Spectral response characteristics and identification of typical plant species in Ebinur lake wetland national nature reserve (ELWNNR) under a water and salinity gradient

Abstract

The spectral characteristics of variable selection are particularly important for wetland vegetation mapping. In the present study, we combined soil salt and water content with spectral data collected in the Ebinur Lake Wetland National Nature Reserve (ELWNNR) in Western China to understand the effects of soil salt and water content on plant spectra. The results showed the following: (1) the distribution of plants reflect the macroscopic response characteristics of plants on water and salt environment; (2) a certain response rule exists between the spectra of different plants under a water and salt gradient, e.g., with increase in water and salt gradient, the spectral reflectivity of salt-dilution plant decreases, and salt-exclusion plant increases; (3) a response pattern is formed between the “trilateral” characteristics of plant spectrum and water salt gradient. With the increase of salinity gradient, the “red edge”, “blue edge”, and “yellow edge” shows the most obvious changes in the 0.8 order derivatives, e.g., when the soil salt content was range from 4.2 to 8.8g/kg, the spectral characteristics of the plants were the most obvious; (4) Fisher linear discriminant analysis found that during fractional order to integer promotion, classification accuracy of the 0.8 order derivative was higher than the integer order derivatives. Therefore, the “trilateral” characteristics of plants spectra in the 0.8 order derivatives were more accurate than the first derivative. The 0.8 order derivative was more advantageous to distinguishing plants, with a classification accuracy of 89.37%, indicating the potential of 0.8-order derivative for hyperspectral remote sensing of plants. This study introduced a fractional order derivative to hyperspectral remote sensing for the quantitative analysis of differences in the vegetation spectrum, provided new insights to the research and application of vegetation remote sensing.