide investigation and reference of land use plan.