Semi-Automatic Monitoring of Miankale Wetland Drought Based on a Hybrid Anomaly Detection Method Using Time Series Satellite Images in Google Earth Engine

Document Type : Original Article

Authors

School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran

10.22059/eoge.2024.374541.1148

Abstract

Wetlands are invaluable ecosystems at risk of destruction due to drought. Continuous monitoring of drought over the years is crucial for environmental management. Traditional mapping methods are expensive and time-consuming. Remote sensing techniques provide a more efficient alternative, allowing for the survey of large geographic areas in a short period. This research aims to detect drought-affected regions of the Miankale wetland, Mazandaran province, Iran, from 2009 to 2021 using a semi-automatic hybrid approach combining multiple anomaly detection algorithms. Time series Landsat satellite images were used to identify effective drought indicators. Clustering-based methods identified anomalous temporal and spatial breakpoints, followed by statistical and machine learning techniques to produce an accurate wetland drought map. Evaluation using ground truth images yielded an overall accuracy of 97.89% and an F1 Score of 98.91%. This study fills a gap by utilizing anomaly detection methods for drought monitoring, presenting a fast and accurate approach that leverages the maximum capacity of satellite data and minimizes errors through a combination of different techniques.

Keywords

Main Subjects

Earth Observation and Geomatics Engineering
Volume 7, Issue 2
December 2023

Files

Share

How to cite

Statistics