Remote sensing data to assess compositional and structural indicators in dry woodland

Abstract

Integrating field-based and remotely sensed data has proven valuable for assessing on-the-ground diversity of plants across a range of spatial scales. Here we assessed whether remotely sensed data is a good indicator of vegetation composition and structure in dry, Prosopis flexuosa-dominated woodlands. Our objectives were (1) to quantify on-the-ground vegetation composition and structure using (A) field-based methods and (B) remotely sensed images and analysis techniques, and (2) to evaluate how well the data extracted from remotely sensed data estimate field-based measures of vegetation composition and structure. We selected 40 individuals of P. flexuosa in Ischigualasto Provincial Park (San Juan, Argentina) and its influence zone. Each individual was the center of a plot (1500-m2) where we recorded richness (compositional indicator) and abundance (structural indicator) of trees, shrubs and other plants (i.e. cacti, grasses and forbs). To assess woodland structure, we evaluated canopy area of each P. flexuosa and the proportion of adult P. flexuosa trees in a plot. In addition, we used Landsat 8 OLI to calculate SATVI (Soil Adjusted Total Vegetation Index) values from the pixel that corresponds with the center of each sample plot, and then estimated first- and second-order texture measures (in 3 × 3 and 5 × 5 moving window sizes). We fitted generalized linear models with different error distributions. Vegetation richness was significantly and directly related to range and entropy (3 × 3 and 5 × 5 windows). Both trees and shrubs, were related to SATVI values and first- and second-order means (3 × 3 and 5 × 5 windows). Moreover, shrub abundance was inversely related to range and entropy (5 × 5 window); and the “other plants” group was inversely related to first- and second-order means in the same window. Variance of the canopy area was directly related to range (5 × 5 window); however, proportion of adults was not related to remote sensing data. Our findings suggest satellite imagery-derived image texture is a valuable tool for management and conservation, and can indicate areas of high plant species richness and abundance of trees and shrubs and help differentiate areas of different canopy sizes in dry P. flexuosa-dominated woodlands of Argentina.