Description

The first online and regularly-updated dataset on land subsidence using Sentinel-1 satellite Synthetic Aperture Radar (SAR) data, over the Netherlands, has been released in https://bodemdalingskaart.nl. More than millions of InSAR (Interferometric SAR) Measurement Points (IMPs) and the associated (deformation) phase time series are publicly available on this website. As such, we are able to directly investigate and analyze the phase time series of all IMPs and skip the complex time series InSAR data processing using such as persistent scatterer interferometry approach. The phase time series per IMP present surface movement in relation to various local human activities and hazards. In order to efficiently and accurately categorize the IMPs in terms of different temporal behavior and identify temporal anomalies, this study will focus on using machine learning approaches to rapidly scrutinizing and accurately analyzing InSAR phase time series.

Objectives and Methodology

This study attempts to explore existing and advanced machine approaches such as a recurrent neural network for time series analysis. The input data – InSAR phase time series can be downloaded from https://bodemdalingskaart.nl, along with the metadata information. The objective of the study is to develop a machine learning method to classify IMPs in terms of different temporal patterns and predict the temporal evolution of IMPs.

Further reading

[1] Ferretti, A., Prati, C., and Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing, 39(1):8–20.
[2] Bas van der kerkhof, Victor Pankratius, Ling Chang, Rob van Swol, and Ramon F. Hanssen (2019). Individual scatterer model learning for satellite interferometry. IEEE Transactions on Geoscience and Remote Sensing, 58(2), PP1273-1280.