Functional diversity rather than species diversity can be accurately assessed by remote sensing in sandy grassland

Abstract

The prediction of grasslands plant diversity using satellite images has been intensively studied. However, the accuracy of functional diversity (FD) is still unknown. Therefore, high spatial resolution Worldview-3 (WV-3) multiple spectral data were used to predict species and FD at the pixel scale (1.2 × 1.2 m) over central Hunshandak Sandland, Inner Mongolia, north China. Data acquired from 120 field plots (6 × 6 m) were used to train and validate several statistical learning methods with a primary objective of linking the satellite spectral and texture indices to the plant diversity indices. Among the several diversity indices tested, functional trait diversity, in particularly Functional Attribute Diversity (FAD1), Modified Functional Attribute Diversity (MFAD) were best predicted (coefficient of determination approximately 0.29 and 0.14, respectively, n=48) using texture indices. However, species diversity (richness, H, E, or D) and other FDs haven’t not been well predicted by WV-3 data. WV data did not significantly improve the prediction accuracy for plant diversity in sandy grassland. Further, high plot-level vegetation coverage can improve the performance of spectral indices for predicting H, E, D and FD. These results highlighted the assessing variability across field conditions and demonstrated the capacity of high spatial-spectral satellite images to monitor plant functional diversity in sandy grasslands.