This dataset includes data used and generated in a PhD research project 'SAR IMAGES FOR BENCHMARK DATASET CREATION, URBAN POLYCENTRICITY AND DEFORMATION DYNAMICS'. The research aimed to: 1) To design a standard framework to generate SAR benchmark datasets. This framework addresses the challenges of aligning radar coordinates with label coordinate systems by defining representative SAR features into three categories according to the feature attributes, e.g., whether these are of single or multi-polarimetric character. Radar-coded geospatial data are transformed using the proposed hybrid radarcoding methods. The quality of the generated datasets is assessed with custom-designed metrics. Effectiveness of the framework is demonstrated through an application in land use and land cover (LULC) classification, a typical machine learning task in remote sensing. 2) To enhance traditional numerical change detection methods for land cover change detection using SAR images. This thesis introduced a novel KI-MM (Kittler and Illingworth with a Modified Model) threshold determination method and a change-type classification approach. This method aims to improve the change detection method and to perform effectively without requiring extensive training data. These enhanced methods categorize ground target changes and are demonstrated on the Google Earth Engine (GEE) platform using large-scale, high spatio-temporal SAR datasets. 3) To develop a comprehensive method for monitoring and predicting land deformation in urban environments. This method aims to tackle the challenges of deformation monitoring in urban areas. It includes a Persistent Scatterer (PS) based concatenation component for wide-range deformation mapping, a deep learning component for automatic deform ation detection and prediction, and robust quality assessment metrics. The proposed approach is demonstrated by means of urban deformation case studies, highlighting its potential for improving monitoring and risk management in urban environments.