Application of Cloud-Based Machine Learning Approach to Analyse Remotely Sensed Data for Coastal Monitoring in Bangladesh

Abstract

AbstractThe implication of artificial intelligence (AI) is becoming popular in the field of Remote Sensing (RS) due to solving critical problems, like big data analysis, advanced classification algorithms, storage, etc. Cloud-based AI and remote sensing services allow users (with little technical background) to analyse imagery with high precision. This research has been conducted using Remote Ecosystem Monitoring Assessment Pipeline (remap) cloud service. The coastal region of Bangladesh has been considered as a study area. All available processing tools in Google Earth Engine (GEE) server have been utilized. Random Forest Classification (RFC) has been employed to classify the two periods (1999–2003 and 2014–2017) of pre-processed cloud-free Landsat satellite images. The classification algorithm produced acceptable Land Use-Land Cover (LULC) maps with the pixel-based error matrix, 88.38% for the first period (past 1999–2003) and 83.07% for the second period (2014–2017), respectively. Maps further highlighted changes in LULC classes over time. The LULC classes had changed substantially due to natural processes and anthropogenic actions. The study demonstrated the potential use of remap to monitoring LULC changes in the coastal region of Bangladesh. The outcomes will help to improve policies related to land management and forest conservation in the region.KeywordsRemapRandom forest regressionLand useLand coverSatellite imageMangrove forestConservationLand policy