Chamoli flash floods of 7th February 2021 and recent deformation: A PSInSAR and deep learning neural network (DLNN) based perspective

Abstract

The February 7, 2021, Joshimath flood scenario was one such event that caused widespread damage and led to complete washout of the many hydroelectric power projects located on the course of Dhauliganga River. The physical monitoring and mapping of such events is a difficult task that often involves deployment of labour force in inhospitable terrains. Therefore, remote sensing techniques are used for the mapping and machine learning model-based predictions for future scenarios. Synthetic Aperture RADAR (SAR) remote sensing has been widely used over the years for accurate estimation of many natural and anthropogenic disaster events. This study utilizes Persistent Scatterer SAR interferometry (PSInSAR) technique to map the surface displacement of the 2021 flood scenario and make predictions for future displacement using a Deep Learning Neural Network (DLNN) model. 16 images of both ascending and descending pass were taken for the estimation of Line of Sight (LOS) displacement velocity mapping between January 2020 and April 2021 for Tapovan area. Further, 36 images from January 2020 to December 2022, in ascending and descending passes were used for prediction and validation of future LOS surface displacement using a DLNN model for Joshimath town to see the possible impact of February 7, 2021 event. The PSInSAR LOS displacements were found to be −1.2 ​cm–1.2 ​cm between January 1, 2020 and April 14, 2021, for Tapovan region where the floods had occurred on February 7, 2021. The predicted LOS displacement was observed to be −10 cm–10 ​cm for December 2022 for Joshimath town. These observations clearly indicate the impact of the event to Joshimath town and as one of the causative factors of recent observations of widespread cracks in the buildings in the region.

Keywords

• Remote sensing
• PSInSAR
• LOS displacement
• Surface displacement
• DLNN model