An enhanced approach to mangrove forest analysis in the Colombian Pacific coast using optical and SAR data in Google Earth Engine

Abstract

Accurate quantification of land cover is crucial for effectively planning and managing territories. For example, the Colombian Pacific Coast (CPC), a biodiversity hotspot, is home to mangrove forests that provide environmental services to local communities. This study utilized robust multi-sensor and multi-annual satellite data, combined with machine learning in the Google Earth Engine (GEE) platform, to improve the detection of mangrove forests in the CPC. Satellite images from the Landsat 5, 7, and 8 missions were utilized along with L-Band Synthetic Aperture Radar (SAR) data (ALOS-2/PALSAR-2) to quantify changes in mangrove cover between 2009 and 2019. Optical indices and SAR textures were used to identify mangrove cover, and a Random Forest supervised machine learning algorithm was applied for image classification. The multisensory composites produced moderate to high overall accuracies (OA) classifications. Our findings indicate that between 2009 (155,394.3 ha) and 2019 (144,704.3 ha), there was a net reduction of 6.88% in mangrove forest cover in the CPC. This study provides important insights into the conservation and management of Colombian coastal territories and demonstrates the effectiveness of using optical and radar data to accurately recognize coastal forest cover, specifically in tropical mangrove forests.