The invention relates to a leaf surface dust fall quantity testing method based on a hyperspectral technique. The method comprises the following steps: firstly, collecting healthy leaves with different leaf surface dust fall quantities, quickly testing spectral information of a single leaf, putting the leaf subjected to the spectral information testing into a room, acquiring leaf surface dust fall quantity data of the leaf by virtue of a leaf area instrument and an electronic scale, and determining a sensitive spectral band subjected to leaf surface dust fall by analyzing the correlation between hyperspectral information and the leaf surface dust fall quantity data; and carrying out modeling by virtue of the data of the sensitive spectral band subjected to the leaf surface dust fall, selecting a model with a minimal root-mean-square error and maximal errors between a determination coefficient and a predicted root-mean-square error and between a sample standard deviation and the predicted root-mean-square error, and predicting the leaf surface dust fall quantity of the model only by virtue of the hyperspectral information of the leaf. Compared with a traditional determination method, the leaf surface dust fall quantity testing method has the beneficial effects of reducing the tedious experimental steps of indoor leaf area testing, cleaning, weighing and the like, being simple, convenient and rapid, and meanwhile, providing references for monitoring of the sand storm strength and environment quality of a dust fall region by virtue of astronautic hyperspectral remote sensing.