Retrieving Dustfall Distribution Based On Ground Spectral Data

Along with the development of the city,people pay more and more attention to the urban environment.The Chinese government has also introduced corresponding policies and increased environmental protection investment,aiming to give residents good air quality.Dust is an important factor in image air quality.When more dust is collected in the air,it will not only affect traffic and travel,but also seriously affect human health.Urban green space can absorb dust and purify air.Among them,plant leaves can retain and adsorb atmospheric particulate matter,characterize air quality,and research on the dust-retaining ability of vegetation can be used for a wide range of environmental monitoring.As a monitoring site data,an important supplementary monitoring method.In this paper,the canopy spectra of common green vegetations in Shanghai,such as Viburnum odoratissimum Ker-Gawl,Photinia serrulata Lindl.,Euonymus japonicus cv.Aureo-ma,Osmanthus fragrans Lour.,Camellia japomica L.,Aucuba japonica Thumb.,Ligustrum vicaryi Rehder.Analyzing the difference of dust between different tree species and the difference of dust in different functional areas,and comparing the influence of dander on the spectral characteristics of plants,select the spectral band and spectral index sensitive to the dust-retaining ability,and adopt the least squares method and neural network.Establish an estimation model between the dust retention capacity and the plant reflectance spectrum.At the same time,the spectrum is resampled into equivalent satellite wide-band data,and the sensitive band and the correlation spectral index are selected to construct the dust-reversing inversion model.The dust of the sampling point and the appropriate image data are used to invert the space of urban dust.Distribution,which lays the foundation for accurate detection of air quality on a large scale.The study found that the canopy spectrum is different before and after the dust retention,mainly in the visible and near-infrared range,and the spectral reflectance decreases after the dust retention.At the same time,the spectral reflectance of canopy is correlated with the amount of dust in the canopy.It is positively correlated between 400 and 700 nm,and negatively correlated with 710 to 1 110 nm.The correlation between 450 and 527,568 to 711 nm is achieved.Significant level.The 749 nm band in the near-infrared region has the highest correlation,and the reflectance in the green region is not sensitive to the effect of leaf dust retention.Moreover,by calculating the correlation between the normalized index,the ratio index,the difference index and the canopy dust content,the correlation between single-band reflectivity and dust retention is improved.Among them,the normalized index at the 395 and 396 nm bands has the highest correlation with the dust retention,and the correlation coefficient is 0.8080.Among the two models constructed,the neural network-based model has higher estimation accuracy.Under the same data conditions,the model prediction coefficient based on partial least squares regression(PLSR)is 0.8291,and the prediction coefficient based on neural network construction model reaches 0.8791.The latter is better.The vegetation spectral data was resampled according to the satellite broad-band spectrum,and the differences before and after the dysphagia were analyzed.It was found that the band range of the same high-spectrum difference was the same,both in the visible and near-infrared bands,and the spectral reflectance decreased after the dust retention.Compared with the Landsat 8 image,the differences between the Tian Gong 2 and the Landsat 8 images are significant,especially in the 9~14 band of Tian Gong 2,corresponding to the 4~5 band of Landsat 8.The equivalent broad-band spectral reflectance is correlated with the amount of dust retention,and is positively correlated across the entire band.The maximum correlation band of Landsat 8 is band 4,the corresponding interval is 630nm~680nm,and the maximum correlation band of Tian Gong 2 is band 7.The corresponding interval is 655nm~675nm,and the correlation coefficients are 0.5896 and 0.5902 respectively.At the same time,the correlation coefficient of the band reflectivity normalization index,the ratio index,the difference index and the dust retention amount is calculated.Among them,the ratio index of the two equivalent satellites has the highest correlation,reaching 0.6742 and 0.6749 respectively.Under the same data conditions,the model estimation based on partial least squares regression of Landsat 8 is higher than that of neural network model.The model estimation based on partial least squares regression of Tian Gong 2 is lower than that of neural network model.The same tree species have different air-difference conditions in different environments.Analysis of the dust-retaining ability of tree species at different sampling points can indirectly reflect the environmental pollution situation in the sampling area.The dust distribution map based on the data of the dust retention capacity of the sample points can reflect the spatial distribution characteristics of dust in Xu Hui District and Min Hang District.Based on the spatial resolution and model fitting effect,Landsat 8 data is used to invert the spatial distribution of dust using partial least squares method to characterize the distribution of dust in the study area.The spatial continuity and the distribution of the dust collection by sampling points are obtained.In comparison,the effect is better.