AI-Based Framework for Rapid Extraction of InSAR Coherence from COMET-LiCSAR Portal

Abstract. Interferometric Synthetic Aperture Radar (InSAR) coherence analysis is a powerful remote sensing technique for detecting ground surface changes caused by natural disasters like wildfires. However, its application has been limited by the large file sizes and high computational demands of Synthetic Aperture Radar (SAR) data, which makes processing on standard personal computers difficult and time-consuming. While the COMET-LiCS Portal democratized access to open-source InSAR data, its complex database created a new bottleneck, making it difficult for users to navigate and download the correct data frames efficiently.

To address these challenges, this study introduces a novel AI-based framework, developed in the Google Colab cloud environment, designed to streamline the rapid extraction and analysis of InSAR coherence data from the COMET-LiCS Portal. The framework uses an AI tool to parse the database and presents the data through a simplified, user-friendly interface that bypasses local computational limitations. Key features of the tool include automated batch downloading to acquire multiple datasets simultaneously. It also integrates OpenStreetMap for overlaying the data and satellite basemaps to improve spatial orientation with province and district boundaries. Additionally, a real-time overlay of active fire detections from the NASA FIRMS API enables immediate cross-comparison between coherence changes and known fire events. The model's effectiveness is demonstrated through case studies of large-scale wildfires in Türkiye, including in Antalya-Manavgat and Muğla-Mazıköy. The analysis confirms that the tool can reliably visualize burned areas by identifying the significant increase in coherence that occurs post-fire due to vegetation loss. The study also identifies a key limitation: the 8-bit data conversion used by the COMET-LiCS Portal reduces data precision, which can hinder the detection of smaller-scale fires. Ultimately, this work delivers a practical, fast, and highly accessible tool that makes coherence-based analysis for disaster monitoring available on both standard computers and mobile devices.