Inferring Plant α Species Diversity In Marsh Wetlands Based On UAV Remote Sensing Technology

Wetlands are one of the richest biodiversity hotspots,which are currently one of the most stressed ecosystems due to their global fragmentation and loss.Rapid and large-scale changes in wetland biodiversity far beyond the ability to monitor through on-site sampling alone.However,it is challenging to track the change of wetlands plant diversity in a large range and for a long time,resulting in very limited data about wetland plant diversity,which restricts the wetland conservation work.The spatial resolution of traditional satellite remote sensing images is too coarse to be used for the accurate monitoring of wetland plant diversity.The rapid development of UAV remote sensing technology has provided new possibilities for wetland biodiversity assessment,which can also provide theoretical basis and technical guidance for evaluating plantαspecies diversity based on satellite remote sensing images in the future.In this study,the Sanjiang National Nature Reserve was selected as the research area.High spatial resolution multi-spectral images and field verification data were obtained based on UAV and field survey,which were used to calculate various spectral andαspecies diversity indices,and then the spectral-species diversity prediction models were constructed by univariate and multivariate regression methods.The optimal spectral and species diversity indices and the factors that affecting the relationship between spectral and plant species diversity were discussed.Specifically,the main research content of this research includes the following three parts.1.Comparison the predictive ability of different spectral indices:Comparison the predictive ability of different vegetation indices for wetland plant species diversity and their sensitivity to image noise using univariate and multivariate regression methods.The results showed that compared with commonly used vegetation indices such as NDVI,the ability of MTCI and NDREI were stronger to predict the wetlands plant species diversity.In addition,NDVI,CTVI and other indices are more sensitive to image noise,while MTCI,NDREI and MSAVI are less sensitive to image noise.The combined use of VI_MEAN and VI_SD significantly improved the predictive ability for plant species diversity,which proved the complementarity of the VI_MEANand VI_SD in predicting plant species diversity.2.Application potential of Hill numbers in spectral-species diversity studies:By comparing the predictive ability of spectral indices to commonly used species diversity indices（Richness,Shannon and Simpson）and Hill numbers.The results showed that the best predictive ability of spectral indices to Hill number was stronger than that of commonly used species diversity indices.Using only commonly used species diversity indices will underestimates the real predictive ability based on spectral information of remote sensing images,proving the limitations of commonly used species diversity indices and the application potential of Hill number in the study of spectral-species diversity relationship.3.The dominant role of vegetation coverage in community canopy reflectance:By comparing the predictive ability of spectral indices to species diversity indices that calculated based on the density and coverage of each species respectively.The results showed that,compared to the species diversity indices that calculated based on the density of each species in the community,spectral indices exhibited stronger predictive ability to the species diversity indices that calculated based on the species coverage in most cases.Moreover,this relationship was more robust across the entire continuum of community species diversity,demonstrating the significant impact of vegetation coverage on the spectral-species diversity relationship.