Predicting plant diversity in beach wetland downstream of Xiaolangdi reservoir with UAV and satellite multispectral images

Accurate and timely acquisition of plant diversity information downstream of the reservoir is helpful to understand the impact mechanism of reservoir operation on wetland plant diversity and formulate reasonable water and sediment regulation strategies. In this study, we conducted field surveys in two communities (Phragmites australis and Tamarix chinensis) at a typical wetland in the lower reaches of Xiaolangdi Reservoir on the Yellow River, and employed UAV and Gaofen 1B multispectral images to estimate the wetland plant diversity. Results showed that most diversity indexes had a higher correlation with the mean of spectral vegetation indexes (DVI, RVI, NDVI, SAVI, and MSAVI). The diversity indexes (C_SP and C_SW) constructed by relative coverage had a better overall correlation with spectral indexes. Interestingly, opposite correlations were found between Tamarix chinensis and Phragmites australis plots. We further gave a deep insight into the interspecific associations in Phragmites australis and Tamarix chinensis plots with the variance ratio (VR) method. It was found that plant species in Tamarix chinensis plot showed positive association (VR > 1), with a VR value of 1.095. Plant species in Phragmites australis plot had a negative association (VR < 1), with a VR value of 0.983. In Phragmites australis plot, C_SP and C_SW showed a significant decreasing trend (r2 of 0.36 and 0.33 respectively, and P values less than 0.001) with the increase of Phragmites australis coverage. Moreover, the effect of spatial resolution was not significant on plant diversity estimation. Correlations between remote sensing indexes and diversity indexes were improved with the quadrat size changing from 2 m × 2 m to 4 m × 4 m. These findings demonstrate promising approaches for remote sensing of wetland plant diversity and indicate that the type of wetland plant community determines the relationship between diversity index and spectral index.