A new time series vegetation–water index of phenological–hydrological trait across species and functional types for Poyang Lake wetland ecosystem

Abstract

Despite the long history of vegetation mapping with remote sensing, challenges remain in effectively linking remote sensing observations with plant traits and environmental disturbance across species and functional types. This study proposes a classification of aquatic and wetland vegetation plant functional types (PFTs) using time series remotely sensed data applicable to wetland ecosystems with large annual water level changes like Poyang Lake, the largest freshwater lake–wetland of China. We first developed the aquatic and wetland PFT classification scheme which included perennial C4 grasses, perennial C3 reed, C3 (sedges and taller forbs), short C3 forbs, floating aquatic macrophytes, and submerged aquatic macrophytes. Using this scheme, time series normalized difference vegetation index (time series NDVI) images and time series vegetation–water index (time series VWI) images extracted from 32-m spatial resolution Beijing-1 microsatellite data were applied to perform two PFT classifications with a support vector machine (SVM) algorithm. We found that the time series NDVI-based SVM classification method mapped aquatic and wetland PFT distribution with an overall accuracy of 81.3%. However, when the information on water level fluctuation was combined with time series NDVI into a new time series index (referred to as time series VWI), the overall accuracy increased to 90.4%. Results suggest that time series VWI has a stronger capability in distinguishing wetland PFTs with different submersion times and flood tolerances than time series NDVI, and thus may be considered as a key input dataset for future PFT classifications in seasonally dynamic wetlands.