Change Detection and Tracking of the Coastline of Pakistan using Machine Learning and Remote Sensing

Abstract

Shorelines are becoming increasingly fragile due to the combined effects of climate change and natural disasters, leading to erosion and flooding in coastal areas. This study used satellite imagery data spanning 20 years to analyze the changes in five coastal regions: Karachi, Manora, Jiwani, Kemari, and Gwadar. The Random Forest classification algorithm achieved high accuracy (95% to 99%) in classifying images, supported by a kappa coefficient of 0.85 to 0.90. Additionally, a time series model was employed, showing a mean absolute percentage error of 4.5% to 4.95%. The findings revealed dynamic changes in the coastlines of the studied regions from 2000 to 2020. Karachi experienced significant accretion at a rate of 20 meters per year from 2000 to 2006, followed by erosion of -10 meters per year in 2014. However, the shoreline has since experienced accretion, with an annual increase of 15 meters per year until 2020. Similarly, Kemari initially witnessed severe erosion of -11 meters per year from 2000 to 2006, leading to substantial accretion. This trend improved in 2014, with an erosion rate of -22 meters per year. From 2014 to 2021, the shoreline consistently experienced an annual increase in the accretion rate of 11 meters per year. The study reveals varying trends in coastline change across different locations, with a mix of expansion and erosion. Future coastal management should address these

changes and their impacts. Regions experiencing negative trends (Manora, Kemari, Ji- wani) may need erosion mitigation strategies, while areas with positive trends (Karachi,

Gwadar) should focus on sustainable development. Comprehensive coastal manage- ment is crucial to protect ecosystems while accommodating human activities, consid- ering the dynamic nature of coastlines.