Neural network model for leaf chlorophyll content of roadside trees based on hyperspectral approaches

This paper compared the changes of leaf chlorophyll content between roadside trees affected by urban communication environment and corresponding trees growing in a less polluted environment—Jingyue National Forest Park,with the hyperspectral data of leaf samples obtained by using ASD Field NVIR Spectroradiometer.The results showed that urban communication environment had a strong effect on the chlorophyll content of roadside tree leaves,and deciduous tree leaves had more intensive response than conifers tree leaves.The hyperspectral reflectance of roadside tree leaves had a close relation with the chlorophyll content,and the relationship varied with wavelength.At 740～760 nm,the correlation coefficient was 0.72.PSSR(pigment specific simple ratio) index had a close relation with chlorophyll content,and a power regression model was established,with the determination coefficient around 0.82.BP(back propagation) neural network showed a more promising potential for predicting the tree leaf chlorophyll content with hyperspectral reflectance data,and the determination coefficient was 0.97.It was suggested that hyperspectral reflectance data could be used to detect the change of roadside tree leaf chlorophyll content caused by urban communication environment.