The effects of water depth on estimating Fractional Vegetation Cover in mangrove forests

Abstract

Maps of mangroves have often been limited to showing the presence or absence of mangrove trees and seldom have studies shown an important indicator of ecosystem integrity such as vegetation cover. Fractional Vegetation Cover (FVC) is used to assess ecosystem health, land cover and carbon stocks, hence accurately measuring FVC is an important task for scientists and land managers. Many methods have been proposed to measure FVC and simple linear models are commonly used. We created an experiment that allowed us to: 1) acquire very detailed hyperspectral imagery (1 mm pixel size) from a simulated mangrove forest, 2) measure the effect of water depth on FVC estimations, and 3) compare the relationship of eight spectral bands and indices with FVC using linear and non-linear models. After acquiring the imagery we corrected for dark signal and a white reference, performed spectral and spatial resampling, and created linear and non-linear models across four pixel sizes. Our results suggest that 1) linear and beta models have similar performance across all pixel sizes; 2) Soil Adjusted Vegetation Index (SAVI), Modified Soil Adjusted Vegetation Index2 (MSAVI2) and Enhanced Vegetation Index (EVI) perform better than the Normalized Difference Vegetation Index (NDVI), and, 3) our models perform better at fine pixel sizes than coarse scales. We tested our results on high-resolution satellite imagery with similar results and, therefore, recommend using SAVI, EVI or MSAVI2 when predicting FVC instead of NDVI.