Atmospheric particulate pollution has become one of the serious urban environmental problems on a global scale. In order to explore the influence of leaf surface dust on the spectral characteristics of the leaves, a prediction model of leaf surface dross based on high-spectrum data was established. In order to explore the influence of leaf surface dust on the spectral characteristics of the leaves, a prediction model of leaf surface dross based on high-spectrum data was established. Our study focused on the common greening tree species (Euonymus japonicus) in Beijing, and collected 720 leaf samples in high, medium and low dust pollution gradient environments, and then used the ASDF Fildsoec Handheld spectrometer to obtain hyperspectral data. The results showed that the spectral reflection peaks were at 560 and 900 nm, respectively, and the absorption valleys were in the range of 400~500, 600~700 and 1 000~1 050 nm. The leaf reflectivity with or without dust retention showed different rules in different bands. The spectral reflectances in the range of 400~760 and 760~1 100 nm were as follows: dust-retaining leaves>dust-removing leaves, dust-retaining leaves