Hyperspectral Remote Sensing Of Plant Leaf Surface Dust And Its Influence On PROSPECT Inversions

Plants have the function of adsorbing dust as the pollutants in the atmosphere can adhere to the leaves,forming leaf surface dust(LSD).LSD has been an important index on detecting plant growth stress and indicating atmospheric pollution.This study measured the LSD contents of typical urban trees including photinia,camphor tree and laurel in Jinhua City,as well as the hyperspectral reflectance of the leaves in dusty and clean conditions.The effect of LSD on the reflection spectra of leaves was analyzed.Sensitivity bands were screened and hyperspectral indexes for estimating LSD were established.In addition,the effects of LSD on inversions through the PROSPECT model have been discussed.The main results are as follows:(1)LSD covering will alter the leaf reflected spectra.The variations of the spectra of photinia,camphor tree and laurel at 400-700 nm and longer than 1300 nm were inconsistent.The reflectance increases as the LSD content decreases from 800 to 1200 nm.Comparing the reflected spectra from the dust and clean leaves,suggested that the reflection spectra of the dust leaf were substantially smaller than those of the clean leaves.In addition,the LSD does not change the three-sided position of the leaf reflectivity and cause red shift.But the influence of LSD has caused the amplitude of the three sides to decrease and the area has increased.(2)The spectra of dust leaves were sensitive to LSD content,mainly in the visible and near-infrared domains.Analysis of the existing index found that the applicable indexes for the three species were different.Correlation analysis was performed on normalization index,ratio index,difference index and leaf dust content.Finally,parameters sensitive to LSD are extracted.The results show that thecorrelation coefficient from high to low is the normalization index,the difference index and the ratio index,respectively.(3)Sensitivity analysis based on the first derivative,second derivative,log and original spectra were implemented,from which the sensitive bands and indices to establish LSD were identified.The estimation of the laurel dust retention capacity was better.Laurel's R~2 value was higher than 0.7 and the highest was 0.809.The R~2 values of the photinia model was 0.7036 and 0.748,respectively.In addition,the model established with the sensitive band of the camphor tree was more accurate;(4)LSD has a certain influence on the reflectance spectra and thus causes changes in the optical properties of the leaf surface.However,inversion using the current PROSPECT model,in general had low outputs due premolar to LSD.In this paper,a modification of the current model has been attempted by adding the leaf surface roughness parameters to the PROSPECT model.The modification improved biochemical parameter inversions.The coefficient of determination for inversion of CHL(Leaf chlorophyll a+b concentration),EWT(Leaf equivalent water thickness),and LMA(Leaf mass per area)reached 0.7405,0.7274,and 0.4203,respectively.The RMSE was less than 0.018.