Estimation Of The Distribution Of Dust Deposition In Urban Shrubs Based On The PROSAIL Radiative Transfer Model

Monitoring and estimation of urban vegetation classification and dust deposition capacity is the basis for urban management and landscape planning.It can provide technical methods and data support for research fields such as atmospheric particulate pollution prevention,urban ecological environmental protection,and human health risk assessment.Based on Sentinel-2A satellite remote sensing image data and field sampling of leaf dust deposition parameters,leaf biochemical component parameters,and hyperspectral data,this paper selected Xuhui District and Minhang District of Shanghai for research,and explored the classification and extraction of urban shrub information,and estimation of dust deposition capacity.In terms of vegetation classification,Spectral Mixture Analysis technology was used,four different combination schemes of categorical feature variables were designed,and Support Vector Machine classifier（SVM）and Random Forest classifier（RF）were used for classification,and the optimal scheme was selected to extract shrub information.In the estimation of the dust deposition capacity of shrubs,the PROSAIL radiative transfer model was used to simulate the reflectance spectrum of the shrub canopy,and compared with the measured canopy data to optimize the model parameter settings suitable for different dust deposition capacities.Combined with the monitoring data of the dust deposition capacity of the samples,several red-edge spectral indexes were established based on the simulated canopy spectrum,and the correlation analysis was then carried out.According to the results of model accuracy verification,combined with satellite images,this study successfully analyzed the spatial distribution of the shrub dust deposition capacity.The main research conclusions are as follows:（1）The correlation between the main input parameters of the PROSAIL model and the dust deposition capacity of shrubs was significant.Combined with EFAST quantitative sensitivity analysis and local qualitative sensitivity analysis,,₍(6（7）andwere three parameters with great sensitivity.According to the chlorophyll content and other parameters corresponding to different dust deposition capacities,the Root Mean Square Error（RMSE）between the measured canopy spectrum and the simulated canopy spectrum was small,all below 0.1098.And the spectral red edge information obtained by PROSAIL simulation was relatively consistent with the actral measured data.Therefore,PROSAIL model was useful to simulate the shrub canopy spectrum and using its red edge band was an effective simulation method.（2）The correlation between the simulated canopy red-edge spectral indexes based on the PROSAIL model and the dust deposition capacity of shrubs was significant.Among all spectral indexes,NDVI had the highest correlation coefficient（-0.843）,followed by OSAVI（-0.829）and RERI₇₄₀（-0.829）,while CI_green and CI_rededge showed lower correlations.TCARI and RERI₇₀₅ were significantly positively correlated with the dust deposition capacity of shrubs,and OSAVI,RERI₇₄₀,NDVI,NDVI_RE were significantly negatively correlated with the dust deposition capacity of shrubs.These six spectral indexes can be used as sensitive spectral vegetation indexes to establish an estimation model for the dust deposition capacity of shrubs.（3）The multiple stepwise regression model had a better effect on the estimation and prediction of the dust deposition capacity of urban shrubs.Under the single regression models,the four spectral indexes of TCARI,OSAVI,RERI₇₄₀ and NDVI had high accuracy,and the estimation model constructed by univariate cubic regression had the best estimation effect,and the highest coefficient of determination（R²）can reach 0.7934.Under the multiple regression models,the introduction of independent variable can be realized and optimized based on multiple stepwise regression,and the fitting degree was great.The R² of the dust deposition capacity estimation model based on the variable combination of NDVI,OSAVI and RERI₇₄₀ can reach 0.851.（4）The shrub information can be extracted by using the method of spectral mixture analysis and the Support Vector Machine classifier.At the same time,the combined application of the PROSAIL radiative transfer model and the multivariate stepwise regression model can achieve the visualization of the spatial distribution of dust deposition of shrubs.Observing the spatial distribution results,the dust deposition capacity of shrubs generally showed a trend of high in the central urban area and low in the suburban area.The dust deposition capacity of shrubs in the northern and central-southern parts of Xuhui District was relatively high and concentrated.The high dust deposition capacity of Minhang District was distributed in strips,and the dust deposition capacity of shrubs in the south parts was higher than that in others.