Tracking coastal tidal wetland change using Landsat time series
Abstract
Coastal tidal wetlands are the most stressed ecosystems due to frequent land use change and sea level rise, which leads to altered ecological functions and loss of climate mitigation efficacy. Tracking tidal wetland change requires detection and inventory of the frequency, intensity, and characteristics of both naturally-occurring and human-influenced land use changes. We applied the Continuous monitoring of Land Disturbance (COLD) algorithm with the long-term Landsat record. However, the original COLD algorithm would not work well for changes occurring in the tidal wetland regions because the effect of tidal water level variation and their irregular conditional change. In this study, we refined and improved the COLD algorithm to track the variabilities between wetland and dryland, between sub-classes of the wetland (open water bodies, marshes and tidal flat), and conditional and ephemeral change of the wetland (marshes). Two major improvements were achieved: (i) Several water and vegetation related spectral indices were incorporated as variables beside the surface reflectance values, which could increase the sensitivity to conditional changes. (ii) A water level regressor, derived from NOAA coastal tide predictions, was introduced into the time series model to suppress the noise caused by the tidal variation. Against our samples in Sandy Hook Bay and Delaware Bay, the improved approach achieved more than 20% improvement of overall accuracy (F1 score) compared to COLD (75.34% versus 54.25%).
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMB048.0018Y
- Keywords:
-
- 0414 Biogeochemical cycles;
- processes;
- and modeling;
- BIOGEOSCIENCES;
- 0428 Carbon cycling;
- BIOGEOSCIENCES;
- 0442 Estuarine and nearshore processes;
- BIOGEOSCIENCES;
- 0469 Nitrogen cycling;
- BIOGEOSCIENCES