Unmixing marsh vegetation species across multiple sensors and spatial scales

Coastal marsh ecosystems are critical for providing essential habitats, buffering coastlines against erosion, and sequestering carbon. However, there is a global trend where these ecosystems are disappearing due to rising sea levels, wave-drive erosion, and human activities. Accurate and repeatable...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhicheng Yang, Tegan Blount, Conner Lester, Nat Blackford, Andrea D’Alpaos, Marco Marani, Brad Murray, Sonia Silvestri
Format: Article
Language:English
Published: Taylor & Francis Group 2025-12-01
Series:GIScience & Remote Sensing
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/15481603.2025.2481689
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Coastal marsh ecosystems are critical for providing essential habitats, buffering coastlines against erosion, and sequestering carbon. However, there is a global trend where these ecosystems are disappearing due to rising sea levels, wave-drive erosion, and human activities. Accurate and repeatable observations of marsh vegetation distributions are essential for understanding marsh resilience and informing conservation strategies. Mapping the fractional abundance (FA) of each marsh species using standard airborne/satellite sensor pixels can provide accurate information about marsh vegetation distribution and a deeper insight into marsh dynamics. This study explores the feasibility of estimating FA for marsh vegetation species across the diverse pixel sizes of different remote sensors, based on ground truthing information obtained from Unmanned Aerial Vehicle (UAV) data and field surveys. Specifically, the FA values of each species within the pixels from airborne, WorldView-2 (WV2), and Sentinel-2 (SL2) data, with pixel sizes of 0.15 m, 0.5 m, and 10 m, respectively, were estimated by using a Rescaled Random Forest Regression (RRFR) algorithm. Our results suggest that Random Forest Classification can accurately classify marsh vegetation, with extremely high levels of accuracy when applied to UAV data. This demonstrates that UAVs are cost-effective and efficient for acquiring ground truthing to inform FA estimation algorithms. Our analyses also indicate that the RRFR can accurately unmix dominant marsh species (Spartina and Juncus), as well as bare soil, across various spatial resolutions. However, the unmixing of Salicornia, a minor species, proved to be challenging, indicating the need for additional ground truthing data to capture information relating to minor species. The approach proposed in this study can facilitate the analysis and monitoring of marsh vegetation dynamics, supporting effective conservation and management practices to enhance marsh resilience.
ISSN:1548-1603
1943-7226