BioOne.org will be down briefly for maintenance on 17 December 2024 between 18:00-22:00 Pacific Time US. We apologize for any inconvenience.
How to translate text using browser tools
6 November 2024 Integrating a High-Resolution Canopy Height Model With a Sentinel-1/2 Fused Product in Google Earth Engine for Mangrove Species Mapping in the Caroni Swamp
Deanesh Ramsewak, Arvind Jagassar
Author Affiliations +
Abstract

Small island states are amongst the most vulnerable to the effects of climate change. Many of these effects have been occurring along the coastal zones. Mangrove ecosystems are amongst the most effective natural protection against coastal erosion and inundation. Mangroves are particularly important to the islands of the Caribbean which are subject to the impacts of hurricanes and sea-level rise. Mapping of mangrove ecosystems is key for monitoring changes in their structure and health. This is critical for their conservation and rehabilitation and aids in improving coastal protection and the numerous other ecosystem functions that mangroves provide. Mangrove extent mapping and monitoring have been effectively carried out using advanced geospatial techniques and data. Mangrove species mapping from remotely sensed data, however, has been challenging due to the spectral similarity among several species. Despite this, species maps remain an integral component of monitoring changes within mangroves and provide a core metric towards their conservation. In this study, we combined radar backscatter data from Sentinel-1 with the spectral reflectance information from Sentinel-2 and tree height data from a global high resolution Canopy Height Model (CHM) produced by World Resources Institute (WRI) and Meta, to produce mangrove species maps for the Caroni Swamp region of Trinidad. A machine learning model was trained within Google Earth Engine's (GEE) open-source platform to produce five classes. We also included Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Green Normalized Difference Vegetation Index (GNDVI) to improve the detection of the two main mangrove species. Independent sets of GPS field data were collected and used to train the machine learning model and then conduct an accuracy assessment of the final mangrove species map. The combined optical-radar-CHM classified product resulted in an overall combined accuracy for both seasons of 98.95% and outperformed a classification based on 3 m PlanetScope SuperDove 8-band imagery at 96.55%. This study recommends that the combined multispectral, radar, CHM, GEE approach is a simple, cost-effective, highly accurate method that can be applied to mangrove species mapping within the Caribbean region.

Deanesh Ramsewak and Arvind Jagassar "Integrating a High-Resolution Canopy Height Model With a Sentinel-1/2 Fused Product in Google Earth Engine for Mangrove Species Mapping in the Caroni Swamp," Caribbean Journal of Science 54(2), 428-442, (6 November 2024). https://doi.org/10.18475/cjos.v54i2.a21
Published: 6 November 2024
RIGHTS & PERMISSIONS
Get copyright permission
Back to Top