Estimation of atmospheric dust deposition on plant leaves based on spectral features

Urban atmospheric dust is a significant problem and becoming a considerable pollution source in many cities. This study was based on a comparison of spectral reflectance on the surfaces of dusty and clean leaves. A significant linear relationship (r = 0.811) correlation between the dust weight and near-infrared band region (700-1000 nm) was found through analysis of the spectral data. This relationship obtained from near-infrared band regions, based on the main effects and cluster and interval analysis, was more distinct and stable than that of blue, green, red, and middle-infrared band regions. Thus, the use of near-infrared band data is a reliable method to estimate the amount of dust deposition on plant leaves. A regression model (R2= 64.3%) was constructed based on dust deposition on plant leaves and a near-infrared ratio. The model proved to be accurate as regards an estimation of dust weight, based on a comparison of residuals (normal distribution) and accuracy tests (slope = 0.8437). This model could provide a methodological basis for spatial dust distribution analysis and has the potential for evaluating air pollution levels.